Abstract
Breast Cancer (BC) remains a leading cause of morbidity and mortality among women globally, accounting for 30% of all new cancer cases (with approximately 44,000 women dying), according to recent American Cancer Society reports. Therefore, accurate BC screening, diagnosis, and classification are crucial for timely interventions and improved patient outcomes. The main goal of this paper is to provide a comprehensive review of the latest advancements in BC detection, focusing on diagnostic BC imaging, Artificial Intelligence (AI) driven analysis, and health disparity considerations. We first examine diverse imaging techniques such as Mammography, Ultrasound, and Dynamic Contrast-Enhanced Magnetic Resonance Imaging, and provide an overview of their pros and cons. Then, we provided an intensive review of the State-of-the-Art (SOTA) literature on the role of AI in BC classification and segmentation. Lastly, we examined the role of AI in BC health disparities. A key contribution of this work lies in its integrative approach, consolidating insights from multiple research areas, imaging methods, AI-driven methodologies, and health disparities in a single resource. This paper evaluates the effectiveness of modern AI-based tools in enhancing diagnostic accuracy and discusses their potential to address biases in BC diagnosis, thus promoting equitable healthcare access. By integrating clinical, technical, and equity perspectives, this review aims to inform real-world decision-making, supporting the development of bias-aware AI tools, guiding equitable screening policy, and enhancing clinical practice in breast cancer care. Additionally, our critical analysis and discussion of recent SOTA highlights the strengths, limitations, and knowledge gaps for future directions of AI roles in BC. In total, these findings and future venue suggestions serve as a practical reference for researchers, clinicians, and policymakers, underscoring the need for interdisciplinary collaboration to harness AI’s full potential in BC diagnosis and reduce global health disparities.
Keywords: Breast Cancer, Multi-modal, Health Disparity, Mammograms
1. Introduction
Breast Cancer (BC) is the most common cancer in women, with more than one million cases detected globally, resulting in 600,000 deaths annually [1, 2]. According to the latest estimates from the American Cancer Society in 2023, BC accounts for 30% of all new cancer cases in the United States, and approximately 44,000 women die from BC annually [3]. As a pervasive and heterogeneous cancer type, the correct identification of BC stage is essential to inform targeted treatments and improve outcomes. Early detection through regular screenings, combined with increased awareness and improved therapies, has resulted in a 43% decline in BC mortality as of 2020 [4]. Despite these gains, existing screening practices are far from perfect. Results from screening recalls approximately 10% of women for further imaging, and over 50% of the subsequent biopsies reveal benign results. [5–7]. Therefore, there is an opportunity to improve the accuracy while reducing unnecessary interventions.
Clinicians typically diagnose BC using image-guided biopsy, an invasive procedure prone to sampling errors and false negatives. [8]. As a result, there is a growing interest and ongoing research in non-invasive diagnostic alternatives, particularly imaging modalities such as Mammography, Ultrasound (US), and Magnetic Resonance Imaging (MRI) [9, 10]. The BC screening workflow is guided by the Breast Imaging Reporting and Data System (BI-RADS) [11, 12], which provides a standardized framework for escalating patients from screening to diagnostic imaging and, if necessary, to biopsy. Biopsy and surgical resection results play a critical role in diagnosing BC, as they determine lesion characteristics and guide treatment decisions and patient management [13]. However, the relatively low cancer detection rate in screening poses challenges. Out of every 1000 women screened, clinicians recall about 100 for diagnostic imaging, perform biopsies on 30–40, and ultimately diagnose only 5–8 with cancer [14].
Screening compliance is also an ongoing challenge, with lower compliance among Black and minority women contributing to delayed detection and poorer outcomes [15–17]. Additionally, Social Determinants of Health (SDOH), along with breast density, genetic predispositions, and personal or family history, further complicate risk stratification and call for more personalized approaches. These concerns have heightened efforts to identify high-risk BC patients more accurately who would benefit from more invasive interventions rather than subjecting all patients to aggressive approaches like chemotherapy, which can have severe side effects and long-term health consequences [18]. Thus, there is an urgent need for advanced, comprehensive AI-based screening and diagnostic tools that integrate breast imaging with clinical and demographic data to provide objective and unbiased personalized results effectively, thereby increasing accuracy and streamlining biopsy referrals.
BC screening is undergoing a rapid transformation because of the integration of AI, particularly through trials that assess AI as a second reader for radiologists. According to recent research, AI has the potential to improve the detection rate by detecting subtle lesions [19], especially in dense breast tissue, while also reducing false negatives. In Europe and the U.S., clinical trials have shown that AI-assisted reading enhances diagnostic accuracy and may reduce the workload of radiologists, thereby addressing workforce shortages [20–22]. These AI systems incorporate an additional layer of analysis to enhance cancer detection without significantly increasing recall rates [23].
AI–driven tools, especially Computer-Assisted Diagnosis (AI-CAD), have emerged as promising solutions to enhance breast imaging interpretation and streamline diagnostic pathways. Lee et al. demonstrated that AI-CAD in ultrasound significantly improved specificity, positive predictive value, and accuracy across experience levels [24, 25]. Likewise, deep learning–based CAD systems offer early, non-invasive lesion detection and support clinical risk stratification, potentially reducing unnecessary biopsies [26, 27]. Furthermore, researchers have widely reviewed the use of AI and imaging in BC diagnosis; however, the issue of health disparities, particularly in terms of race, ethnicity, socioeconomic status, and geographic access, has received limited attention. Most prior reviews either omit equity considerations entirely or treat them peripherally.
Given the growing awareness of algorithmic bias and unequal screening outcomes in the underserved population, this oversight represents a critical literature gap. At the same time, growing interest from clinical trials and regulatory bodies in AI-assisted screening tools underscores the urgency of evaluating both current progress and future directions. Unlike previous reviews that solely focused on imaging techniques (e.g., [28–30]), AI applications for BC classification and segmentation (e.g., [31–34]), or health disparities in screening and diagnosis (e.g., [35–38]), our review uniquely combines all three domains into a unified framework to reflect the complexity of real-world clinical workflows. It bridges technical, clinical, and socio-ethical dimensions of BC diagnosis. To our knowledge, no existing review offers such an integrative perspective, offering actionable insights that may inform clinical decision-making, support equitable screening policies, and guide AI model developers in designing bias-aware diagnostic systems.
Importantly, the timing of this review is significant as the 2020–2025 period represents a turning point for AI in BC, marked by several converging advances: the rise of Vision Transformers (ViTs), the fusion of multimodal datasets (imaging, histology, genomics, clinical records), and the advent of privacy-preserving techniques like Federated Learning (FL). At the same time, Large Language Models (LLMs) are being adapted for radiology reporting and screening triage, while regulatory interest in AI deployment in screening workflows is accelerating. Taken together, these trends underscore the urgency of synthesizing current progress to guide both implementation and future research. This review is therefore both timely and necessary. While primarily a literature review, this paper also serves as a strategic guide for aligning AI development in BC imaging with equity-driven clinical priorities.
This paper presents its content in seven structured sections. First, we cover the imaging techniques widely employed in BC diagnosis in section 2. Also detailed in this section are the publicly available datasets for BC research. Next, section 3 provides a comprehensive overview of related literature and AI-based methodologies developed for BC imaging research. It includes a description of the most recent SOTA research findings, along with a discussion of the literature covering segmentation and classification tasks, including how researchers evaluate such systems. Section 5 focuses on health disparities in the BC imaging context and AI’s role in mitigating such biases. This section covers factors contributing to disparities in BC screening and diagnosis, and how the adoption of AI offers promising solutions to address these disparities. Section 6 summarizes the key findings and implications of recent literature, reviewing the strengths and weaknesses of the SOTA. Further, the section also focuses on identifying gaps and opportunities for future innovations and research directions. Section 7 outlines emerging trends and strategic research priorities that researchers are actively developing to shape the next generation of AI-driven BC diagnostics. It builds on prior analysis to propose actionable pathways for addressing current limitations, enhancing clinical integration, and ensuring equitable and trustworthy innovation. Finally, we review our conclusions in section 8.
2. Imaging Techniques in BC Diagnosis
BC screening and diagnosis rely on various imaging techniques, including Mammography, Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI), and US. These imaging modalities help detect and characterize abnormalities in breast tissue, aiding in early detection and treatment planning. Average-risk women in the U.S. undergo screening mammography annually, depending on age. High-risk women may additionally receive screening with US or MRI [39–41]. Figure 1 shows the typical clinical workflow for BC screening, diagnosis, and biopsy decisions based on BI-RADS assessments and symptom presentation.
Figure 1:
Schematic illustration of the overall screening process from initial annual screening until final biopsy diagnosis.
2.1. Mammography
Mammography remains the cornerstone of BC screening, widely adopted for its proven impact on early detection and mortality reduction [42, 43]. Techniques such as 3D Digital Breast Tomosynthesis (DBT) and Contrast-Enhanced Mammography (CEM) offer enhanced lesion visualization through multi-angle imaging and contrast agents, respectively [44, 45]. While mammography remains vital for early BC detection and mortality reduction, it has limitations, including high levels of false positives leading to unnecessary follow-ups, patient anxiety, and reduced accuracy in dense breasts, often requiring supplementary imaging [43, 46, 47]. Deep learning (DL) approaches, particularly Convolutional Neural Networks (CNNs), have demonstrated strong potential to augment mammogram interpretation, improve diagnostic precision, and mitigate some of these limitations in both retrospective and prospective settings [48]
2.2. Contrast-Enhanced Mammography (CEM)
CEM combines dual-energy X-ray imaging with an intravenous iodinated contrast agent to provide both structural and functional breast imaging [49, 50]. CEM offers diagnostic accuracy and sensitivity comparable to DCE-MRI at a lower cost, making it a viable option in rural and resource-limited settings [51, 52]. Additional advantages include shorter exam times, improved patient comfort relative to MRI, and versatility across screening, staging, lesion characterization, and treatment monitoring [53, 54]. However, challenges such as increased radiation exposure, limited standardization, and implementation barriers limit wider adoption [55]. Figure 2 (a), (b), and (c) show examples of CEM images.
Figure 2:
Examples of breast images (a) a contrast-enhanced mammogram in Cranio-Caudal (CC) view that shows signs of BC from low energy contrast; and T1-weighted MRI images of (b) a special type carcinoma and (c) ductal carcinoma.
2.3. Ultrasound (US)
In addition to mammography, US aids in diagnosing BC by assessing lesion shape, acoustic features (shadowing, posterior acoustic enhancement), and blood flow [56]. Its high sensitivity, accessibility, cost-effectiveness, and real-time imaging make it particularly valuable for dense breast tissue [57, 58]. However, US has limitations, including operator dependence and challenges in detecting micro-calcifications, which indicate early-stage BC [59, 60]. In some cases, image quality issues like noise and low contrast further hinder accurate diagnosis through US images [29, 61]. Advanced computer-aided systems and multi-modality approaches, such as the integration of US with MRI, improve diagnostic accuracy, sensitivity, and specificity, and offer a more reliable assessment of breast lesions [62].
2.4. Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI)
Dynamic Contrast-Enhanced MRI (DCE-MRI) plays a critical role in BC diagnosis, breast conservation therapy planning, radiotherapy guidance, and treatment planning [63–67]. DCE-MRI effectively identifies cancerous lesions and is highly sensitive to breast carcinoma, including Nonmass Enhancement (NME) lesions, which are uniquely visible on MRI and can indicate malignancy [66]. Image analysis systems extract texture features from DCE-MRI, which serve as virtual prognostic biomarkers, aiding in the prediction of clinical parameters. Additionally, Background Parenchymal Enhancement (BPE) observed in DCE-MRI is an emerging marker of BC risk, with higher BPE levels correlated with an increased risk of initial BC diagnosis and recurrence [68, 69]. While DCE-MRI offers higher sensitivity than US and similar efficacy to CEM, it carries a higher risk of false positives, leading to over-treatment [70, 71]. Figure 2(b,c) shows examples of MRI images.
2.5. Public Datasets for BC Research
To support the development and validation of imaging-based AI models, publicly available datasets have played a critical role by offering standardized benchmarks and enabling reproducible research across modalities, histopathology, and tabular data. Mammography datasets such as EMBED, CBIS-DDSM, and NYU BCSD offer annotated images that have enabled the development and benchmarking of DL models. In particular, CBIS-DDSM provides lesion-level annotations and curated diagnostic labels, which have supported numerous CNN-based architectures for tasks such as mass and calcification classification [72, 73]. Researchers commonly use smaller but high-quality collections like INBreast and MIAS, which focus on expert-annotated lesions, to evaluate segmentation and cross-domain generalization performance [74]. In total, publicly accessible mammogram datasets (such as EMBED, CBIS-DDSM, MIAS, NYU BCSD, and INBreast) have played a foundational role in advancing DL for BC diagnosis by reducing data access barriers, supporting model reproducibility, and enabling the development of AI-driven tools with improved accuracy, interpretability, and generalizability for early detection and personalized treatment.
Furthermore, US datasets, including BUSI, UDIAT, and BLUI, facilitate non-invasive detection through pixel-level lesion annotations, serving as crucial resources for training and validating AI models in low-resource settings. Histopathology datasets such as BreakHis, BCSS, and BRACS provide microscopic tumor classifications and tissue segmentation labels, enabling fine-grained learning of morphological features. Researchers have explored alternative imaging modalities using thermography and tomography datasets like Visual Lab DMRIR and EIT, which support multimodal research. Meanwhile, tabular datasets such as WBCD, SEER, and BCCD contain clinical, demographic, and biomarker information that contribute to predictive modeling and outcome estimation. Table 1 provides a list of publicly available datasets for BC detection via various modalities.
Table 1:
A summary of the publicly available datasets for BC imaging modalities.
| Dataset | # Subjects/ Instances (#Classes) | Description |
|---|---|---|
| EMBED [75] | 110,000 (Two: Normal, Cancer) | A mammography dataset containing 60,000 annotated lesions spanning 7 years, encompassing approximately 3.4 million 2D and 3D images, with equal representation of Black and White women. |
| CBIS-DDSM [76] | 1,600 (Two: Benign, Malignant) | A mammography dataset of breast mammograms, encompassing approximately 10,200 2D images designed for training DL models for BC detection. |
| INBreast [77] | 115 (Two: Benign, Malignant) | A mammography dataset based on FFDMs, encompassing 410 2D images with expert annotations for masses, calcifications, and other lesions, used in BC detection studies. |
| MIAS [78] | 161 (Three: Normal, Benign, Malignant) | A mammography dataset based on digital films that are approximately 3,800 2D images, including normal, benign, and malignant images with labels for abnormality class, severity, and mass center coordinates. |
| Mini-MIAS [79] | 161 (Two: Benign, Malignant) | A small dataset containing 2D images of detailed radiologist annotations from mammograms downscaled from the original MIAS database based on current abnormalities present. |
| BCDR [80] | 1,734 (Three: Normal, Benign, Malignant) | A mammography dataset containing approximately 1,700 2D images of Mammograms from various sources used for training CAD and other AI models. |
| CMMD [81] | (Two: Normal, Cancer) | A mammography dataset containing approximately 3,700 2D images of mammograms and medical records from Chinese hospitals. |
| NYU BCSD [82] | 141,473 (Three: Normal, Benign, Malignant) | A mammography dataset containing a large collection of high-resolution mammograms, encompassing approximately 1 million 2D images based on four mammographic views: Left-CC, Right-CC, Left-MLO, and Right-MLO. |
| DDSM [83] | 2,620 (Three: Normal, Benign, Malignant) | A mammography dataset containing approximately 10,000 2D images of pixel-level annotations for ROIs, labeled according to BI-RADS categories. |
| Mini-DDSM [84] | 1,500 (Three: Normal, Benign, Malignant) | A mammography dataset containing approximately 4,800 2D images, which researchers reduced to only the essential images with fewer detailed annotations based on the full DDSM. |
| Magic-5 [85] | 345 (Three: Benign, Malignant, Normal) | A mammography database having approximately 1,200 2D images classified according to lesion type and morphology, breast tissue, and pathology type with a dedicated graphical user interface. |
| BancoWeb LAPIMO Database [86] | 320 (Three: Normal, Benign, Malignant) | A diverse mammography database of 2 containing approximately 1,000 2D images from multiple hospitals, along with medical reports and clinical information. |
| OPTIMAM Medical Image Database [87] | 170,000 (Four: Normal, Benign, Cancerous, Interval) | A mammography database of approximately 2.5 million 2D de-identified images and associated clinical data collected from multiple UK breast screening centers. |
| BUSI / Baheya / Data in Brief [88] | 600 (Three: Normal, Benign, Malignant) | A US dataset containing 780 2D images used in BC diagnosis, including annotations for all cases. |
| UDIAT [89] | 163 (Two: Benign, Malignant) | A small US dataset consisting of 2D images with benign lesions and cancerous lesions of DCIS, lobular carcinoma, fibroadenomas, and cysts. |
| BLUI [90] | 256 (Two: Benign, Malignant) | A US dataset consisting of 500 2D images based on manually annotated lesions, with tumor and image-level details including BIRADS categories and descriptors, and supplemented with patient demographic and clinical data. |
| ENCI [91] | 60 (Seven: ER, PR, HERS2, Ki-67, TNBC, Lymph Node, Menopausal) | A small US dataset that includes 2D images with patient demographics, tumor characteristics, and treatment data. |
| BreakHis [92] | 82 (Two: Benign, Malignant) | A histopathology dataset containing approximately 7,000 2D microscopic images of breast tumor tissue, captured at varying magnification levels. |
| Duke [93] | 922 (Various: Two Instances, Four Instances Depending on Data Type) | A small histopathology dataset comprising 3D DCE-MRI from BC patients, annotated with detailed clinical, imaging, and pathological features. |
| BCSS [94] | 20,000 (Five: Tumor, Stroma, Inflammatory, Necrosis, Other) | A histopathology dataset containing approximately 5,000 2D images with annotated tissue regions from BC images, facilitating the development of machine-learning models for detailed tissue segmentation. |
| BreCaHAD [95] | 162 (Six: Mitosis, Apoptosis, Tumor Nuclei, Non-tumor Nuclei, Tubule, Non-tubule) | A small, meticulously annotated histopathology dataset containing 2D images to aid in the development of automated diagnostic systems for detecting and analyzing BC at the microscopic level. |
| Kaggle IDC [96] | 162 (Two: Benign, Malignant) | A small histopathology dataset containing 2D images identified as non-IDC and IDC. |
| BACH [97] | 400 (Four: Normal, Benign, DCIS, IC) | A small histopathology dataset composed of H&E stained breast histology microscopy and WSIs. |
| BRACS [98] | 187 (Seven: Normal, Benign, UDH, FEA, ADH, DCIS, IC) | A histopathology dataset containing a collection of approximately 5,000 2D H&E-stained histopathological images, aimed at facilitating automated BC detection and classification. |
| AIDPATH [99] | 50 (Five: HE, ER, PR, Ki-67, HER2) | A histopathology dataset includes WSIs and 120 TMAs stained with biomarkers. |
| Visual Lab / DMRIR [100, 101] | 56 (Three: Normal, Benign, Malignant) | A thermography dataset contains a collection of 400 2D thermal breast images (thermograms) used for BC screening |
| EIT [102] | 64 (Three: Normal, Benign, Malignant) | A tomography dataset that includes 348 2D EIT images from breast tissues, both healthy and cancerous, collected from patients undergoing BC screening or diagnostic imaging. |
| WBCD [103] | 699 (Eight: Group 1 through 8) | A tabular dataset featuring attributes like clump thickness and uniformity of cell size to help distinguish between benign and malignant breast tumors. |
| WDBC [104] | 569 (Thirty-Two Attributes) | A tabular dataset with attributes from digitized breast mass biopsy images, used to classify instances as benign or malignant based on cell nucleus characteristics. |
| WPBC [105] | 198 (Thirty-Four Attributes) | A tabular dataset having diagnostic and prognostic information derived from image-processed breast mass features to help in the prediction of BC recurrence, with attributes such as tumor thickness, cell shape uniformity, and mitotic count playing crucial roles in prognostic assessments. |
| UHCC [106] | 116 (Two: Benign, Cancerous) | A tabular dataset containing nine predictive variables, which include anthropometric and routine blood analysis parameters like age, BMI, glucose, insulin, HOMA, leptin, adiponectin, resistin, and MCP-1 levels. The target variable indicates whether a patient has BC or not (binary classification). |
| BBCD [107] | 10,114 (Six: circRNA, lncRNA, mRNA, miRNA, piRNA, tRFRNA) | A tabular dataset containing blood-derived RNA biomarkers for the early detection of various cancers, including BC. |
| SEER [108] | 4,024 (Four: HR+/HER2−, HR−/HER2−, HR+/HER2+, HR−/HER2+) | A tabular dataset containing a collection of cancer incidence and survival data from population-based cancer registries, providing detailed information on BC subtypes categorized by HR- and HER2-positive/negative status based on demographic, diagnostic, and treatment variables. |
ADH: Atypical Ductal Hyperplasia; AIDPATH: Annotated Images Dataset for Pathology; BACH: Breast Cancer Histology Challenge; BBCD: Blood-based Biomarkers for Cancer; BCDR: Breast Cancer Digital Repository; BCSS: Breast Cancer Segmentation Challenge; BLUI: Breast Cancer Histopathological Annotation Dataset; BMI: Body Mass Index; BreCaHAD: Breast Cancer Histopathological Annotation Dataset; BRACS: Breast Cancer Radiology and Screening; BUSI: Breast Ultrasound Images Dataset; CBIS-DDSM: Curated Breast Imaging Subset of DDSM; CC: Cranio-Caudal; circRNA: Circular RNA; CMMD: Chinese Mammography Database; DCE-MRI: Dynamic Contrast-Enhanced Magnetic Resonance Imaging; DCIS: Ductal Carcinoma In Situ; DDSM: Digital Database for Screening Mammography; DMRIR: Database for Mastology Research with Infrared Image; EIT: Electrical Impedance Tomography; EMBED: Emory Breast Imaging Dataset; ENCI: Egyptian National Cancer Institute; ER: Estrogen Receptor; FEA: Flat Epithelial Atypia; FFDMs: Full-Field Digital Mammograms; H&E: Hematoxylin and Eosin Stain; HER2: Human Epidermal Growth Factor Receptor 2; HOMA: Homeostasis Model Assessment; HR: Hormone Receptor; IC: Invasive Carcinoma; IDC: Inductive Directive Carcinoma; Ki-67: Antigen Kiel 67; LAPIMO: Laboratório de Análise e Processamento de Imagens Médicas e Odontológicas (English: Laboratory of Analysis and Processing of Medical and Dental Images); lncRNA: Long Non-coding RNA; mRNA: Messenger RNA; MCP-1: Monocyte Chemoattractant Protein-1; MIAS: Mammographic Image Analysis Society; miRNA: Micro RNA; MLO: Medio-Lateral Oblique; NYU BCSD: New York University Breast Cancer Screening Dataset; OPTIMAM: Optimized Mammography Database; piRNA: Piwi-interacting RNAs; PR: Progesterone Receptor; RNA: Ribonucleic Acid; ROIs: Region of Interests; SEER: Surveillance, Epidemiology, and End Results; TMAs: Tissue Microarrays; TNBC: Triple-Negative Breast Cancer; tRFRNA: Transfer RNA Derived Fragments; UDH: Usual Ductal Hyperplasia; UDIAT: Unitat de Diagnòstic per la Imatge de l’Alt Taulí (English: Diagnostic Imaging Unit of the Alt Taulí); UHCC: University Hospital Cologne Cancer; UK: United Kingdom; WBCD: Breast Cancer Wisconsin Original; WDBC: Breast Cancer Wisconsin Diagnostic; WPBC: Breast Cancer Wisconsin Prognosis; WSIs: Whole-Slide Images.
3. Role of Artificial Intelligence in BC
There are many challenges associated with BC detection and prognosis, including under-and over-detection, tissue sampling, and treatment. In the U.S., 40 million undergo screening mammography annually, which requires an extremely large workforce of radiologists to interpret these studies; however, this is an ongoing and predicted worsening shortage of radiologists to interpret these exams, and these challenges are even more severe in international settings [109]. To address these challenges, AI-CAD systems have become a focal point of ongoing research [34]. Initial results are promising: for instance, when radiologists employed AI support, their cancer detection rates increased by 13.8%, according to a multi-center study [110]. The Explainable-AI (XAI) model outperformed certain radiologists in the classification of breast lesions, achieving 94.34% accuracy [111]. These findings reflect the growing value of AI tools in augmenting human expertise and improving diagnostic accuracy. To trace the progression of AI applications in BC imaging, we now explore key developments across traditional CAD, classification algorithms, segmentation techniques, and fusion approaches.
3.1. Traditional Computer-Aided Diagnosis (CAD)
Traditional CAD systems laid the foundation for today’s AI-driven models in BC screening and diagnosis by assisting radiologists in lesion detection and minimizing human error [112, 113]. These systems, which evaluate imaging modalities like mammography and MRI, serve as an additional layer of scrutiny to enhance detection rates, especially in recognizing small anomalies in dense breast tissue [114, 115]. Early CAD approaches relied heavily on rule-based logic and handcrafted feature extraction, using explicitly programmed criteria derived from expert knowledge. These approaches apply techniques such as thresholding, region growing, and morphological filtering to identify calcification or masses, often mimicking radiologist reasoning [116, 117]. One of the earliest examples is the system by Giger et al. in the late 1980s that used morphological and template-matching filters to detect microcalcifications in digitized mammograms [118]. These deterministic approaches laid the groundwork for the formalization of lesion detection and classification as computational tasks.
In the 1990s, CAD systems evolved with the adoption of statistical learning techniques, including Linear Discriminant Analysis and Support Vector Machines (SVMs). Researchers have introduced feature descriptors such as Local Binary Patterns (LBP) and Histogram of Oriented Gradients (HOG) to improve sensitivity and specificity [119, 120]. Publicly available datasets like the DDSM, introduced in 1997, played a pivotal role in training and benchmarking these systems [121]. These developments paved the way for the emergence of DL–based AI models, which fully automate feature extraction [122, 123]. Despite the greater accuracy of AI models like CNNs, classical CAD continues to be pertinent, particularly in low-resource environments where sophisticated AI is not yet available [124]. They have also contributed to standardizing radiological interpretations and reducing inter-reader variability, thereby improving diagnostic consistency. As AI technologies evolve, the integration of XAI techniques into CAD workflows will be essential to improve trust, transparency, and decision support in clinical settings [125].
Building on these traditional foundations, the rapid advancement of AI methodologies has led to more sophisticated systems, from early ML models like SVMs and Logistic Regression (LR) to complex DL architectures, hybrid models, and emerging frameworks such as LLMs [126, 127] and consequently Vision Language Models (VLMs) [128, 129]. These models offer new ways to interpret and synthesize imaging and textual data, expanding AI’s role in BC care. Table 2 presents a detailed comparison of these approaches, highlighting their advantages, disadvantages, and role in improving BC imaging and diagnosis.
Table 2:
Comparison of various AI/ML Methods for BC Diagnosis.
| References | Advantages | Disadvantages | |
|---|---|---|---|
| Traditional ML Methods | Examples include: SVM [130, 131] KNN [132, 133] LR [132, 134, 135] RF [134, 136] K Means Clustering [137–139] |
|
|
| Modern ML Methods | Examples include: CNN [140–145] ViT [146–148] Transfer Learning [149] Segmentation-based [150, 151] |
|
|
| Hybrid Methods | Examples include: [152–154] |
|
|
| New Trends | LLMs [155–157] |
|
|
KNN: K-Nearest Neighbor; LR: Logistic Regression; RF: Random Forest; SVM: Support Vector Machine; CNN: Convolutional Neural Network; ViT: Vision Transformer; LLM: Large Language Models;
Researchers and practitioners use two core methodologies in AI applications for BC: classification and segmentation. They apply classification algorithms to determine the likelihood that a lesion, image, or exam contains cancer, while they use segmentation algorithms to outline tumor boundaries in imaging data. These processes can complement each other by facilitating early detection, enhancing diagnostic accuracy, and supporting therapy planning. The following subsections explore both classification and segmentation methods and their commonly used performance evaluation metrics in detail, beginning with traditional ML and modern DL applications, followed by a discussion of recent advances, including the integration of transformer models to further optimize AI performance in BC screening and diagnosis, and multimodal fusion methods.
3.2. Classification Methods
Classification methods focus on categorizing breast lesions as benign or malignant by analyzing imaging data, biopsy results, and clinical features [158]. The integration of AI in BC screening and diagnosis has evolved significantly, with both traditional ML and modern DL methods playing crucial roles. Recent studies have highlighted the effectiveness of these approaches in enhancing screening and diagnostic accuracy and efficiency [33, 159, 160].
3.2.1. Traditional Machine Learning Approaches
Traditional models such as LR, SVM, K-nearest neighbors (KNN), Decision Trees (DT), and ensemble learners like Random Forest (RF) or XGBoost (XGB) have served as a foundational framework for early BC classification studies [132–135, 137–139]. These methods typically rely on hand-crafted features, including shape descriptors, texture, or radiomic features, which require labor-intensive feature engineering and may not always capture the most discriminative features of BC. Despite these limitations, LR and RF have achieved high accuracy in distinguishing between benign and malignant tumors [130–134]. In addition, a hybrid ensemble model combining Artificial Neural Networks (ANN) and Deep Belief Network (DBN) characteristics outperformed various traditional methods, achieving even better accuracy [152].
3.2.2. Modern Deep Learning Approaches
While traditional ML methods provide a solid foundation for BC detection, integrating DL techniques offers enhanced accuracy and efficiency, resulting in improved diagnostic and screening tools in clinical settings. [161] As an example of DL methods, CNNs automatically learn complex patterns from medical images, detecting subtle features such as textural changes, micro-calcifications, and architectural distortions with high precision, achieving up to 99.16% accuracy for breast mass classification [162]. Transfer Learning (TL) integration further enhances DL model efficiency by leveraging pre-trained knowledge from large datasets, which enables generalization across various medical centers, and improves the results on smaller BC datasets. TL models, such as DenseNet121, ResNet50, and AlexNet, have achieved accuracies exceeding 98% while effectively addressing the challenges of small sample size datasets [149, 163–167].
3.2.3. Vision Transformers (ViTs)
The adoption of ViTs in BC imaging reflects a significant shift from CNNs toward architectures capable of capturing richer spatial dependencies across medical images. Unlike CNNs, which rely on localized convolutional filters, ViTs employ self-attention mechanisms that allow the model to learn global context and long-range feature interactions. Such capability is especially beneficial in BC diagnosis, where both localized lesions and broader structural patterns (e.g., tissue density, architectural distortion) provide critical diagnostic cues. ViTs divide images into fixed-size patches and process them as sequences with positional embeddings, enabling the capture of both local details (such as small tumors) and global patterns (such as tissue density). This architecture facilitates greater flexibility and scalability in analyzing diverse imaging modalities and resolutions [146–148, 168–172].
Recent studies have validated these advantages across multiple datasets and imaging tasks. For example, a ViT-based TL model achieved state-of-the-art performance in breast mass classification on mammography, confirming the effectiveness of pretrained ViTs in medical imaging scenarios with limited annotated data [173]. Similarly, in 3D breast imaging, a Swin Transformer model outperformed both ResNet101 and vanilla ViTs on a large Digital Breast Tomosynthesis (DBT) dataset comprising about 4,500 scans [174]. Despite this, ViTs can underperform CNNs when trained from scratch on small datasets due to their higher data and computational demands [175].
3.2.4. Hybrid Models
To mitigate the limitations of standalone ViTs and harness the strengths of other architectures, hybrid models combining traditional ML, ViTs, and CNNs have gained traction in BC imaging. These approaches can integrate handcrafted features of ML, local inductive biases, and parameter efficiency of CNNs with the global attention and contextual modeling of ViTs, offering a robust solution for diverse imaging tasks [176]. For instance, a multi-scale hybrid ViT-CNN model demonstrated 97% binary classification accuracy and 93% subtype accuracy on the BreakHis histopathology dataset, outperforming either ViT or CNN alone. Token-mixing strategies used in these hybrids help reduce training time and computational load while improving classification precision [177]. Additionally, hybrid ViT frameworks tailored for mammography have shown significant gains in diagnostic reliability and interpretability, with attention maps highlighting clinically relevant regions for expert review [178]. Integrating histopathological or genetic data with hybrid models may further enable the identification of specific BC subtypes, such as triple-negative BC, which remains a challenging and aggressive form requiring specialized treatment approaches [179].
3.2.5. Evaluation Metrics for Classification
Evaluating the performance of classification tasks in ML is crucial for ensuring model efficacy. Classification models define thorough evaluation measures in terms of True Positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN) rates of test outcomes. Each measure has its strengths and limitations. Accuracy measures overall correct predictions (can be misleading for imbalanced data) [180]. Precision & recall show positive prediction accuracy and ability to detect all positives [181, 182]. Classification models calculate the F1-score as the harmonic mean of precision and recall, which helps balance the two metrics. ROC-AUC is a graphical depiction that measures a model’s ability to differentiate between classes [183, 184]. Specificity & Sensitivity show the percentage of actual negatives/positives correctly identified [185]. Metric choice should typically align with dataset characteristics and clinical requirements.
3.3. Segmentation Methods
While classification focuses on predicting disease presence, segmentation methods are essential for localizing and delineating cancerous regions, providing critical insights into tumor size, shape, and boundaries.
3.3.1. Traditional Segmentation Techniques
Segmentation methods play a critical role in BC screening by detecting cancerous regions in mammograms, US, and MRIs. Traditional methods include region-growing, thresholding, K-means, and fuzzy C-means clustering. These approaches demonstrate varying effectiveness in identifying anomalies, particularly micro-calcifications, and tumor structures [133, 137, 186, 187]. Often, these segmentation methods are employed in manual feature engineering to enhance image quality during imaging pre-processing stages, enabling more effective downstream classification by ML or DL methods [188, 189]. Techniques like multilevel thresholding further enhance performance by categorizing image regions based on pixel intensity, particularly when combined with metaheuristic algorithms [190]. However, these methods often struggle with high-dimensional data and lack generalizability, requiring manual tuning.
3.3.2. Deep Learning-based Segmentation
Researchers increasingly advocate for advanced algorithms to address the limitations of traditional approaches, improve segmentation performance, and reduce human bias in BC detection [32]. Modern methods for segmentation tasks leverage DL architectures, particularly Fully Convolutional Networks (FCNs), U-Net & Variants (U-Net++, U-Net3++), and ViTs, and have achieved SOTA results in medical imaging [162]. FCNs utilize convolutional layers to produce pixel-level segmentation maps, but their limited use of skip connections may result in some spatial information loss [191]. In contrast, U-Net heavily relies on skip connections, preserving spatial information and enhancing segmentation accuracy, particularly for detailed tasks like medical imaging [192].
Studies comparing variations of U-Net, including U-Net++ and U-Net3++, highlight U-Net3++’s superior performance in metrics like surface distance and dice coefficient, though all models demonstrated strong accuracy [151]. Advanced models, such as SegResNet, HighResNet, VNet, and R2U-Net, have also shown high accuracy in segmenting cancerous tissues [193]. Hybrid models such as R2U-Net, integrating residual and recurrent connections, have reported accuracy up to 95.6%. Additionally, researchers have effectively used DeepLabV3+ for tumor segmentation in US images, achieving global accuracy rates of up to 96.5% [194].
3.3.3. Vision Transformers in Segmentation
ViTs, on the other hand, excel at capturing complex tumor boundaries and, in some cases, outperform traditional CNN-based segmentation models like U-Net. For instance, a retrospective study using ViTs achieved a dice coefficient of 91%, compared to 90% for nnUNet in multi-institutional MRI data [171]. A 1% difference (dice score 91% vs. 90%) may not always be clinically significant, given the potential variability in multi-institutional factors such as generalizability, interpretability, and efficiency, which makes ViTs attractive [171]. Hybrid ViT models continue to evolve, offering improved generalization and robustness [153, 195–197].
3.3.4. Evaluation Metrics for Segmentation
Unlike classification, segmentation uses pixel-wise and overlap-based metrics because they aim to delineate the exact boundaries of cancerous regions in breast images (e.g., mammograms, MRIs). The Dice Coefficient (F1-score for segmentation) evaluates the degree to which the ground truth mask and the expected segmentation mask overlap. Intersection over Union (IoU), also called the Jaccard Index, is another overlap-based metric stricter than Dice, as it penalizes false positives more heavily, which can often lead to bias in larger objects. Researchers have proposed fine-grained mIoUs to mitigate this bias, providing a more nuanced evaluation [198]. Pixel accuracy refers to the proportion of correctly classified pixels across the entire image (tumor vs. non-tumor). Although simple, pixel accuracy may not be as informative in cases where the background (non-tumor pixels) dominates the image. As in classification metrics, researchers use precision and recall in segmentation tasks to evaluate how accurately models classify tumor pixels, measuring both the proportion of predicted tumor pixels that are correct and the proportion of actual tumor pixels the model correctly identifies.
Other metrics, such as boundary IoU, focus on the accuracy of the predicted boundary of the segmented object, evaluating how well the predicted and true boundaries match. It is beneficial for medical imaging tasks, like BC detection, where accurate boundary delineation is critical for tumor assessment and surgical planning. The Hausdorff distance calculates the maximum distance between points on the predicted boundary and the ground truth boundary. It helps evaluate segmentation models where boundary precision is critical, like tumor size estimation. Volume Overlap Error (VOE) is another metric that quantifies the error in volume estimation of the predicted region compared to the ground truth. Practitioners often rely on it to assess tumor size estimation in 3D imaging, such as MRI.
3.4. Multimodal Fusion Methods
Multimodal fusion represents a significant advancement in BC research. By integrating diverse data modalities such as imaging, histopathology, genomics, and clinical records, these approaches offer a more comprehensive view of disease characteristics for improved diagnosis, subtype classification, and personalized treatment planning. Initial multimodal approaches combined imaging features with clinical metadata using ensemble ML methods. Early efforts employed models like RFs and SVMs to merge radiologic and clinical data, resulting in improved diagnostic performance compared to single-modality models [199, 200]. With the advent of DL, more sophisticated multimodal fusion strategies have emerged. DL-based methods employ early, joint, and decision-level fusion strategies across modalities such as mammography, histopathology, and clinical data [201].
Cross–attention–based models that integrate histopathology images with genomic profiles have shown superior performance in subtype classification and prognostic prediction [202]. Similarly, pathomic fusion fuses histology and genomic data via attention mechanisms, improving prognostic prediction across multiple cancer types, including BC [203]. Recent innovations in transformer and Graph Neural Networks (GNNs) have shown great potential for multimodal integration. For example, researchers have used DL models that combine histopathology and genomic data to improve survival prediction. In contrast, self-attention–based models integrating histopathology with genetic and clinical data have enhanced survival risk stratification in BC [204, 205].
While multimodal fusion offers significant promise by integrating imaging and molecular data for more personalized BC care, challenges remain in data heterogeneity, modality imbalance, and interpretability. Emerging solutions such as cross-modal transformers and graph-based models are addressing these limitations [206, 207]. By bridging image-based and molecular insights, these approaches offer the potential to support precision oncology and inform patient-specific care strategies more effectively than unimodal systems. Looking ahead, foundation models like LLMs and VLMs aim to unify multimodal inputs at scale, enabling more generalizable and human-aligned AI systems, as discussed in Section 7.
4. Literature Search and Review Methodology
This comparative analysis highlights the breadth of research across traditional ML and modern DL approaches, illustrating their roles in classification, segmentation, and detection tasks. This systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [208] to ensure a rigorous selection process as shown in Figure 3. We conducted a comprehensive search using predefined Boolean queries consistent across databases, including PubMed, IEEE Xplore, Scopus, and Google Scholar, related to AI, BC imaging techniques, and health disparities. Table A.1 in Appendix A lists full search terms and strategies.
Figure 3:
PRISMA flowchart showing the study selection process in this paper.
Our review followed a structured selection process consistent with PRISMA guidelines. The initial search yielded 391 articles. We removed 41 duplicates via Zotero’s duplication and manual verification, leaving 350 records for further screening. Inclusion criteria required studies to (1) present empirical evidence of AI application in BC imaging, (2) focus on at least one imaging modality (mammography, ultrasound, or MRI), or (3) address disparities in screening, diagnosis, or outcomes. Furthermore, we included studies if they were (4) published in English from 2019 to 2024, (5) appeared in peer-reviewed journals, (6) applied clinically relevant or explainable AI methods, and (7) contained sufficient methodological detail. The inclusion excluded preprints, conference papers, and non-English articles.
To enhance completeness and ensure thematic saturation, we supplemented our primary database searches with backward and forward citation tracking using Google Scholar and Scopus. We scanned the reference lists of relevant reviews and primary articles for additional studies. We repeated this process iteratively until no new themes or study types emerged; the absence of new findings signaled that we had reached saturation. Two independent authors screened titles and abstracts for relevance, followed by full-text assessment. A third senior editor resolves any disagreements through discussion or by consulting.
This process yielded 146 eligible full-text articles. We then summarized the selected studies based on their objectives, methods, imaging modalities, year of publication, and relevance. This systematic approach ensures a well-rounded and evidence-based discussion of the advancements and limitations of AI-driven BC research. Tables B.1, B.2, and B.3 from Appendix B provide a complete list of the included articles along with citation metadata, publication year, imaging modality, dataset, AI method, best performing algorithm for imaging modality, performance of each study or range of studies, health disparity focus, AI role for health disparity, equity contribution, and limitations for every domain.
Studies included in AI methods show a sharp increase in publication volume from 2022 onward. Notably, 2023 marked the peak with 23 studies, followed closely by 19 in 2024. This trend reflects the growing research attention to AI applications in BC imaging, particularly in recent years. Figure 4 presents a conceptual overview of the key domains, methodologies, and thematic relationships addressed in this systematic review.
Figure 4:
Conceptual map illustrating the thematic structure of the review.
4.1. Structured Synthesis of Findings
In our review, we grouped the final studies into three primary domains. First, we classified AI Applications (n = 56) based on an algorithmic approach (e.g., traditional ML, CNNs, ViTs, hybrid models). Second, we categorized Imaging Techniques (n = 47) based on the primary imaging modality (mammography, US, MRI). Third, we included Health Disparity Studies (n = 43) based on focus on racial/ethnic disparities, geographic or socio-economic barriers, or screening adherence. We assigned each article based on its primary research question and method. When studies overlapped categories (e.g., AI applied to disparity analysis), we categorized them according to their central aim.
To provide a comparative view across the included studies, we synthesized key performance metrics (accuracy, AUC) stratified by imaging modality and model architecture. Table 3 presents a cross-sectional summary derived from 52 representative AI method studies that reported sufficient quantitative detail. Figure 5 visualizes the average performance of these studies by year based on the dataset modality (Figure 5a) or AI methods (Figure 5b), highlighting how different data types or AI methods have evolved in diagnostic accuracy over time. Figure 6 also shows the occurrence of the number of reviewed studies by year based on the AI method (Figure 6a) or modality (Figure 6b), providing insight into the evolving distribution of research focus for the two domains.
Table 3:
Summary of model performance by modality and algorithm type
| Model Type | Modality | Median AUC | Best Accuracy | Rep. Studies |
|---|---|---|---|---|
| CNN (n=17) | Mammography, US, MRI | 0.89 | 99.17% | [144, 209, 210] |
| Transformer-based (n=11) | Histopathology, MRI, US | 0.94 | 99.99% | [147, 170, 171] |
| Traditional ML (n=10) | Tabular, US, Mammography | 0.86 | 98.2% | [133, 138, 211] |
| Hybrid CNN+ViT (n=9) | Mixed (Multi-modal) | 0.93 | 99.16% | [162, 195, 197] |
| LLMs (n=5) | Clinical Text / Scenarios | N/A | 98% | [212–214] |
Figure 5:
Heatmap summaries of performance by year across (a) dataset modality and (b) AI method.
Figure 6:
Bar charts of reviewed literature by year across (a) AI method and (b) dataset modality.
Performance metrics varied widely due to differences in dataset size, class balance, imaging modality, and evaluation methodology. Nevertheless, our synthesis highlights a trend toward improved outcomes in ViT-based and hybrid CNN-transformer architectures, particularly for histopathology and multimodal datasets. Transformer-based models generally yielded superior generalization across classification and segmentation tasks. Meanwhile, traditional ML models demonstrated moderate performance but remain valuable for tabular and low-dimensional data settings.
Despite performance gains, this review identified several cross-cutting limitations. A major concern raised via the review process is that the evaluation of many models from private or single-institution datasets lacks generalizability. [143, 215], which restricts reproducibility and cross-domain applicability. Additionally, fairness assessments remain scarce; only a few studies examined subgroup performance or addressed algorithmic bias, impeding equitable clinical deployment. There is also a notable absence of multimodal and longitudinal integration, with only a handful of works [140, 166, 197] incorporating diverse data types or temporal progression, both of which are critical for robust risk prediction. Furthermore, many segmentation and classification pipelines remain handcrafted, limiting automation, scalability, and adaptability to new data contexts. Finally, explainability remains underdeveloped, especially in transformer-based architectures, where few studies leverage saliency maps or attention heatmaps to enhance interpretability and clinical trust.
To advance the field meaningfully, future research should prioritize several directions. First, establishing open benchmarks and cross-institutional datasets is critical for evaluating the generalizability of AI models across diverse clinical settings and imaging modalities. Second, integrating multimodal and longitudinal data, such as combining MRI, clinical metadata, and histopathology, will better capture disease heterogeneity and improve predictive accuracy over time. Third, equity and bias auditing should become a standard component of AI evaluation; disaggregating model performance by race, age, and other sociodemographic factors is essential to ensure fairness and avoid perpetuating healthcare disparities. Finally, deeper integration of LLMs with imaging pipelines holds promise for bridging unstructured clinical narratives and structured image-based analyses, paving the way for more holistic and deployable diagnostic systems. We summarize included studies using structured Tables B.1, B.2, and B.3 in Appendix B for AI methods, imaging modalities, and health disparities, respectively.
5. Health Disparity in BC Diagnosis
Despite a growing body of work on AI in BC imaging, health disparities have often been underexamined or only briefly mentioned in previous reviews. Most existing literature has focused on algorithmic performance or imaging modality development without systematically analyzing how such technologies may reinforce or alleviate inequities in care. This section addresses that critical gap by reviewing how disparities, particularly those rooted in race, geography, socioeconomic status, and systemic bias, intersect with AI development and deployment in BC diagnosis.
Health disparities in cancer research primarily manifest in one or more of the following forms: (1) Unequal burden of cancer incidence, where certain racial and ethnic groups experience higher rates of specific cancers due to genetic, environmental, and socioeconomic factors; (2) Disparities in access and availability of healthcare services to obtain a mammogram screening [216], which can lead to delayed diagnoses and treatment; (3) Differences in treatment outcomes, where minority groups may receive suboptimal care or have limited access to advanced therapies due to financial and geographic barriers; (4) Varying mortality rates [217] are often linked to delayed detection, limited access to care, and differences in treatment adherence; (5) Lack of health education among different groups of women [218]; and (6) Lack of participation in clinical trials [219], rooted in past unethical research practices,limits the development of treatments that address the specific needs of these populations, further widening health disparities. Social, economic, and environmental disadvantages directly create disparities that lead to unequal healthcare experiences and outcomes [220–223]. Addressing these disparities is crucial to improving equity in BC care, as they directly impact the survival rates and quality of life for affected individuals [224].
BC remains a leading health concern globally, with significant disparities in outcomes among different demographic groups [225]. Black women face a 40% higher mortality rate despite a lower incidence [218], while Hispanic women often receive late-stage diagnoses, worsening their prognosis. Black women experience higher mortality rates due to factors such as socioeconomic status, healthcare access, and biological differences in tumor characteristics [226]. In the year 2024, Black women have a BC mortality rate 38% higher than that of White women, despite having a 5% lower incidence [227].
5.1. Factors Contributing to Disparities
Several factors contribute to these disparities, with socioeconomic status being one of the most significant. Women with lower socioeconomic status often face barriers to accessing healthcare services, including BC screening and treatment [220]. These barriers may include limited access to health insurance, which restricts their access to regular mammograms, follow-up tests, and advanced treatments [228]. Lower-income women often face challenges such as struggling to take time off work, lacking transportation, and other logistical barriers that hinder timely access to necessary care [220].
Geographical location is another critical factor contributing to health disparities in BC screening, diagnosis, and treatment. Women living in rural areas often have limited access to healthcare facilities, including those that offer specialized BC diagnostic services [229–231]. This geographic disparity can result in delays in diagnosis and treatment, leading to worse outcomes compared to women in urban areas. Moreover, regional differences in healthcare policies and resources can exacerbate these disparities, particularly in areas that lack the infrastructure needed for effective BC care [232–234].
Cultural and language barriers further compound these disparities. Cultural beliefs and practices can significantly influence how women perceive BC and their willingness to seek screening and treatment [235]. In some cultures, the stigma associated with cancer can significantly delay individuals seeking timely medical care [236]. Cultural and societal perceptions, including beliefs rooted in certain religious or spiritual ideologies, can influence the stigma surrounding cancer [237]. For example, reliance on traditional or natural medicine practices, often viewed as more culturally acceptable or aligned with spiritual beliefs, may lead individuals to prioritize these alternatives over conventional medical treatments [238]. Practitioners have observed this preference for traditional medicine in various cultures, and it can influence how people seek healthcare [239]. Fear of social ostracization or judgment, along with misconceptions that a cancer diagnosis equates to a death sentence, further compounds individuals’ reliance on avoidance and discourages proactive health-seeking behavior [240]. These factors collectively contribute to late-stage diagnoses and poorer outcomes, underscoring the importance of culturally sensitive education and interventions to reduce stigma and improve access to early cancer detection and care [241].
Language barriers can impede communication between patients and healthcare providers, resulting in misunderstandings, misdiagnoses, and delayed treatment. Women who do not speak the local predominant language may find it challenging to navigate the healthcare system, leading to disparities in the quality of care they receive [242, 243]. Additionally, medical mistrust, rooted in historical injustices and unethical practices, is another key factor that contributes to disparities in BC diagnosis [244]. This mistrust leads to delayed screenings, reduced adherence to medical recommendations, and lower participation in clinical trials, further widening disparities [245].
-
Healthcare provider bias contributes to disparities in BC screening, diagnosis, and treatment. Implicit biases [246] among healthcare providers can lead to differences in the quality of care received by minority women. Studies have shown that minority women may receive less aggressive treatment recommendations or face longer wait times for diagnostic procedures compared to their White counterparts [247]. These biases, often unconscious, can contribute to the higher mortality rates observed in minority populations [248, 249].
An under-recognized yet critical aspect of provider bias lies in postpartum care and breastfeeding support—both of which play key roles in breast cancer prevention. Black women in the U.S. are statistically less likely to breastfeed, due to factors like shorter maternity leave, workplace constraints, and limited access to lactation counseling [250, 251]. Studies have shown that breastfeeding reduces the risk of aggressive subtypes such as triple-negative and estrogen receptor–negative breast cancers—both of which disproportionately affect Black women [250, 252]. However, disparities in postpartum follow-up—such as fewer clinical breast exams and missed preventive care visits—may contribute to delayed detection of early-stage cancers among Black women of childbearing age [253]. These findings underscore the importance of equitable postpartum care, including culturally sensitive provider engagement, early breast health monitoring, and improved breastfeeding support services.
Racial and ethnic disparities make inequalities more noticeable in BC diagnosis. Black women, for example, face significant disparities in BC outcomes. Although they have a lower BC incidence rate, clinicians more often diagnose them with aggressive subtypes such as triple-negative BC, which are harder to treat and linked to poorer prognoses. [254]. Clinicians are also more likely to diagnose Black women at later stages, which further reduces their chances of survival [255, 256]. Hispanic and Latina women also face barriers to early BC diagnosis, often due to socio-cultural factors and their immigration status, which can limit their access to healthcare services [257]. Other minority groups, such as Native American, Asian American, and Pacific Islander women, also experience disparities in BC screening, diagnosis, and treatment outcomes, often due to lower screening rates and limited access to specialized care [258, 259]. Furthermore, the lack of participation in clinical trials due to historical mistrust of medical research in minority communities means that anti-cancer medications may not work as well in minority populations [219].
5.2. AI Role to Reduce BC Health Disparity
Before examining the role of AI in addressing BC health disparities, it is essential to recognize the ongoing importance of non-AI interventions in reducing inequities in care. These include expanding access to high-quality screening and diagnostic services, especially in underserved and rural areas, and promoting public health initiatives that improve awareness, education, and culturally sensitive outreach to encourage early detection [260]. Policy efforts to address socioeconomic barriers, such as improving insurance coverage, reducing out-of-pocket costs, and enhancing patient navigation, are also critical for ensuring timely and appropriate treatment [261]. Additionally, increasing workforce diversity and training healthcare providers to recognize implicit biases can strengthen the patient-provider communication and trust [262]. These foundational efforts collectively address systemic issues beyond the reach of technology alone and establish the necessary conditions for effective and equitable AI implementation.
The development and application of AI have the potential to either mitigate or exacerbate bias in BC diagnosis and screening [263]. Researchers and developers can use diverse datasets to train AI models, thereby enhancing diagnostic accuracy and mitigating disparities for underserved populations, including women with dense breast tissue or racial and ethnic minorities [264]. Nevertheless, developers may perpetuate existing inequities by training AI on biased or incomplete data, which leads to higher false negative rates and lower accuracy for underrepresented populations [265]. AI tools that fail to account for variations in access to care, socioeconomic status, and biological distinctions are at risk of favoring well-represented groups and excluding vulnerable patients [266]. Consequently, it is imperative to guarantee algorithm transparency, data diversity, and continuous evaluation to prevent AI from exacerbating preexisting biases in BC care [263].
AI technology offers promising solutions to address health disparities in BC screening and diagnosis. AI-driven tools, such as ML algorithms and CAD systems, can enhance the accuracy of mammograms and other screening and diagnostic tools, helping to detect cancer earlier and more reliably, even in underserved populations [267]. AI can improve access to BC screening by enabling mobile health applications and telemedicine platforms that can reach remote and low-income communities. Additionally, researchers and clinicians can use AI to develop personalized risk assessment models that consider demographic, genetic, and lifestyle factors unique to each patient, leading to earlier interventions and better outcomes for women at higher risk of developing BC [268].
Despite these benefits, one persistent challenge has been the “black box” nature of many AI systems, particularly DL models, where decision-making processes are often opaque, hindering clinical adoption [269, 270]. To address this, the development of XAI solutions has become increasingly important. XAI methods aim to make AI model predictions more transparent, interpretable, and clinically actionable by providing visual or textual justifications for the decisions made [269, 271]. In the context of BC diagnosis, techniques such as saliency maps [272], Grad-CAM [273], SHAP (SHapley Additive exPlanations) [274], and LIME (Local Interpretable Model-agnostic Explanations) [275] can highlight image regions or features that influenced the AI’s prediction, allowing clinicians to verify that the model is focusing on relevant anatomical structures or tumor characteristics.
Moreover, XAI not only enhances clinical interpretability but also supports bias detection and error analysis, which are critical for ensuring equitable care across diverse populations [276, 277]. For example, XAI tools can help uncover whether a model’s performance degrades for patients of specific racial, socioeconomic, or age groups, thus guiding fairness-aware model improvements [276]. By increasing the transparency and accountability of AI systems, XAI has the potential to foster greater clinician trust, regulatory acceptance, and ultimately, patient safety in real-world diagnostic settings [271, 277].
To develop fair AI models, datasets must also be diverse. AI models trained on homogeneous datasets may not perform well across diverse populations, potentially perpetuating existing disparities. Clinicians and researchers must encourage women in minoritized communities to participate in clinical trials and ensure that they train AI algorithms on diverse and representative datasets to begin addressing systemic inequities in healthcare. Wider clinical trial participation strengthens research generalizability and supports equitable healthcare outcomes. Moreover, diverse datasets are essential to mitigate biases in AI systems, ensuring that predictive models and healthcare interventions are effective across different demographic groups. This dual approach combats the perpetuation of health disparities, fosters trust, and improves outcomes for historically underserved communities [278].
Several research endeavors have explored the use of AI to mitigate health disparities in BC diagnosis. In regions with high mortality rates due to screening and diagnostic delays, cloud-based AI solutions enable remote mammogram analysis by specialized radiologists. Gao et al. [279] developed a MobileNetV2-based system for sentinel lymph nodes, improving accuracy and reducing delays. Zhang et al. [280] introduced ‘Dr.J’, an AI-powered ultrasonography tool that facilitates accessible BC screening via cloud computing without specialized medical expertise. AI also enhances diagnostic accuracy and efficiency. Hamid et al. [281] demonstrated that AI improves mammography density assessment, reduces unnecessary biopsies, and supports BI-RADS classification with substantial agreement among radiologists. Xavier et al. [282] reported a 68.3% workload reduction with 93.1% sensitivity in BC screening, optimizing resources and alleviating workforce shortages. Zarcaro et al.[283] highlighted AI’s ability over traditional CAD systems in breast lesion identification and false positive reduction, enabling personalized screening. Despite significant advancements in research leveraging AI to address health disparities, there remains a critical need for further U.S.-based studies to expand and refine these efforts.
The use of heterogeneous datasets in BC screening plays a crucial role in mitigating racial disparities by enhancing the accuracy and robustness of diagnostic models. These datasets incorporate diverse patient demographics and biological data, which help address the variability in tumor biology and socioeconomic factors contributing to racial disparities in BC outcomes [284]. However, existing datasets often underrepresent minority groups, e.g., Black women, and lack detailed annotations on socioeconomic status, hindering the understanding of biases in BC outcomes [285]. Therefore, researchers must conduct more work in data collection and annotation, especially for minority and traditionally underserved populations.
Some studies have developed techniques and methods to address racial bias in datasets and enhance model performance across diverse populations. Specifically, Eshun et al. [286] used a Deep Convolutional Generative Adversarial Network (DCGAN) to generate synthetic data for the minority class in the imbalanced BreakHis dataset, achieving high accuracy in distinguishing benign and malignant tumors by fine-tuning a ResNet50 model. Similarly, Subasree et al. [287] proposed an Improved Generative Adversarial Network (I-GAN) with a Modified Convolutional Neural Network (MCNN) to balance the Wisconsin dataset, improving classification metrics, Geometric Mean, and Matthews’s Correlation Coefficient. Park et al. [288] developed race/ethnicity-specific survival ML models using the SEER dataset (322,348 women) to address racial disparities in BC survival predictions, outperforming general models, and demonstrating the value of individualized oncology care in mitigating representation bias. Smith et al. [289] tackled ancestral bias in genomic datasets by developing PhyloFrame, an ML method integrating big data interaction networks and population variation data, significantly improving prediction accuracy across ancestries. Contributions from studies focused on producing heterogeneous BC detection datasets include Jeong et al. [75], who created a dataset of 3.65 million mammograms from 116,000 racially diverse patients, including a significant representation of Black women, to improve AI generalizability. In another study, Frazer et al. [290] curated the ADMANI dataset, comprising 4.4 million images from 629,863 women across multiple screening episodes and countries, enabling real-world AI evaluation in mammographic screening and promoting equitable healthcare outcomes.
Several studies have incorporated XAI techniques in BC diagnosis. Specifically, Maouche et al. [291] used LIME explainer to assess the influence of patient and treatment attributes on BC metastasis, identifying key factors from substantial (e.g., absence of adjuvant chemotherapy) to minimal (e.g., use of oral contraceptives). In another study, Islam et al. [292] applied the SHAP XAI technique to an XGB model trained on 500 mammograms from Dhaka Medical College Hospital, determining which breast regions most influenced model decisions. Furthermore, Sobhana et al. [293] integrated ML and XAI to predict BC using fine needle aspiration data, demonstrating improved accuracy with RF, SVM, and KNN while emphasizing early BC detection’s role in improving survival rates, providing insights into cancerous cell features to support informed clinical decisions, and mitigating health disparities in BC screening, diagnosis, and treatment outcomes.
In addition to addressing BC health disparities, multimodal AI approaches offer opportunities to integrate imaging, pathology, genomics, and clinical data to improve diagnostic accuracy and personalized care [294]. These tools can help address workforce shortages by supporting clinicians in interpreting complex data and guiding treatment decisions through risk prediction models that incorporate clinical and genomic information, such as Oncotype DX [295]. Researchers and healthcare providers are also exploring AI-powered LLMs, like ChatGPT, to enhance patient education and support shared decision-making [296]. Additionally, AI can improve screening guidelines to prioritize high-risk populations and plays an increasing role in clinical trial matching to ensure diverse patient inclusion. When thoughtfully implemented, these innovations hold the potential to reduce disparities in BC care by promoting precision, equity, and broader access to advanced diagnostics and treatments [297]. In summary, the reviewed literature highlights efforts to mitigate various forms of health disparities.
6. Discussion
The development of streamlined computational techniques using AI-based algorithms for BC screening, diagnosis, and bias mitigation remains an active and impactful area of oncological research. Advances in AI-powered screening tools hold significant potential to enhance patient outcomes and healthcare utilization through earlier detection and more effective treatment planning [31–34]. As these technologies continue to mature, their integration into breast imaging workflows offers a pathway to more accessible, efficient, and equitable cancer care.
Despite promising progress, a critical examination of the 56 AI-focused studies included in this review reveals substantial heterogeneity in model design, dataset composition, validation strategies, and outcome metrics. While many studies reported high diagnostic performance, several methodological limitations persist. Only 24.7% of the studies employed external validation cohorts (e.g., [143, 167]), which restricts generalizability. Fewer than 6.5% explicitly assessed fairness through subgroup performance stratification (e.g., [212, 298]), limiting insight into potential biases. Additionally, only 14 studies incorporated explainability tools such as Grad-CAM or SHAP (e.g., [299, 300]), which highlights the underutilization of interpretability techniques in current AI pipelines. Moreover, reliance on widely used but demographically narrow datasets, such as CBIS-DDSM and INbreast, raises concerns about algorithmic bias and representational fairness. These limitations underscore the need for robust, externally validated, and demographically inclusive AI pipelines.
Nevertheless, the field has witnessed notable advances in AI-based learning architectures that have improved tumor and BC detection across diverse datasets. DL algorithms have demonstrated strong capabilities in accurately classifying tumor types and identifying malignant tissues. Among these, ViT has outperformed traditional neural network models such as ResNet, UNet, and EfficientNet in terms of diagnostic accuracy and computational efficiency. Their ability to process high-resolution imaging data enables precise detection of malignant features. This shift in architecture design reflects models that capture both global context and fine-grained patterns via self-attention mechanisms, overcoming the limitations inherent in conventional CNNs, which often suffer from restricted receptive fields and a bias toward local features.
The conceptual map in Figure 4, the heatmaps in Figure 5a and 5b and bar charts in Figure 6a and 6b synthesize findings across AI methods and imaging modalities, visualizing performance trends, and identifying gaps. Figure 5a shows that both modern DL and traditional ML methods consistently get high diagnostic accuracy. However, segmentation methods experienced a significant decline in 2024. Hybrid and transformer-based methods continue to perform well, but LLMs have become less accurate and more variable in recent years. Figure 5b illustrates that mammography, MRI, and histopathology have consistently demonstrated high accuracy over the years, whereas tabular and US data remain unchanged. On the other hand, clinical text models performed poorly in 2024, indicating the challenges of applying AI to unstructured clinical data.
Figure 6(a) indicates that the number of single-modality and multi-modality BC AI studies has been steadily rising over time, with single-modality studies always being the most common. In 2023, multi-modality studies gained popularity, indicating that an increasing number of people are interested in combining different types of data to improve diagnosis accuracy. Figure 6(b) tracks how AI methods have changed from 2019 to 2024, with modern DL being the most popular method every year. Recently, hybrid approaches and LLMs have become more common. This shift indicates that the field is shifting towards the use of language-based insights and integrating various methods. Together, these visual tools identify patterns, limitations, and opportunities for more robust, equitable, and clinically relevant AI deployment across modalities.
Table B.1 summarizes over 50 AI-focused studies with CNNs, ResNets, and ViTs, which often achieved over 95% accuracy. However, many models rely on small or uniform datasets, limiting generalizability. While DeepLabV3+ and nnUNet perform well for segmentation on MRI and ultrasound, dataset heterogeneity and limited public benchmarks hinder reproducibility. Similarly, Table B.2 confirms that AI performance varies between imaging methods, as mammography and tabular modalities maintain the highest accuracy. At the same time, ultrasound and MRI suffer from overfitting and generalization issues. Researchers have yielded promising results in many studies, but they often limit their findings by using relatively undiverse datasets or focusing on a single modality. These findings highlight the importance of having standardized, large datasets and custom preprocessing. Finally, Table B.3 outlines AI-driven strategies addressing BC disparities, emphasizing dataset diversity, fairness frameworks, and explainability. Foundations that are not technical, like public health policy, access to clinical trials, and ethical governance, are equally important for making sure that AI-driven tools are fair, scalable, and effective for everyone.
Despite these advances, several persistent challenges remain. Researchers must adapt DL models to the specific constraints of each imaging modality, such as differences in resolution, noise, and artifacts. Moreover, to maximize AI’s clinical value, models must go beyond image interpretation to incorporate contextual clinical information, handle incomplete data, and undergo rigorous validation in prospective clinical scenarios [301]. Multimodal AI systems that integrate imaging, genomic, and clinical records represent a vital step toward personalized, comprehensive diagnosis and treatment planning.
Real-world deployments have begun to bridge research and practice. Specialized breast radiologists have reviewed mammogram images from any location by using cloud-ready AI solutions that researchers developed to address geographical barriers. Beyond cloud-based image transfer, recent advancements have integrated real-time AI support directly into clinical workflows. AI-assisted reading stations in radiology suites allow algorithms to highlight suspicious regions, prioritize urgent cases, and suggest risk scores in real time [302–304]. These tools integrate with existing Picture Archiving and Communication Systems (PACS) and improve both efficiency and diagnostic accuracy, particularly in high-volume screening programs.
Some triage systems automatically identify normal cases for single-read protocols, reducing workload and allowing radiologists to focus on high-risk exams [303, 304]. These decision support systems have proven especially valuable in settings with limited access to radiologists or where clinician fatigue may compromise diagnostic accuracy [302, 305]. By reducing diagnostic delays and standardizing interpretation across sites, these systems help bridge access gaps and promote equity [304, 306]. Similarly, lightweight mobile-ready models such as MobileNetV2 have reduced diagnostic delays and improved access to specialized care. In addition, AI tools have enhanced breast lesion detection, improved BI-RADS categorization, and reduced unnecessary biopsies, optimizing both clinical decision-making and healthcare resource allocation. Tools like “Dr.J” have expanded accessibility by allowing non-specialist operators to conduct screenings.
Researchers have begun to tackle racial disparities in BC detection by creating heterogeneous datasets to ensure broader representation. Additionally, developers are mitigating cultural and healthcare provider biases by developing XAI solutions that cultivate trust among healthcare professionals and promote equitable diagnostic processes tailored to the unique needs of individual patients. However, continued studies are necessary to address the existing disparities and ensure these interventions translate into tangible, real-world impacts. Challenges related to equitable AI deployment remain. Many current datasets underrepresent racial and socioeconomic minority populations, particularly Black women, leading to potential biases in diagnostic performance. There is an urgent need to include racially, ethnically, and socioeconomically diverse populations in BC datasets to ensure AI systems perform reliably across all demographic groups. Furthermore, research must assess the broader impact of AI systems on BC mortality, health equity, and cost effectiveness, especially in low-resource settings.
Researchers and practitioners must address ethical and practical considerations. Integrating AI into routine healthcare workflows will require investment in provider training, infrastructure, and technical capacity, particularly in underserved areas. Moreover, concerns around data privacy and patient trust in AI-driven systems are especially salient for cloud-based or federated applications. Transparency, accountability, and equitable access must remain central to AI development and deployment. Regulatory and ethical bodies such as the EU, FDA, WHO [307], and OECD [308] are shaping AI’s transition from research settings into clinical use through frameworks like the EU AI Act [309, 310], the FDA’s Good Machine Learning Practice (GMLP) guidelines [311, 312], and global standards. These frameworks emphasize key principles such as transparency, bias mitigation, human oversight, and fairness to ensure that clinical AI systems are safe, effective, and broadly beneficial. Adherence to these standards will be essential for regulatory approval and the sustained trust of clinicians, institutions, and patients alike.
To advance the field, future efforts should prioritize open benchmarks, cross-institutional datasets, and standardized pipelines to support robust validation. Integrating multimodal and longitudinal data will enhance accuracy and reflect clinical complexity. Routine fairness auditing, disaggregating performance by race, age, and other demographics, is essential. Deeper integration of LLMs with imaging workflows also shows promise for bridging structured and unstructured data.
7. Future Directions
While current applications of AI in BC detection have demonstrated strong performance and practical utility, significant challenges remain around data privacy, fairness, interpretability, regulatory alignment, and real-world deployment. To transition from research to widespread clinical adoption, the next phase of development must focus on building systems that are not only technically accurate but also ethically grounded, transparent, and broadly accessible.
Federated Learning (FL) offers a privacy-preserving approach to collaborative AI development across institutions. This decentralized framework is a promising strategy to address the concerns around data privacy and multi-institutional collaboration. Rather than sharing sensitive patient data, FL shares model updates from local nodes, enabling multi-institutional training while complying with strict regulatory frameworks such as HIPAA and GDPR [313, 314]. The FL paradigm not only preserves patient confidentiality but also enables researchers to use diverse and geographically distributed datasets to develop robust and generalizable AI models. [315, 316]. In BC imaging, FL has enabled the integration of imaging, histopathology, and EHR data from diverse centers, thus enhancing generalizability [317, 318]. However, standardization of FL protocols, auditing for fairness, and addressing performance degradation under data heterogeneity remain active areas of research. Regulatory agencies like the FDA and the forthcoming EU AI Act should pair their further integration with formal guidelines.
Self-Supervised Learning (SSL) and Multiple Instance Learning (MIL) can reduce dependency on annotated datasets. In BC imaging, where expert annotation requires specialized radiological or pathological expertise, SSL presents a promising direction for scaling up AI training without the bottleneck of labeled datasets. Recent advances in SSL frameworks, such as contrastive learning and masked image modeling, allow models to capture meaningful visual representations from raw images alone. In parallel, methods like MIL offer another viable strategy in low-annotation contexts. MIL frameworks allow models to be trained on sets of image patches (or “bags”) with only bag-level labels, making them especially useful for histopathology and mammography, where detailed region annotations are scarce. Several recent MIL approaches, including spherical separation-based classification [319], heuristic linear separation methods [320], and other instance-to-bag algorithms [319], have shown strong potential in medical image classification, particularly when ground truth-pixel level or lesion-level labels are unavailable. Together, these strategies offer scalable solutions to overcome annotation bottlenecks and expand AI utility in low-resource and underrepresented populations. Looking ahead, integrating SSL into BC AI pipelines could accelerate model development, expand access to robust diagnostic tools in under-annotated domains, and enhance generalizability by exposing models to more diverse real-world data. Its role will likely be central to future AI systems that aim to balance performance, scalability, and equity in diagnostic imaging.
Building on foundational advances like FL and SSL, recent research is shifting toward scalable AI paradigms that bridge modalities and enable richer contextual reasoning. Notably, large language and vision-language models offer new frontiers in understanding complex clinical narratives and aligning imaging data with textual evidence.
Large Language Models (LLMs), built on NLP-based transformer architectures, extract diagnostic insights from unstructured clinical data such as radiology reports, pathology notes, and electronic health records (EHRs) [321]. Recent advances in LLMs, including GPT, Gemini, and LLaMA, have expanded NLP capabilities. These transformer-based models help extract diagnostic information, identify symptom risk patterns, generate clinical summaries, and translate findings into patient-friendly language [155–157, 213, 214, 298, 322]. However, their integration into clinical workflows necessitates rigorous validation due to concerns around data provenance, fairness, and interpretability [323].
Complementing LLMs, Vision Language Models (VLMs) fuse visual data from mammograms or ultrasound with textual records to enable richer diagnostic interpretation. Algorithms like CLIP, BLIP, and Flamingo have shown promise in aligning imaging biomarkers with patient-specific clinical context, supporting risk stratification and treatment planning [324–327]. VLMs also aid in automated report generation and patient education through visual textual summarization [328, 329]. While promising, their safe deployment requires careful evaluation across diverse demographics, imaging platforms, and health system workflows [330]. Improving explainability, reducing biases, and integrating VLMs into real-time decision processes are priorities for future work.
The need for XAI has never been greater. Clinician trust, regulatory approval, and patient safety depend on the interpretability of model outputs. Techniques such as saliency maps, class activation mapping, and uncertainty quantification can provide insight into AI decision-making. These features are critical in flagging ambiguous cases and establishing accountability. Additionally, Uncertainty quantification can signal when AI systems require review by human experts, reducing errors. By focusing on scalable, accessible, and equitable AI technologies, the future of BC diagnosis will enhance diagnostic precision and ensure these advancements benefit populations worldwide.
AI in BC diagnosis sits at a critical inflection point. Future systems must not only improve technical metrics but also demonstrate ethical robustness, regulatory compliance, and real-world readiness. Only then can AI fulfill its promise of improving outcomes for all patients, regardless of geography, race, or socioeconomic status.
Limitations of the Review
This review has several limitations. First, although we adhered to PRISMA guidelines for systematic reviews, we did not conduct a formal meta-analysis due to the heterogeneous nature of the included studies and the narrative focus of our review. Instead, we present Table 3, Figures 5, 6, and Appendix B, which offer aggregated overviews of key trends across study clusters, including AI methods, imaging modalities, and health disparities. Second, we included only studies published in English, which may introduce a language bias and exclude relevant non-English research. Third, many included studies relied on datasets from Western populations, which may limit the generalizability of findings to global and underrepresented communities. Fourth, there may be publication bias due to the overrepresentation of studies reporting positive or novel results in AI applications. Lastly, as most of the reviewed studies did not explicitly report demographic breakdowns or fairness assessments, evaluating the equity impact of AI models remains challenging. Readers should consider these limitations when interpreting the review’s findings. To ensure transparency and reproducibility of reporting, we provide the completed PRISMA 2020 checklist, aligned with McKenzie et al. [208] as supplementary material in Appendix C.
8. Conclusions
In conclusion, over the last two decades, the screening and diagnosis of BC have evolved significantly with the integration of AI, offering promising advancements in early detection and treatment planning. AI’s ability to analyze vast datasets and recognize complex patterns in imaging and genetic data has the potential to enhance diagnostic precision, thereby improving healthcare outcomes. This paper has comprehensively reviewed modern AI algorithms’ role in BC diagnosis. We covered design aspects and surveyed recent papers, highlighting their findings and limitations. This review discussed details of imaging technologies used in screening and diagnosis, as well as the types of learning algorithms used. Additionally, recent research has shown the potential of AI-driven tools to play a crucial role in mitigating BC disparities by providing equitable access to high-quality screening and diagnostic technologies, especially in underrepresented populations. Research opportunities in this area still exist, including the ethical implications and biases inherent in AI systems for fair and inclusive healthcare solutions. Finally, continued tools and algorithms development, interdisciplinary team collaboration, and a commitment to equity will be key to fully realizing AI’s potential in BC screening and diagnosis, and the potential for AI to reduce global health disparities.
This review highlights critical translational obstacles—namely, explainability, validation, regulatory awareness, and equity—that researchers must resolve to transition AI systems from prototype to clinical application. It defines the progression of the field from fundamental image classification to multimodal models that incorporate the clinical environment. Future research must emphasize external validation, fairness assessment, and public benchmarking to guarantee effective real-world implementation. This analysis offers a strategic roadmap that integrates current outcomes and highlights essential gaps and priorities for ethical innovation in AI-driven BC treatment.
Supplementary Material
Examine diverse imaging techniques such as Mammography, Ultrasound, and Magnetic Resonance Imaging (MRI), evaluating their advantages and limitations.
Provide an extensive review of state-of-the-art AI applications in BC classification and segmentation, including both traditional machine learning and deep learning techniques.
Analyze the role of AI in addressing health disparities in BC screening and diagnosis, highlighting both the potential benefits and challenges.
Funding
The research reported in this paper was supported by AIM-AHEAD Coordinating Center at the University of North Texas Science Center at Fort Worth. This research was, in part, funded by the National Institutes of Health (NIH) Agreement No. 1OT2OD032581. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the NIH.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Credit authorship contribution statement
All authors contributed significantly to the study’s conception, analysis, and manuscript preparation and editing. Material preparation: O.A., J.D., J.A., F.W., G.S., and F.K. Data analysis, O.A., J.A., Md.R., H.T., J.W.G. Validation and Visualization: O.A., J.A., F.W., G.S., H.T.; J.W.G. Investigation: H.T.; J.W.G., G.S., Md. R.; T.M., and F.K. Methodology and Supervision: T.M.; Md.R.; and F.K. Fund Acquisition: Md.R., T.M., H.T., J.W.G., and F.K. The first draft of the manuscript was written by O.A., J.D., J.A., F.W., and F.K. Writing–review & editing: H.T., J.W.G., T.M., F.W., and F.K. All authors commented, edited, and approved the final manuscript before submission.
Ethics Statement
Ethics approval is not required for this study because no patients’ data were used/reported.
References
- [1].World Health Organization, Breast cancer (2024).
- [2].Kim J, Harper A, McCormack V, Sung H, Houssami N, Morgan E, Mutebi M, Garvey G, Soerjomataram I, Fidler-Benaoudia MM, Global patterns and trends in breast cancer incidence and mortality across 185 countries, Nature Medicine (Feb. 2025). doi: 10.1038/s41591-025-03502-3. URL https://doi.org/10.1038/s41591-025-03502-3 [DOI] [PubMed] [Google Scholar]
- [3].Arzanova E, Mayrovitz HN, The epidemiology of breast cancer, Exon Publications (2022) 1–19. [PubMed] [Google Scholar]
- [4].Siegel RL, Miller KD, Wagle NS, Jemal A, Cancer statistics, 2023., CA: a cancer journal for clinicians 73 (1) (2023). [DOI] [PubMed] [Google Scholar]
- [5].Ghorbian M, Ghorbian S, Usefulness of machine learning and deep learning approaches in screening and early detection of breast cancer, Heliyon 9 (12) (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Qaseem A, Lin JS, Mustafa RA, Horwitch CA, Wilt TJ, C. G. C. of the American College of Physicians*, Screening for breast cancer in average-risk women: a guidance statement from the american college of physicians, Annals of internal medicine 170 (8) (2019) 547–560. [DOI] [PubMed] [Google Scholar]
- [7].Long H, Brooks JM, Harvie M, Maxwell A, French DP, How do women experience a false-positive test result from breast screening? a systematic review and thematic synthesis of qualitative studies, British journal of cancer 121 (4) (2019) 351–358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Thunnissen E, Kerr KM, Herth FJ, Lantuejoul S, Papotti M, Rintoul RC, Rossi G, Skov BG, Weynand B, Bubendorf L, et al. , The challenge of nsclc diagnosis and predictive analysis on small samples. practical approach of a working group, Lung cancer 76 (1) (2012) 1–18. [DOI] [PubMed] [Google Scholar]
- [9].Atrey K, Singh BK, Bodhey NK, Pachori RB, Mammography and ultrasound based dual modality classification of breast cancer using a hybrid deep learning approach, Biomedical Signal Processing and Control 86 (2023) 104919. [Google Scholar]
- [10].Sapate S, Talbar S, Mahajan A, Sable N, Desai S, Thakur M, Breast cancer diagnosis using abnormalities on ipsilateral views of digital mammograms, Biocybernetics and Biomedical Engineering 40 (1) (2020) 290–305. [Google Scholar]
- [11].Hild S, Johanet M, Valenza A, Thabaud M, Laforest F, Ferrat E, Rat C, t. F. N. C. o. G. P. DEDICACES Group, Quality of decision aids developed for women at average risk of breast cancer eligible for mammographic screening: Systematic review and assessment according to the international patient decision aid standards instrument, Cancer 126 (12) (2020) 2765–2774. [DOI] [PubMed] [Google Scholar]
- [12].Salem MRH, Chalabi NAMT, Mohammed AAGB, Yacoub GEE, The incidence of breast cancer in egyptian females in correlation to different mammographic acr densities, Folia Medica 66 (2) (2024) 213–220. [DOI] [PubMed] [Google Scholar]
- [13].De Ruvo V, Catacchio C, Pallara C, Lorusso V, Moschetta M, et al. , Role of biopsy techniques in breast cancer diagnosis: overview and update, Journal of radiological review 10 (4) (2023) 233–240. [Google Scholar]
- [14].Rahman WT, Helvie MA, Breast cancer screening in average and high-risk women, Best Practice & Research Clinical Obstetrics & Gynaecology; 83 (2022) 3–14. doi: 10.1016/j.bpobgyn.2021.11.007. URL https://www.sciencedirect.com/science/article/pii/S1521693421001711 [DOI] [PubMed] [Google Scholar]
- [15].Castells X, Torá-Rocamora I, Posso M, Román M, Vernet-Tomas M, Rodríguez-Arana A, Domingo L, Vidal C, Baré M, Ferrer J, et al. , Risk of breast cancer in women with false-positive results according to mammographic features, Radiology 280 (2) (2016) 379–386. [DOI] [PubMed] [Google Scholar]
- [16].Hofvind S, Sagstad S, Sebuødegård S, Chen Y, Roman M, Lee CI, Interval breast cancer rates and histopathologic tumor characteristics after false-positive findings at mammography in a population-based screening program, Radiology 287 (1) (2018) 58–67. [DOI] [PubMed] [Google Scholar]
- [17].Ko NY, Hong S, Winn RA, Calip GS, Association of insurance status and racial disparities with the detection of early-stage breast cancer, JAMA oncology 6 (3) (2020) 385–392. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Metwali E, Pennington S, Mass spectrometry-based proteomics for classification and treatment optimisation of triple negative breast cancer, Journal of Personalized Medicine 14 (9) (2024) 944. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Altobelli E, Angeletti PM, Ciancaglini M, Petrocelli R, The future of breast cancer organized screening program through artificial intelligence: A scoping review, in: Healthcare, Vol. 13, MDPI, 2025, p. 378. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Hernström V, Josefsson V, Sartor H, Schmidt D, Larsson A-M, Hofvind S, Andersson I, Rosso A, Hagberg O, Lång K, Screening performance and characteristics of breast cancer detected in the Mammography Screening with Artificial Intelligence trial (MASAI): a randomised, controlled, parallel-group, non-inferiority, single-blinded, screening accuracy study, The Lancet Digital Health 7 (3) (2025) e175–e183. doi: 10.1016/S2589-7500(24)00267-X. URL https://www.sciencedirect.com/science/article/pii/S258975002400267X [DOI] [PubMed] [Google Scholar]
- [21].Laws E, Palmer J, Alderman J, Sharma O, Ngai V, Salisbury T, Hussain G, Ahmed S, Sachdeva G, Vadera S, Mateen B, Matin R, Kuku S, Calvert M, Gath J, Treanor D, McCradden M, Mackintosh M, Gichoya J, Trivedi H, Denniston AK, Liu X, Diversity, inclusivity and traceability of mammography datasets used in development of Artificial Intelligence technologies: a systematic review, Clinical Imaging 118 (2025) 110369. doi: 10.1016/j.clinimag.2024.110369. URL https://www.sciencedirect.com/science/article/pii/S0899707124002997 [DOI] [PubMed] [Google Scholar]
- [22].Mahedi RA, Iqbal H, Azmee R, Azmee M, Jakir F, Nishan MA, Uddin MB, Afrin S, Current trends and future prospects of artificial intelligence in transforming radiology, Journal of Current Health Sciences 4 (2) (2024) 95–104. [Google Scholar]
- [23].Díaz O, Rodríguez-Ruíz A, Sechopoulos I, Artificial intelligence for breast cancer detection: Technology, challenges, and prospects, European journal of radiology (2024) 111457. [DOI] [PubMed] [Google Scholar]
- [24].Lee SE, Han K, Youk JH, Lee JE, Hwang J-Y, Rho M, Yoon J, Kim E-K, Yoon JH, Differing benefits of artificial intelligence-based computer-aided diagnosis for breast us according to workflow and experience level, Ultrasonography 41 (4) (2022) 718–727. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Lee S, Lee HS, Lee E, Kim WH, Kim J, Yoon JH, Improving breast ultrasonography education: the impact of ai-based decision support on the performance of non-specialist medical professionals, Ultrasonography 44 (2) (2025) 124–133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [26].Kim S-Y, Choi Y, Kim E-K, Han B-K, Yoon JH, Choi JS, Chang JM, Deep learning-based computer-aided diagnosis in screening breast ultrasound to reduce false-positive diagnoses, Scientific Reports 11 (1) (2021) 395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Baek J, Kim J, Kim HJ, Yoon JH, Park HY, Lee J, Kang B, Zakiryarov I, Kultaev A, Saktashev B, et al. , Clinical application of artificial intelligence in breast ultrasound, Journal of the Korean Society of Radiology 86 (2) (2025) 216. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Debelee TG, Schwenker F, Ibenthal A, Yohannes D, Survey of deep learning in breast cancer image analysis, Evolving Systems 11 (1) (2020) 143–163. [Google Scholar]
- [29].Hussain S, Mubeen I, Ullah N, Shah SSUD, Khan BA, Zahoor M, Ullah R, Khan FA, Sultan MA, Modern diagnostic imaging technique applications and risk factors in the medical field: a review, BioMed research international 2022 (1) (2022) 5164970. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Iranmakani S, Mortezazadeh T, Sajadian F, Ghaziani MF, Ghafari A, Khezerloo D, Musa AE, A review of various modalities in breast imaging: technical aspects and clinical outcomes, Egyptian Journal of Radiology and Nuclear Medicine 51 (2020) 1–22. [Google Scholar]
- [31].Nassif AB, Talib MA, Nasir Q, Afadar Y, Elgendy O, Breast cancer detection using artificial intelligence techniques: A systematic literature review, Artificial intelligence in medicine 127 (2022) 102276. [DOI] [PubMed] [Google Scholar]
- [32].Salh CH, Ali AM, Comprehensive study for breast cancer using deep learning and traditional machine learning, Zanco Journal of Pure and Applied Sciences 34 (2) (2022) 22–36. [Google Scholar]
- [33].Patel RM, Jain A, Onyechi AD, Ohemeng-Dapaah J, Shaba WO, Onyechi EA, Ogunlana YO, Patel ZV, Patel MM, Patel DC, et al. , Current status of artificial intelligence and machine learning in breast cancer screening: A systematic review, Magna Scientia Advanced Research and Reviews 11 (1) (2024) 060–067. [Google Scholar]
- [34].Yeasmin MN, Al Amin M, Joti TJ, Aung Z, Azim MA, Advances of ai in image-based computer-aided diagnosis: A review, Array (2024) 100357. [Google Scholar]
- [35].Makurumidze G, Lu C, Babagbemi K, Addressing disparities in breast cancer screening: A review, Applied Radiology 51 (6) (2022) 24–28. [Google Scholar]
- [36].Chanakira EZ, Thomas CV, Balen J, Mandrik O, A systematic review of public health interventions to address breast cancer inequalities in low-and middle-income countries, Systematic Reviews 13 (1) (2024) 195. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [37].Hand T, Rosseau NA, Stiles CE, Sheih T, Ghandakly E, Oluwasanu M, Olopade OI, The global role, impact, and limitations of community health workers (chws) in breast cancer screening: a scoping review and recommendations to promote health equity for all, Global Health Action 14 (1) (2021) 1883336. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [38].Advani P, Advani S, Nayak P, VonVille HM, Diamond P, Burnett J, Brewster AM, Vernon SW, Racial/ethnic disparities in use of surveillance mammogram among breast cancer survivors: a systematic review, Journal of Cancer Survivorship (2022) 1–17. [DOI] [PubMed] [Google Scholar]
- [39].Muradali D, Kennedy EB, Eisen A, Holloway CM, Smith CR, Chiarelli AM, Breast screening for survivors of breast cancer: a systematic review, Preventive Medicine 103 (2017) 70–75. [DOI] [PubMed] [Google Scholar]
- [40].Peters JA, Rubinstein WS, Genetics and the multidisciplinary breast center, Surgical Oncology Clinics of North America 9 (2) (2000) 367–396. [PubMed] [Google Scholar]
- [41].Kattepur AK, Gopinath K, Management of hereditary breast cancer: An overview, Breast Cancer: Comprehensive Management (2022) 353–397. [Google Scholar]
- [42].Lima ZS, Ebadi MR, Amjad G, Younesi L, Application of imaging technologies in breast cancer detection: a review article, Open Access Macedonian Journal of Medical Sciences 7 (5) (2019) 838. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [43].Tabár L, Dean PB, Chen TH-H, Yen AM-F, Chen SL-S, Fann JC-Y, Chiu SY-H, Ku MM-S, Wu WY-Y, Hsu C-Y, et al. , The incidence of fatal breast cancer measures the increased effectiveness of therapy in women participating in mammography screening, Cancer 125 (4) (2019) 515–523. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [44].Jagannath M, Adalarasu K, Nathiya N, Vijayakumar P, Dhivya RK, Digital breast tomosynthesis versus two-dimensional mammography: a comparison on breast imaging for cancer screening, in: Intelligent Computing Techniques in Biomedical Imaging, Elsevier, 2025, pp. 139–147. [Google Scholar]
- [45].Ghaderi KF, Phillips J, Perry H, Lotfi P, Mehta TS, Contrast-enhanced mammography: current applications and future directions, Radiographics 39 (7) (2019) 1907–1920. [DOI] [PubMed] [Google Scholar]
- [46].Welch HG, Bergmark R, Cancer screening, incidental detection, and overdiagnosis, Clinical Chemistry 70 (1) (2024) 179–189. [DOI] [PubMed] [Google Scholar]
- [47].Załuska-Kusz J, Litwiniuk M, Follow-up after breast cancer treatment, reports of practical Oncology and radiotherapy 27 (5) (2022) 875–880. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [48].Wen C, Wang S, Ma M, Xu Z, Zeng F, Zeng H, Liao X, He Z, Xu W, Chen W, Breast masses with rim enhancement on contrast-enhanced mammography: morphological and enhancement features for diagnosis and differentiation of benign and malignant, British Journal of Radiology 97 (1157) (2024) 1016–1021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [49].Taylor DB, Kessell MA, Parizel PM, Contrast-enhanced mammography improves patient access to functional breast imaging, Journal of Medical Imaging and Radiation Oncology (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [50].Savaria D, Kaushik C, Brief communication: The current status of contrast-enhanced mammography in breast imaging, Clinical Imaging (2024) 110213. [DOI] [PubMed] [Google Scholar]
- [51].Bartolović N, Car Peterko A, Avirović M, Šegota Ritoša D, Grgurević Dujmić E, Valković Zujić P, Validation of contrast-enhanced mammography as breast imaging modality compared to standard mammography and digital breast tomosynthesis, Diagnostics 14 (14) (2024) 1575. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [52].Porkodi D, Devimeenal J, Role of contrast-enhanced mammogram as an adjunct to tomosynthesis in evaluation of circumscribed breast lesions, British Journal of Radiology 97 (1162) (2024) 1696–1705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [53].Ong A-HJ, Goh Y, Quek ST, Pillay PG, Lee H-S, Chou C-P, The utility of contrast-enhanced mammography in the evaluation of bloody nipple discharge—a multicenter study in the asian population, Diagnostics 14 (20) (2024) 2297. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [54].Corines MJ, Sogani J, Hogan MP, Mango VL, Bryce Y, The role of contrast-enhanced mammography after cryoablation of breast cancer, American Journal of Roentgenology 222 (2) (2024) e2330250. [DOI] [PubMed] [Google Scholar]
- [55].Phillips J, Fein-Zachary VJ, Slanetz PJ, Pearls and pitfalls of contrast-enhanced mammography, Journal of Breast Imaging 1 (1) (2019) 64–72. [DOI] [PubMed] [Google Scholar]
- [56].Hu N, Research progress on ultrasound medicine in diagnosis and evaluation of chemotherapy effect of breast cancer, Transactions on Materials, Biotechnology and Life Sciences 2 (2024) 77–89. [Google Scholar]
- [57].Sood R, Rositch AF, Shakoor D, Ambinder E, Pool K-L, Pollack E, Mollura DJ, Mullen LA, Harvey SC, Ultrasound for breast cancer detection globally: a systematic review and meta-analysis, Journal of global oncology (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [58].Lin L, Wang LV, The emerging role of photoacoustic imaging in clinical oncology, Nature Reviews Clinical Oncology 19 (6) (2022) 365–384. [DOI] [PubMed] [Google Scholar]
- [59].Iacob R, Iacob ER, Stoicescu ER, Ghenciu DM, Cocolea DM, Constantinescu A, Ghenciu LA, Manolescu DL, Evaluating the role of breast ultrasound in early detection of breast cancer in low-and middle-income countries: A comprehensive narrative review, Bioengineering 11 (3) (2024) 262. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [60].Almalki YE, Soomro TA, Irfan M, Alduraibi SK, Ali A, Impact of image enhancement module for analysis of mammogram images for diagnostics of breast cancer, Sensors 22 (5) (2022) 1868. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [61].Ramadan SZ, Methods used in computer-aided diagnosis for breast cancer detection using mammograms: A review, Journal of healthcare engineering 2020 (1) (2020) 9162464. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [62].Sushanki S, Bhandari AK, Singh AK, A review on computational methods for breast cancer detection in ultrasound images using multi-image modalities, Archives of Computational Methods in Engineering 31 (3) (2024) 1277–1296. [Google Scholar]
- [63].Lobbes M, Heuts E, Moossdorff M, Van Nijnatten T, Contrast enhanced mammography (cem) versus magnetic resonance imaging (mri) for staging of breast cancer: The pro cem perspective, European journal of radiology 142 (2021) 109883. [DOI] [PubMed] [Google Scholar]
- [64].Galati F, Rizzo V, Trimboli RM, Kripa E, Maroncelli R, Pediconi F, Mri as a biomarker for breast cancer diagnosis and prognosis, BJR| Open 4 (1) (2022) 20220002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [65].Cho SM, Cha JH, Kim HH, Shin HJ, Chae EY, Choi WJ, Eom HJ, Kim HJ, Assessing internal mammary lymph node metastasis by breast magnetic resonance imaging in breast cancer, Medicine 102 (47) (2023) e36301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [66].Ahmadinejad N, Azizinik F, Khosravi P, Torabi A, Mohajeri A, Arian A, Evaluation of features in probably benign and malignant nonmass enhancement in breast mri, International Journal of Breast Cancer 2024 (1) (2024) 6661849. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [67].Washington I, Palm RF, White J, Rosenberg SA, Ataya D, The role of mri in breast cancer and breast conservation therapy, Cancers 16 (11) (2024) 2122. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [68].Khoshpouri P, Kazerouni A, Parsian S, Hippe D, Walsh O, Koch L, Partridge S, Rahbar H, Abstract po5-07-07: The biological basis of breast mri background parenchymal enhancement in women with high breast cancer risk, Cancer Research 84 (9_Supplement) (2024) PO5–07. [Google Scholar]
- [69].Arefan D, Zuley ML, Berg WA, Yang L, Sumkin JH, Wu S, Assessment of background parenchymal enhancement at dynamic contrast-enhanced mri in predicting breast cancer recurrence risk, Radiology 310 (1) (2024) e230269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [70].Cozzi A, Di Leo G, Houssami N, Gilbert FJ, Helbich TH, Álvarez Benito M, Balleyguier C, Bazzocchi M, Bult P, Calabrese M, et al. , Screening and diagnostic breast mri: how do they impact surgical treatment? insights from the mipa study, European radiology 33 (9) (2023) 6213–6225. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [71].Terzoni A, Basile P, Gambaro AC, Attanasio S, Rampi AM, Brambilla M, Carriero A, Locoregional staging of breast cancer: contrast-enhanced mammography versus breast magnetic resonance imaging, La radiologia medica 129 (4) (2024) 558–565. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [72].Shen L, Margolies LR, Rothstein JH, Fluder E, McBride R, Sieh W, Deep learning to improve breast cancer detection on screening mammography, Scientific reports 9 (1) (2019) 12495. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [73].Agarwal R, Diaz O, Lladó X, Yap MH, Martí R, Automatic mass detection in mammograms using deep convolutional neural networks, Journal of Medical Imaging 6 (3) (2019) 031409–031409. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [74].Moreira IC, Amaral I, Domingues I, Cardoso A, Cardoso MJ, Cardoso JS, Inbreast: toward a full-field digital mammographic database, Academic radiology 19 (2) (2012) 236–248. [DOI] [PubMed] [Google Scholar]
- [75].Jeong JJ, Vey BL, Bhimireddy A, Kim T, Santos T, Correa R, Dutt R, Mosunjac M, Oprea-Ilies G, Smith G, et al. , The emory breast imaging dataset (embed): A racially diverse, granular dataset of 3.4 million screening and diagnostic mammographic images, Radiology: Artificial Intelligence 5 (1) (2023) e220047. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [76].Sawyer-Lee R, Gimenez F, Hoogi A, Rubin D, Curated breast imaging subset of digital database for screening mammography (cbis-ddsm), The Cancer Imaging Archive (2016). [Google Scholar]
- [77].Moreira IC, Amaral I, Domingues I, Cardoso A, Cardoso MJ, Cardoso JS, Inbreast: toward a full-field digital mammographic database, Academic radiology 19 (2) (2012) 236–248. [DOI] [PubMed] [Google Scholar]
- [78].Suckling J, Parker J, Dance D, Astley S, Hutt I, Boggis C, Ricketts I, Stamatakis E, Cerneaz N, Kok S, et al. , Mammographic image analysis society (mias) database v1. 21 (2015).
- [79].Suckling J, The mammographic image analysis society digital mammogram database, Exerpta Medica. International Congress Series 1069 (1994) 375–378. [Google Scholar]
- [80].Lopez MG, Posada N, Moura DC, Pollán RR, Valiente JMF, Ortega CS, Solar M, Diaz-Herrero G, Ramos I, Loureiro J, et al. , Bcdr: a breast cancer digital repository, in: 15th International conference on experimental mechanics, Vol. 1215, 2012, pp. 113–120. [Google Scholar]
- [81].Cui C, Li L, Cai H, Fan Z, Zhang L, Dan T, Li J, Wang J, The chinese mammography database (cmmd): An online mammography database with biopsy confirmed types for machine diagnosis of breast, The Cancer Imaging Archive 1 (2021). [Google Scholar]
- [82].Wu N, Phang J, Park J, Shen Y, Kim SG, Heacock L, Moy L, Cho K, Geras KJ, The nyu breast cancer screening dataset v1. 0, New York Univ., New York, NY, USA, Tech. Rep (2019). [Google Scholar]
- [83].Heath M, Bowyer K, Kopans D, Moore R, Kegelmeyer WP, The digital database for screening mammography, Proceedings of the Fifth International Workshop on Digital Mammography (2001) 212–218. [Google Scholar]
- [84].Lekamlage CD, Afzal F, Westerberg E, Cheddad A, Mini-ddsm: Mammography-based automatic age estimation, in: Proceedings of the 2020 3rd International Conference on Digital Medicine and Image Processing, 2020, pp. 1–6. [Google Scholar]
- [85].Tangaro S, Bellotti R, De Carlo F, Gargano G, Lattanzio E, Monno P, Massafra R, Delogu P, Fantacci M, Retico A, et al. , Magic-5: an italian mammographic database of digitised images for research magic-5: un database mammografico italiano di immagini digitalizzate per scopi di ricerca, Radiol med 113 (2008) 477–485. [DOI] [PubMed] [Google Scholar]
- [86].Matheus BRN, Schiabel H, Online mammographic images database for development and comparison of cad schemes, Journal of digital imaging 24 (2011) 500–506. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [87].Halling-Brown MD, Warren LM, Ward D, Lewis E, Mackenzie A, Wallis MG, Wilkinson LS, Given-Wilson RM, McAvinchey R, Young KC, Optimam mammography image database: a large-scale resource of mammography images and clinical data, Radiology: Artificial Intelligence 3 (1) (2020) e200103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [88].Al-Dhabyani W, Gomaa M, Khaled H, Fahmy A, Dataset of breast ultrasound images, Data in Brief 28 (2020) 104863. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [89].Yap MH, Pons G, Marti J, Ganau S, Sentis M, Zwiggelaar R, Davison AK, Marti R, Automated breast ultrasound lesions detection using convolutional neural networks, IEEE journal of biomedical and health informatics 22 (4) (2017) 1218–1226. [DOI] [PubMed] [Google Scholar]
- [90].Pawłowska A, Ćwierz-Pieńnkowska A, Domalik A, Jaguś D, Kasprzak P, Matkowski R, Fura Ł, Nowicki A, Żołek N, Curated benchmark dataset for ultrasound based breast lesion analysis, Scientific Data 11 (1) (2024) 148. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [91].Hamouda SKM, Wahed ME, Alez RHA, Riad K, Robust breast cancer prediction system based on rough set theory at national cancer institute of egypt, Computer methods and programs in biomedicine 153 (2018) 259–268. [DOI] [PubMed] [Google Scholar]
- [92].Spanhol FA, Oliveira LS, Petitjean C, Heutte L, A dataset for breast cancer histopathological image classification, Ieee transactions on biomedical engineering 63 (7) (2015) 1455–1462. [DOI] [PubMed] [Google Scholar]
- [93].Lew CO, Harouni M, Kirksey ER, Kang EJ, Dong H, Gu H, Grimm LJ, Walsh R, Lowell DA, Mazurowski MA, A publicly available deep learning model and dataset for segmentation of breast, fibroglandular tissue, and vessels in breast mri, Scientific reports 14 (1) (2024) 5383. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [94].Amgad M, Elfandy H, Hussein H, Atteya LA, Elsebaie MA, Abo Elnasr LS, Sakr RA, Salem HS, Ismail AF, Saad AM, et al. , Structured crowdsourcing enables convolutional segmentation of histology images, Bioinformatics 35 (18) (2019) 3461–3467. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [95].Aksac A, Demetrick DJ, Ozyer T, Alhajj R, Brecahad: a dataset for breast cancer histopathological annotation and diagnosis, BMC research notes 12 (2019) 1–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [96].Janowczyk A, Madabhushi A, Deep learning for digital pathology image analysis: A comprehensive tutorial with selected use cases, Journal of pathology informatics 7 (1) (2016) 29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [97].Aresta G, Araújo T, Kwok S, Chennamsetty SS, Safwan M, Alex V, Marami B, Prastawa M, Chan M, Donovan M, et al. , Bach: Grand challenge on breast cancer histology images, Medical image analysis 56 (2019) 122–139. [DOI] [PubMed] [Google Scholar]
- [98].Brancati N, Anniciello AM, Pati P, Riccio D, Scognamiglio G, Jaume G, De Pietro G, Di Bonito M, Foncubierta A, Botti G, et al. , Bracs: A dataset for breast carcinoma subtyping in h&e histology images, Database 2022 (2022) baac093. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [99].Qaiser T, Mukherjee A, Reddy Pb C, Munugoti SD, Tallam V, Pitkäaho T, Lehtimäki T, Naughton T, Berseth M, Pedraza A, et al. , Her 2 challenge contest: a detailed assessment of automated her 2 scoring algorithms in whole slide images of breast cancer tissues, Histopathology 72 (2) (2018) 227–238. [DOI] [PubMed] [Google Scholar]
- [100].Pérez-Martín J, Sánchez-Cauce R, Quality analysis of a breast thermal images database, Health Informatics Journal 29 (1) (2023) 14604582231153779. [DOI] [PubMed] [Google Scholar]
- [101].Conci A, Thermal image (2014).
- [102].Jossinet J, Variability of impedivity in normal and pathological breast tissue, Medical and biological engineering and computing 34 (1996) 346–350. [DOI] [PubMed] [Google Scholar]
- [103].Leisch F, Dimitriadou E, Leisch MF, No Z, Package ‘mlbench’ (2024).
- [104].Wolberg MOSN, William W Street, Breast Cancer Wisconsin (Diagnostic) (1993).
- [105].Wolberg W, Street W, Mangasarian O, Breast Cancer Wisconsin (Prognostic) (1995).
- [106].Patrício M, Pereira J, Crisóstomo J, Matafome P, Gomes M, Seiça R, Caramelo F, Using resistin, glucose, age and bmi to predict the presence of breast cancer, BMC cancer 18 (2018) 1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [107].Zuo Z, Hu H, Xu Q, Luo X, Peng D, Zhu K, Zhao Q, Xie Y, Ren J, Bbcancer: an expression atlas of blood-based biomarkers in the early diagnosis of cancers, Nucleic Acids Research 48 (D1) (2020) D789–D796. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [108].TENG J, Seer breast cancer data (2019).
- [109].Oluyemi ET, Grimm LJ, Goldman L, Burleson J, Simanowith M, Yao K, Rosenberg RD, Rate and timeliness of diagnostic evaluation and biopsy after recall from screening mammography in the national mammography database, Journal of the American College of Radiology 21 (3) (2024) 427–438. [DOI] [PubMed] [Google Scholar]
- [110].Rodriguez-Ruiz A, Lång K, Gubern-Merida A, Broeders M, Gennaro G, Clauser P, Helbich TH, Chevalier M, Tan T, Mertelmeier T, et al. , Stand-alone artificial intelligence for breast cancer detection in mammography: comparison with 101 radiologists, JNCI: Journal of the National Cancer Institute 111 (9) (2019) 916–922. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [111].Oztekin PS, Katar O, Omma T, Erel S, Tokur O, Avci D, Aydogan M, Yildirim O, Avci E, Acharya UR, Comparison of explainable artificial intelligence model and radiologist review performances to detect breast cancer in 752 patients, Journal of Ultrasound in Medicine 43 (11) (2024) 2051–2068. [DOI] [PubMed] [Google Scholar]
- [112].Freitas V, Ghai S, Au F, Muradali D, Kulkarni S, The Transformative Power of Digital Breast Tomosynthesis and Artificial Intelligence in Breast Cancer Diagnosis, Canadian Association of Radiologists Journal 76 (2) (2025) 302–312, _eprint: https://doi.org/10.1177/08465371241301957. doi: 10.1177/08465371241301957. URL https://doi.org/10.1177/08465371241301957 [DOI] [PubMed] [Google Scholar]
- [113].Jin A, Tosato D, The Implementation of Artificial Intelligence in Breast Cancer Screening, Journal of Student Research 13 (2) (may 2024). doi: 10.47611/jsrhs.v13i2.6530. URL https://www.jsr.org/hs/index.php/path/article/view/6530 [DOI] [Google Scholar]
- [114].Rao MN, Reddy LV, Ganesh D, Reddy LU, Mohana RM, Usage of Computer Aided Methods for Detection and Evaluation of Breast Cancer in Mammograms, in: 2024 5th International Conference on Smart Electronics and Communication (ICOSEC), 2024, pp. 1884–1889. doi: 10.1109/ICOSEC61587.2024.10722661. [DOI] [Google Scholar]
- [115].Jannatdoust P, Valizadeh P, Saeedi N, Valizadeh G, Salari HM, Saligheh Rad H, Gity M, Computer-Aided Detection (CADe) and Segmentation Methods for Breast Cancer Using Magnetic Resonance Imaging (MRI), Journal of Magnetic Resonance Imaging n/a (n/a), _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/jmri.29687 (2025). doi: 10.1002/jmri.29687. URL https://onlinelibrary.wiley.com/doi/abs/10.1002/jmri.29687 [DOI] [PubMed] [Google Scholar]
- [116].Chan H-P, Doi K, Vyborny CJ, Schaefer TM, Wagman LR, Schmidt RA, Computerized detection of microcalcifications in mammograms: Ai in diagnosis and decision making, Radiographics 15 (3) (1995) 653–663.7624570 [Google Scholar]
- [117].Suzuki K, Overview of deep learning in medical imaging, Radiological Physics and Technology 10 (3) (2017) 257–273. [DOI] [PubMed] [Google Scholar]
- [118].Giger ML, Doi K, MacMahon H, Nishikawa RM, Schmidt RA, Computerized detection of clustered microcalcifications in digitized mammograms, Medical Physics 17 (4) (1990) 601–610. [Google Scholar]
- [119].Elter M, Horsch A, Cadx of mammographic masses and clustered microcalcifications: A review, Medical Physics 34 (6Part1) (2007) 2052–2068. [DOI] [PubMed] [Google Scholar]
- [120].Dutra I, Medeiros M, Almeida J, de Medeiros F, Computer-aided diagnosis for breast cancer: Review, Computational and Mathematical Methods in Medicine 2018 (2018) 1–17. [Google Scholar]
- [121].Heath M, Bowyer K, Kopans D, Moore R, Kegelmeyer P, The digital database for screening mammography, Proceedings of the 5th International Workshop on Digital Mammography (2001) 212–218. [Google Scholar]
- [122].Dada EG, Oyewola DO, Misra S, Computer-aided diagnosis of breast cancer from mammogram images using deep learning algorithms, Journal of Electrical Systems and Information Technology 11 (1) (2024) 38. doi: 10.1186/s43067-024-00164-y. URL https://doi.org/10.1186/s43067-024-00164-y [DOI] [Google Scholar]
- [123].Dihmani H, Bousselham A, Bouattane O, A New Computer-Aided Diagnosis System for Breast Cancer Detection from Thermograms Using Metaheuristic Algorithms and Explainable AI, Algorithms 17 (10) (2024). doi: 10.3390/a17100462. URL https://www.mdpi.com/1999-4893/17/10/462 [DOI] [Google Scholar]
- [124].Lorenzo-Ramírez MA, Pérez-Alvarado VM, Ramírez-Villavicencio FP, Sánchez-Cárdenas J, Orozco-Padilla AT, Avilés-Rodríguez GJ, Computer aided diagnosis in mammograms for breast cancer screening: Diagnóstico asistido por computadora en mamografías para detección del cáncer de mama, Revista de la Facultad de Medicina Humana 24 (4) (2024) 159–179. doi: 10.25176/RFMH.v24i4.6554. URL https://revistas.urp.edu.pe/index.php/RFMH/article/view/6554 [DOI] [Google Scholar]
- [125].Moroz-Dubenco C, Bajcsi A, Andreica A, Chira C, Towards an interpretable breast cancer detection and diagnosis system, Computers in Biology and Medicine 185 (2025) 109520. doi: 10.1016/j.compbiomed.2024.109520. URL https://www.sciencedirect.com/science/article/pii/S0010482524016056 [DOI] [PubMed] [Google Scholar]
- [126].Carriero A, Groenhoff L, Vologina E, Basile P, Albera M, Deep learning in breast cancer imaging: State of the art and recent advancements in early 2024, Diagnostics 14 (8) (2024) 848. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [127].Kaklamanis MM, Filippakis ME, A comparative survey of machine learning classification algorithms for breast cancer detection, in: Proceedings of the 23rd pan-hellenic conference on informatics, 2019, pp. 97–103. [Google Scholar]
- [128].Khan W, Leem S, See KB, Wong JK, Zhang S, Fang R, A comprehensive survey of foundation models in medicine, arXiv preprint arXiv:2406.10729 (2024). [DOI] [PubMed] [Google Scholar]
- [129].de Moura LV, Ravazio R, Mattjie C, Kupssinskü LS, Freitas CMDS, Barros RC, Unlocking the potential of vision-language models for mammography analysis, in: 2024 IEEE International Symposium on Biomedical Imaging (ISBI), IEEE, 2024, pp. 1–4. [Google Scholar]
- [130].Kartikasari P, Utami IT, Suparti S, Rahman SDF, Breast cancer classification using support vector machine (svm) and light gradient boosting machine (lightgbm) models, MEDIA STATISTIKA 16 (2) (2023) 182–193. [Google Scholar]
- [131].Liu Z, Improving breast cancer classification using histopathology images through deep learning, Applied and Computational Engineering 73 (2024) 187–196. [Google Scholar]
- [132].Hassan MM, Hassan MM, Yasmin F, Khan MAR, Zaman S, Islam KK, Bairagi AK, et al. , A comparative assessment of machine learning algorithms with the least absolute shrinkage and selection operator for breast cancer detection and prediction, Decision Analytics Journal 7 (2023) 100245. [Google Scholar]
- [133].Priya K, Senthilkumar V, Isaac JS, Kottu S, Ramakrishna V, Kumar MJ, Breast cancer segmentation by k-means and classification by machine learning, in: 2022 International Conference on Automation, Computing and Renewable Systems (ICACRS), IEEE, 2022, pp. 651–656. [Google Scholar]
- [134].Khandezamin Z, Naderan M, Rashti MJ, Detection and classification of breast cancer using logistic regression feature selection and gmdh classifier, Journal of Biomedical Informatics 111 (2020) 103591. [DOI] [PubMed] [Google Scholar]
- [135].Zuluaga-Gomez J, Breast cancer diagnosis using machine learning techniques, arXiv preprint arXiv:2305.02482 (2023). [Google Scholar]
- [136].Yalavarthi S, Makkapati SS, Murari H, Balamurugan K, Rajendran P, Advanced breast cancer diagnostics through a comparative analysis of svm, random forests, and neural networks in mri image analysis, in: 2024 Asian Conference on Communication and Networks (ASIANComNet), IEEE, 2024, pp. 1–7. [Google Scholar]
- [137].Hassan NS, Abdulazeez AM, Zeebaree DQ, Hasan DA, Medical images breast cancer segmentation based on k-means clustering algorithm: a review, Asian Journal of Research in Computer Science 9 (1) (2021) 23–38. [Google Scholar]
- [138].Abdulla S, Sagheer A, Veisi H, Breast cancer segmentation using k-means clustering and optimized region-growing technique (2022).
- [139].Utreja B, Sharma R, Wason A, A survey on segmentation techniques for breast cancer detection, ECS Transactions 107 (1) (2022) 6703. [Google Scholar]
- [140].Liu H, Cui G, Luo Y, Guo Y, Zhao L, Wang Y, Subasi A, Dogan S, Tuncer T, Artificial intelligence-based breast cancer diagnosis using ultrasound images and grid-based deep feature generator, International Journal of General Medicine (2022) 2271–2282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [141].Yurttakal AH, Erbay H, İkizceli T, Karaçavuş S, Detection of breast cancer via deep convolution neural networks using mri images, Multimedia Tools and Applications 79 (21) (2020) 15555–15573. [Google Scholar]
- [142].Kavitha T, Mathai PP, Karthikeyan C, Ashok M, Kohar R, Avanija J, Neelakandan S, Deep learning based capsule neural network model for breast cancer diagnosis using mammogram images, Interdisciplinary Sciences: Computational Life Sciences (2021) 1–17. [DOI] [PubMed] [Google Scholar]
- [143].Polat DS, Nguyen S, Karbasi P, Hulsey K, Cobanoglu MC, Wang L, Montillo A, Dogan BE, Machine learning prediction of lymph node metastasis in breast cancer: Performance of a multi-institutional mri-based 4d convolutional neural network, Radiology: Imaging Cancer 6 (3) (2024) e230107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [144].Baghel SS, Manjhvar AK, Breast cancer prediction using deep learning technique, International Research Journal of Modernization in Engineering Technology and Science (2023). [Google Scholar]
- [145].Akhtar DN, Pant DH, Dwivedi A, Jain V, Perwej DY, A breast cancer diagnosis framework based on machine learning, Int. J. Sci. Res. Sci. Eng. Technol (2023) 118–132. [Google Scholar]
- [146].He C, Diao Y, Ma X, Yu S, He X, Mao G, Wei X, Zhang Y, Zhao Y, A vision transformer network with wavelet-based features for breast ultrasound classification, Image Analysis and Stereology 43 (2) (2024) 185–194. [Google Scholar]
- [147].Gella V, High-performance classification of breast cancer histopathological images using fine-tuned vision transformers on the breakhis dataset, bioRxiv (2024) 2024–08. [Google Scholar]
- [148].Shiri M, Reddy MP, Sun J, Supervised contrastive vision transformer for breast histopathological image classification, in: 2024 IEEE International Conference on Information Reuse and Integration for Data Science (IRI), IEEE, 2024, pp. 296–301. [Google Scholar]
- [149].Wijaya BA, Hulu M, Resel R, Halawa N, Tarigan AA, Classification of breast cancer with transfer learning on convolutional neural network models, Sinkron: jurnal dan penelitian teknik informatika 8 (3) (2024) 1715–1723. [Google Scholar]
- [150].Katar O, Yıldırım Ö, Breast cancer segmentation from ultrasound images using resnext-based u-net model, Bitlis Eren Üniversitesi Fen Bilimleri Dergisi 12 (3) (2023) 871–886. [Google Scholar]
- [151].Ramalakshmi E, Gunisetti L, et al. , Different u-net variants for segmentation of histological breast images: An analytical comparison, International Journal of Computer Information Systems and Industrial Management Applications (jul 2024). [Google Scholar]
- [152].Gupta V, Wadhawan S, Kumar V, Sethi H, Gupta G, Breast cancer classification with ann and dbn, in: 2024 2nd International Conference on Advancement in Computation & Computer Technologies (InCACCT), IEEE, 2024, pp. 246–251. [Google Scholar]
- [153].Tagnamas J, Ramadan H, Yahyaouy A, Tairi H, Joining cnns and transformer networks for enhanced breast ultrasound image segmentation, in: 2024 International Conference on Intelligent Systems and Computer Vision (ISCV), IEEE, 2024, pp. 1–6. [Google Scholar]
- [154].Maistry B, Ezugwu AE, Breast cancer detection and diagnosis: A comparative study of state-of-the-arts deep learning architectures, arXiv preprint arXiv:2305.19937 (2023). [Google Scholar]
- [155].Rao A, Kim J, Kamineni M, Pang M, Lie W, Dreyer KJ, Succi MD, Evaluating gpt as an adjunct for radiologic decision making: Gpt-4 versus gpt-3.5 in a breast imaging pilot, Journal of the American College of Radiology 20 (10) (2023) 990–997. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [156].Choi HS, Song JY, Shin KH, Chang JH, Jang B-S, Developing prompts from large language model for extracting clinical information from pathology and ultrasound reports in breast cancer, Radiation Oncology Journal 41 (3) (2023) 209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [157].Jeblick K, Schachtner B, Dexl J, Mittermeier A, Stüber AT, Topalis J, Weber T, Wesp P, Sabel BO, Ricke J, et al. , Chatgpt makes medicine easy to swallow: an exploratory case study on simplified radiology reports, European radiology 34 (5) (2024) 2817–2825. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [158].Wu J, Hicks C, Breast cancer type classification using machine learning, Journal of personalized medicine 11 (2) (2021) 61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [159].Oladimeji OO, Mcloughlin I, Unnikrishnan S, Deep learning for breast cancer diagnosis: A bibliometric analysis and future research directions, Computational and Structural Biotechnology Reports (2024) 100004. [Google Scholar]
- [160].Amoroso N, Pomarico D, Fanizzi A, Didonna V, Giotta F, La Forgia D, Latorre A, Monaco A, Pantaleo E, Petruzzellis N, et al. , A roadmap towards breast cancer therapies supported by explainable artificial intelligence, Applied Sciences 11 (11) (2021) 4881. [Google Scholar]
- [161].Carriero A, Groenhoff L, Vologina E, Basile P, Albera M, Deep learning in breast cancer imaging: State of the art and recent advancements in early 2024, Diagnostics 14 (8) (2024) 848. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [162].Ahmad J, Akram S, Jaffar A, Ali Z, Bhatti SM, Ahmad A, ur Rehman S, Deep learning empowered breast cancer diagnosis: Advancements in detection and classification, PLOS ONE 19 (7) (2024) e0304757–e0304757. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [163].Hassan SA, Sayed MS, Abdalla MI, Rashwan MA, Breast cancer masses classification using deep convolutional neural networks and transfer learning, Multimedia Tools and Applications 79 (41) (2020) 30735–30768. [Google Scholar]
- [164].Beghriche T, Djerioui M, Brik Y, Attallah B, Deep-modified transfer learning-based cnn networks for enhanced breast cancer prediction, Studies in Engineering and Exact Sciences 5 (2) (2024) e5383–e5383. [Google Scholar]
- [165].Abdelhafiz D, Yang C, Ammar R, Nabavi S, Deep convolutional neural networks for mammography: advances, challenges and applications, BMC Bioinformatics 20 (11) (2019) 281. doi: 10.1186/s12859-019-2823-4. URL https://doi.org/10.1186/s12859-019-2823-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [166].Zakareya S, Izadkhah H, Karimpour J, A new deep-learning-based model for breast cancer diagnosis from medical images, Diagnostics 13 (11) (2023) 1944. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [167].Shen W-J, Zhou H-X, He Y, Xing W, Predicting female breast cancer by artificial intelligence: Combining clinical information and bi-rads ultrasound descriptors, WFUMB Ultrasound Open 1 (2) (2023) 100013. [Google Scholar]
- [168].Salgarkar VS, AK I, Multi-class breast cancer classification from digital mammograms using vision transformers, in: Proceedings of the 2023 Fifteenth International Conference on Contemporary Computing, 2023, pp. 426–436. [Google Scholar]
- [169].Tummala S, Kim J, Kadry S, Breast-net: Multi-class classification of breast cancer from histopathological images using ensemble of swin transformers, Mathematics 10 (21) (2022). [Google Scholar]
- [170].Baroni GL, Rasotto L, Roitero K, Tulisso A, Di Loreto C, Della Mea V, Optimizing vision transformers for histopathology: Pretraining and normalization in breast cancer classification, Journal of Imaging 10 (5) (2024) 108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [171].Müller-Franzes G, Müller-Franzes F, Huck L, Raaff V, Kemmer E, Khader F, Arasteh ST, Lemainque T, Kather JN, Nebelung S, et al. , Fibroglandular tissue segmentation in breast mri using vision transformers: a multi-institutional evaluation, Scientific Reports 13 (1) (2023) 14207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [172].Thisanke H, Deshan C, Chamith K, Seneviratne S, Vidanaarachchi R, Herath D, Semantic segmentation using vision transformers: A survey, Engineering Applications of Artificial Intelligence 126 (2023) 106669. [Google Scholar]
- [173].Ayana G, Choe S.-w., Vision transformers-based transfer learning for breast mass classification from multiple diagnostic modalities, Journal of Electrical Engineering & Technology 19 (5) (2024) 3391–3410. [Google Scholar]
- [174].Kassis I, Lederman D, Ben-Arie G, Giladi Rosenthal M, Shelef I, Zigel Y, Detection of breast cancer in digital breast tomosynthesis with vision transformers, Scientific Reports 14 (1) (2024) 22149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [175].Abimouloud ML, Bensid K, Elleuch M, Aiadi O, Kherallah M, Mammography breast cancer classification using vision transformers, in: International Conference on Intelligent Systems Design and Applications, Springer, 2023, pp. 452–461. [Google Scholar]
- [176].Dixon J, Akinniyi O, Abdelhamid A, Saleh GA, Rahman MM, Khalifa F, A hybrid learning-architecture for improved brain tumor recognition, Algorithms 17 (6) (2024) 221. [Google Scholar]
- [177].Abimouloud ML, Bensid K, Elleuch M, Ammar MB, Kherallah M, Advancing breast cancer diagnosis: token vision transformers for faster and accurate classification of histopathology images, Visual Computing for Industry, Biomedicine, and Art 8 (1) (2025) 1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [178].Ahmed S, Elazab N, El-Gayar MM, Elmogy M, Fouda YM, Multi-scale vision transformer with optimized feature fusion for mammographic breast cancer classification, Diagnostics 15 (11) (2025) 1361. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [179].Batool Z, Kamal MA, Shen B, Advancements in triple-negative breast cancer sub-typing, diagnosis and treatment with assistance of artificial intelligence: a focused review, Journal of Cancer Research and Clinical Oncology 150 (8) (2024) 383. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [180].Naidu G, Zuva T, Sibanda EM, A review of evaluation metrics in machine learning algorithms, in: Computer Science On-line Conference, Springer, 2023, pp. 15–25. [Google Scholar]
- [181].Erickson BJ, Kitamura F, Magician’s corner: 9. performance metrics for machine learning models (2021). [DOI] [PMC free article] [PubMed]
- [182].Geng S, Analysis of the different statistical metrics in machine learning, Highlights in Science, Engineering and Technology 88 (2024) 350–356. [Google Scholar]
- [183].Aina J, Akinniyi O, Rahman MM, Odero-Marah V, Khalifa F, A hybrid learning-architecture for mental disorder detection using emotion recognition, IEEE Access; (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [184].Chicco D, Jurman G, The advantages of the matthews correlation coefficient (mcc) over f1 score and accuracy in binary classification evaluation, BMC genomics 21 (2020) 1–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [185].Rainio O, Teuho J, Klén R, Evaluation metrics and statistical tests for machine learning, Scientific Reports 14 (1) (2024) 6086. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [186].Kushwaha S, Sathish P, Thankam T, Rajkumar K, Kumar MD, Gadde SS, Segmentation of breast cancer from mammogram images using fuzzy clustering approach, in: 2024 International Conference on Advances in Computing, Communication and Applied Informatics (ACCAI), IEEE, 2024, pp. 1–6. [Google Scholar]
- [187].Krasnov D, Davis D, Malott K, Chen Y, Shi X, Wong A, Fuzzy c-means clustering: A review of applications in breast cancer detection, Entropy 25 (7) (2023) 1021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [188].Jasim WN, Mohammed RJ, A survey on segmentation techniques for image processing., Iraqi Journal for Electrical & Electronic Engineering 17 (2) (2021). [Google Scholar]
- [189].Weingart M, Vascan O, Image segmentation processing-some techniques and experimental results a comparative study of the concepts of some segmentation techniques, in: 2013 4th International Symposium on Electrical and Electronics Engineering (ISEEE), IEEE, 2013, pp. 1–6. [Google Scholar]
- [190].Abdel-Basset M, Mohamed R, Abouhawwash M, Askar SS, Tantawy AA, An efficient multilevel threshold segmentation method for breast cancer imaging based on metaheuristics algorithms: Analysis and validations, International Journal of Computational Intelligence Systems 16 (1) (2023) 101. [Google Scholar]
- [191].Long J, Shelhamer E, Darrell T, Fully convolutional networks for semantic segmentation, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 3431–3440. [DOI] [PubMed] [Google Scholar]
- [192].Albarracin J, Cano F, Romero E, Cruz-Roa A, A comparative analysis between two convolutional networks architectures for semantic segmentation of histopathology breast cancer images, in: 2023 19th International Symposium on Medical Information Processing and Analysis (SIPAIM), IEEE, 2023, pp. 1–5. [Google Scholar]
- [193].Karri C, Santinha J, Papanikolaou N, A short review: Semantic segmentation for breast cancer detection in mri images, IETE Journal of Research 70 (6) (2024) 5666–5680. [Google Scholar]
- [194].Deepak GD, Bhat SK, A comparative study of breast tumour detection using a semantic segmentation network coupled with different pretrained cnns, Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization 12 (1) (2024) 2373996. [Google Scholar]
- [195].Zhang H, Lian J, Yi Z, Wu R, Lu X, Ma P, Ma Y, Hau-net: Hybrid cnn-transformer for breast ultrasound image segmentation, Biomedical Signal Processing and Control 87 (2024) 105427. [Google Scholar]
- [196].Anari S, de Oliveira GG, Ranjbarzadeh R, Alves AM, Vaz GC, Bendechache M, Efficientunetvit: Efficient breast tumor segmentation utilizing unet architecture and pretrained vision transformer, Bioengineering 11 (9) (2024) 945. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [197].Iqbal A, Sharif M, Bts-st: Swin transformer network for segmentation and classification of multimodality breast cancer images, Knowledge-Based Systems 267 (2023) 110393. [Google Scholar]
- [198].Wang Z, Berman M, Rannen-Triki A, Torr P, Tuia D, Tuytelaars T, Gool LV, Yu J, Blaschko M, Revisiting evaluation metrics for semantic segmentation: Optimization and evaluation of fine-grained intersection over union, Advances in Neural Information Processing Systems 36 (2024). [Google Scholar]
- [199].Cruz-Roa A, Basavanhally A, González F, Gilmore H-P, Feldman M, Ganesan S, Shih N-Y, Tomaszewski J, Madabhushi A, Automatic detection of invasive ductal carcinoma in whole slide images with convolutional neural networks, in: Medical Imaging 2014: Digital Pathology, Vol. 9041, SPIE, 2014, p. 904103. [Google Scholar]
- [200].Jin X, Demirel E, Zhu M, Kharazi A, Redmond SJ, Ward L, O’Keefe F, Neate S, McGuckin M, O’Neal DN, O’Carroll SJ, McCague P, Warby A-C, Caixeta L, Baade PD, Moore MW, Smith M, McColl SR, Predicting breast cancer recurrence using ensemble machine learning on multi-modal data, Cancers 12 (6) (2020) 1548.32545446 [Google Scholar]
- [201].Nakach A, Raphaël J, Aigrain H, Frouin V, Deep learning–based multimodal fusion: health applications, challenges and opportunities, IEEE Journal of Biomedical and Health Informatics 27 (2) (2023) 885–898. [Google Scholar]
- [202].Honarmandi Shandiz A, Kiranyaz S, Gabbouj M, Parag T, Iosifidis A, Cross-modal attention for breast cancer subtype classification using histopathology images and gene expression data, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2022, pp. 2429–2436. [Google Scholar]
- [203].Chen RJ, Lu MY, Wang J, Williamson DF, Rodig SJ, Lindeman NI, Mahmood F, Pathomic fusion: An integrated framework for fusing histopathology and genomic features for cancer diagnosis and prognosis, IEEE Transactions on Medical Imaging 41 (3) (2022) 757–770. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [204].Unger J, Kather JN, Integrating histopathology and molecular data in cancer research: promises and challenges of multimodal data analysis, British Journal of Cancer 126 (2022) 390–397. [Google Scholar]
- [205].Hamoudi Y, Nguyen-Ngoc A-Q, Amin MN, Xuan J, Iosifidis A, Kiranyaz S, Multimodal survival analysis with deep attention models for integrating histopathology, genomics, and clinical data, Briefings in Bioinformatics 24 (1) (2023) bbac604.36573486 [Google Scholar]
- [206].Lu MY, Williamson DFK, Chen TY, Chen RJ, Barbieri M, Mahmood F, Data-efficient and weakly supervised computational pathology on whole-slide images, Nature Biomedical Engineering 5 (2021) 555–570. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [207].Wang X, Sun Z, Zhang Q, Yang W, Sun L, Mo-mpnet: Multi-omics transformer for integrative analysis of multi-omics data, Bioinformatics 38 (20) (2022) 4708–4716. [Google Scholar]
- [208].Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, Shamseer L, Tetzlaff JM, Akl EA, Brennan SE, et al. , The prisma 2020 statement: an updated guideline for reporting systematic reviews, bmj 372 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [209].Hu Q, Whitney HM, Giger ML, A deep learning methodology for improved breast cancer diagnosis using multiparametric mri, Scientific reports 10 (1) (2020) 10536. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [210].Zaalouk AM, Ebrahim GA, Mohamed HK, Hassan HM, Zaalouk MM, A deep learning computer-aided diagnosis approach for breast cancer, Bioengineering 9 (8) (2022) 391. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [211].Martinez RG, van Dongen D-M, Pre-screening breast cancer with machine learning and deep learning, arXiv preprint arXiv:2302.02406 (2023). [Google Scholar]
- [212].Haider SA, Pressman SM, Borna S, Gomez-Cabello CA, Sehgal A, Leibovich BC, Forte AJ, Evaluating large language model (llm) performance on established breast classification systems, Diagnostics 14 (14) (2024) 1491. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [213].Sorin V, Klang E, Sklair-Levy M, Cohen I, Zippel DB, Balint Lahat N, Konen E, Barash Y, Large language model (chatgpt) as a support tool for breast tumor board, NPJ Breast Cancer 9 (1) (2023) 44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [214].Haver HL, Ambinder EB, Bahl M, Oluyemi ET, Jeudy J, Yi PH, Appropriateness of breast cancer prevention and screening recommendations provided by chatgpt, Radiology 307 (4) (2023) e230424. [DOI] [PubMed] [Google Scholar]
- [215].Guo Y-Y, Huang Y-H, Wang Y, Huang J, Lai Q-Q, Li Y-Z, Breast mri tumor automatic segmentation and triple-negative breast cancer discrimination algorithm based on deep learning, Computational and Mathematical Methods in Medicine 2022 (1) (2022) 2541358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [216].Rossi J, Mullen LA, Oluyemi ET, Panigrahi B, Myers KS, DiCarlo P, Ambinder EB, Patient utilization of weekend/evening appointments for screening mammography: An 8-year observational cohort study, Journal of the American College of Radiology (2024). [DOI] [PubMed] [Google Scholar]
- [217].Wheeler SB, Reeder-Hayes KE, Carey LA, Disparities in breast cancer treatment and outcomes: biological, social, and health system determinants and opportunities for research, The oncologist 18 (9) (2013) 986–993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [218].Huq M, Woodard N, Okwara L, McCarthy S, Knott C, Recommendations for breast cancer education for african american women below screening age, Health Education Research 36 (5) (2021) 530–540. [DOI] [PubMed] [Google Scholar]
- [219].Fairley R, Lillard JW Jr, Berk A, Cornew S, Gaspero J, Gillespie J, Horne LL, Kidane S, Munro SB, Parsons M, et al. , Increasing clinical trial participation of black women diagnosed with breast cancer, Journal of Racial and Ethnic Health Disparities 11 (3) (2024) 1701–1717. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [220].Coughlin SS, Social determinants of breast cancer risk, stage, and survival, Breast cancer research and treatment 177 (2019) 537–548. [DOI] [PubMed] [Google Scholar]
- [221].Saini G, Ogden A, McCullough LE, Torres M, Rida P, Aneja R, Disadvantaged neighborhoods and racial disparity in breast cancer outcomes: the biological link, Cancer Causes & Control 30 (2019) 677–686. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [222].Minas TZ, Kiely M, Ajao A, Ambs S, An overview of cancer health disparities: new approaches and insights and why they matter, Carcinogenesis 42 (1) (2021) 2–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [223].Gehlert S, Hudson D, Sacks T, A critical theoretical approach to cancer disparities: breast cancer and the social determinants of health, Frontiers in public health 9 (2021) 674736. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [224].Wilson IB, Cleary PD, Linking clinical variables with health-related quality of life: a conceptual model of patient outcomes, Jama 273 (1) (1995) 59–65. [PubMed] [Google Scholar]
- [225].Wilkinson L, Gathani T, Understanding breast cancer as a global health concern, The British journal of radiology 95 (1130) (2022) 20211033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [226].Yap Y-S, Outcomes in breast cancer—does ethnicity matter?, ESMO open 8 (3) (2023) 101564. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [227].Giaquinto AN, Sung H, Newman LA, Freedman RA, Smith RA, Star J, Jemal A, Siegel RL, Breast cancer statistics 2024, CA: A Cancer Journal for Clinicians 74 (6) (2024) 477–495, _eprint: https://acsjournals.onlinelibrary.wiley.com/doi/pdf/10.3322/caac.21863. doi: 10.3322/caac.21863. URL https://acsjournals.onlinelibrary.wiley.com/doi/abs/10.3322/caac.21863 [DOI] [PubMed] [Google Scholar]
- [228].Tsapatsaris A, Babagbemi K, Reichman MB, Barriers to breast cancer screening are worsened amidst COVID-19 pandemic: A review, Clinical Imaging 82 (2022) 224–227. doi: 10.1016/j.clinimag.2021.11.025. URL https://www.sciencedirect.com/science/article/pii/S0899707121004563 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [229].Hiatt RA, Brody JG, Environmental determinants of breast cancer, Annual review of public health 39 (1) (2018) 113–133. [DOI] [PubMed] [Google Scholar]
- [230].Grabinski VF, Brawley OW, Disparities in breast cancer, Obstetrics and Gynecology Clinics 49 (1) (2022) 149–165. [DOI] [PubMed] [Google Scholar]
- [231].Yedjou CG, Sims JN, Miele L, Noubissi F, Lowe L, Fonseca DD, Alo RA, Payton M, Tchounwou PB, Health and racial disparity in breast cancer, Breast cancer metastasis and drug resistance: Challenges and progress (2019) 31–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [232].Bhatia S, Landier W, Paskett ED, Peters KB, Merrill JK, Phillips J, Osarogiagbon RU, Rural–urban disparities in cancer outcomes: opportunities for future research, JNCI: Journal of the National Cancer Institute 114 (7) (2022) 940–952. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [233].Obeng-Gyasi S, Obeng-Gyasi B, Tarver W, Breast cancer disparities and the impact of geography, Surgical Oncology Clinics 31 (1) (2022) 81–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [234].Togawa K, Anderson BO, Foerster M, Galukande M, Zietsman A, Pontac J, Anele A, Adisa C, Parham G, Pinder LF, et al. , Geospatial barriers to healthcare access for breast cancer diagnosis in sub-saharan african settings: The african breast cancer—disparities in outcomes cohort study, International journal of cancer 148 (9) (2021) 2212–2226. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [235].Mascara M, Constantinou C, Global Perceptions of Women on Breast Cancer and Barriers to Screening, Current Oncology Reports 23 (7) (2021) 74. doi: 10.1007/s11912-021-01069-z. URL https://doi.org/10.1007/s11912-021-01069-z [DOI] [PubMed] [Google Scholar]
- [236].Pakseresht S, Tavakolinia S, Leili EK, Determination of the Association between Perceived Stigma and Delay in Help-Seeking Behavior of Women with Breast Cancer., Maedica 16 (3) (2021) 458–462, place: Romania. doi: 10.26574/maedica.2021.16.3.463. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [237].Jabeen S, Zakar R, Zakar MZ, Fischer F, Experiences of family caregivers in dealing with cases of advanced breast cancer: a qualitative study of the sociocultural context in Punjab, Pakistan, BMC Public Health 24 (1) (2024) 1030. doi: 10.1186/s12889-024-18404-1. URL https://doi.org/10.1186/s12889-024-18404-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [238].Nambi Namusisi H, Traditional cancer treatments in west africa: Historical practices, current approaches, and future directions, Research Output Journal of Biological and Applied Science (2024). [Google Scholar]
- [239].Afaya A, Anaba EA, Bam V, Afaya RA, Yahaya A-R, Seidu A-A, Ahinkorah BO, Socio-cultural beliefs and perceptions influencing diagnosis and treatment of breast cancer among women in Ghana: a systematic review, BMC Women’s Health 24 (1) (2024) 288. doi: 10.1186/s12905-024-03106-y. URL https://doi.org/10.1186/s12905-024-03106-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- [240].Nwozichi CU, Ojewale MO, Salako O, Brotobor D, Olaogun E, The lived experience of suffering by nigerian female breast cancer survivors: A phenomenological perspective, Journal of Patient Experience 12 (2025) 23743735251314858. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [241].Keegan G, Rizzo J-R, Joseph K-A, Disparities in breast cancer among patients with disabilities: care gaps, accessibility, and best practices, JNCI: Journal of the National Cancer Institute 115 (10) (2023) 1139–1144. [DOI] [PubMed] [Google Scholar]
- [242].Organi ZK, Nazarenia M, Aghaee F, Understanding the challenges of language barriers in healthcare, Interdisciplinary Studies in Society, Law, and Politics 3 (3) (2024) 28–35. [Google Scholar]
- [243].Krishnamurthy N, Smith CB, Lin JJ, Odom JN, Mazor M, “if we don’t speak the language, we aren’t offered the same opportunities”: Qualitative perspectives of palliative care coordination for women of color living with metastatic breast cancer. (2023). [DOI] [PMC free article] [PubMed]
- [244].Bourgeois A, Horrill TC, Mollison A, Lambert LK, Stajduhar KI, Barriers to cancer treatment and care for people experiencing structural vulnerability: a secondary analysis of ethnographic data, International Journal for Equity in Health 22 (1) (2023) 58. doi: 10.1186/s12939-023-01860-3. URL https://doi.org/10.1186/s12939-023-01860-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [245].Mouslim MC, Johnson RM, Dean LT, Healthcare system distrust and the breast cancer continuum of care, Breast cancer research and treatment 180 (2020) 33–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [246].Falcone M, Salhia B, Hughes Halbert C, Roussos Torres ET, Stewart D, Stern MC, Lerman C, Impact of structural racism and social determinants of health on disparities in breast cancer mortality, Cancer Research 84 (23) (2024) 3924–3935. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [247].Zaveri S, Nevid D, Ru M, Moshier E, Pisapati K, Reyes SA, Port E, Romanoff A, Racial Disparities in Time to Treatment Persist in the Setting of a Comprehensive Breast Center, Annals of Surgical Oncology 29 (11) (2022) 6692–6703. doi: 10.1245/s10434-022-11971-w. URL https://doi.org/10.1245/s10434-022-11971-w [DOI] [PubMed] [Google Scholar]
- [248].Wieder R, Adam N, Racial disparities in breast cancer treatments and adverse events in the seer-medicare data, Cancers 15 (17) (2023) 4333. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [249].Stabellini N, Cullen J, Cao L, Shanahan J, Hamerschlak N, Waite K, Barnholtz-Sloan JS, Montero AJ, Racial disparities in breast cancer treatment patterns and treatment related adverse events, Scientific reports 13 (1) (2023) 1233. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [250].Anstey EH, Shoemaker ML, Barrera CM, O’Neil ME, Verma AB, Holman DM, Breastfeeding and breast cancer risk reduction: implications for black mothers, American journal of preventive medicine 53 (3) (2017) S40–S46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [251].Johnson AM, Kirk R, Muzik M, Overcoming workplace barriers: A focus group study exploring african american mothers’ needs for workplace breastfeeding support, Journal of Human Lactation 31 (3) (2015) 425–433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [252].Howard FM, Olopade OI, Epidemiology of triple-negative breast cancer: a review, The Cancer Journal 27 (1) (2021) 8–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [253].Taylor JG, Examining risk factors for early-onset breast cancer, Ph.D. thesis, University of South Carolina; (2024). [Google Scholar]
- [254].Thomas A, Reis-Filho JS, Geyer CE, Wen HY, Rare subtypes of triple negative breast cancer: Current understanding and future directions, npj Breast Cancer 9 (1) (2023) 55. doi: 10.1038/s41523-023-00554-x. URL https://doi.org/10.1038/s41523-023-00554-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- [255].Calip GS, Hoskins KF, Guadamuz JS, Examining the associations among treatment declination, racial and ethnic inequities, and breast cancer survival, JAMA Network Open 7 (5) (2024) e249402–e249402. [DOI] [PubMed] [Google Scholar]
- [256].Proskuriakova E, Aryal BB, Shrestha D, Adams M, Valencia S, Varsani A, Sakhuja A, Jasaraj RB, Kovalenko I, Verda L, et al. , Ethnic disparities in breast cancer: A comprehensive analysis of socio-demographic and clinical characteristics among african american, hispanic, and caucasian women. (2024).
- [257].Ferreira CS, Rodrigues J, Moreira S, Ribeiro F, Longatto-Filho A, Breast cancer screening adherence rates and barriers of implementation in ethnic, cultural and religious minorities: a systematic review, Molecular and clinical oncology 15 (1) (2021) 139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [258].Saldaña-Téllez M, Meneses-Navarro S, Cano-Garduño L, Unger-Saldaña K, Barriers and facilitators for breast cancer early diagnosis in an indigenous community in mexico: voices of otomí women, BMC Women’s Health 24 (1) (2024) 33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [259].Pederson HJ, Al-Hilli Z, Kurian AW, Racial disparities in breast cancer risk factors and risk management, Maturitas (2024) 107949. [DOI] [PubMed] [Google Scholar]
- [260].Ahmad IA, A scoping review of barriers within the cancer care continuum: Addressing screening, diagnosis, and treatment disparities for an equitable future, Advances in Medicine, Psychology, and Public Health 2 (2) (2025) 96–106. [Google Scholar]
- [261].Elmohr MM, Javed Z, Dubey P, Jordan JE, Shah L, Nasir K, Rohren EM, Lincoln CM, Social determinants of health framework to identify and reduce barriers to imaging in marginalized communities, Radiology 310 (2) (2024) e223097. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [262].Kaplan DA, Recognizing and ameliorating provider implicit bias, in: Cultural Responsiveness in Assisted Reproductive Technology: Best Practices for Clinics and Affiliated Providers, Springer, 2024, pp. 313–331. [Google Scholar]
- [263].Chen F, Wang L, Hong J, Jiang J, Zhou L, Unmasking bias in artificial intelligence: a systematic review of bias detection and mitigation strategies in electronic health record-based models, Journal of the American Medical Informatics Association 31 (5) (2024) 1172–1183. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [264].Fazeli S, Narayan A, Mango VL, Wahab R, Mehta TS, Ojeda-Fournier H, Access to breast cancer screening: Disparities and determinants—ajr expert panel narrative review, American Journal of Roentgenology (2024). [DOI] [PubMed] [Google Scholar]
- [265].Hanna M, Pantanowitz L, Jackson B, Palmer O, Visweswaran S, Pantanowitz J, Deebajah M, Rashidi H, Ethical and bias considerations in artificial intelligence (ai)/machine learning, Modern Pathology (2024) 100686. [DOI] [PubMed] [Google Scholar]
- [266].Griffin AC, Wang KH, Leung TI, Facelli JC, Recommendations to promote fairness and inclusion in biomedical ai research and clinical use, Journal of Biomedical Informatics 157 (2024) 104693. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [267].Ahmad J, Akram S, Jaffar A, Ali Z, Bhatti SM, Ahmad A, Rehman SU, Deep learning empowered breast cancer diagnosis: Advancements in detection and classification, Plos one 19 (7) (2024) e0304757. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [268].Shamir SB, Sasson AL, Margolies LR, Mendelson DS, New frontiers in breast cancer imaging: The rise of ai, Bioengineering 11 (5) (2024) 451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [269].Holzinger A, Biemann C, Pattichis CS, Kell DB, What do we need to build explainable ai systems for the medical domain?, arXiv preprint arXiv:1712.09923 (2017). [Google Scholar]
- [270].Rudin C, Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead, Nature Machine Intelligence 1 (5) (2019) 206–215. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [271].Arrieta AB, et al. , Explainable artificial intelligence (xai): Concepts, taxonomies, opportunities and challenges toward responsible ai, Information Fusion 58 (2020) 82–115. [Google Scholar]
- [272].Simonyan K, Vedaldi A, Zisserman A, Deep inside convolutional networks: Visualising image classification models and saliency maps, in: arXiv preprint arXiv:1312.6034, 2013. [Google Scholar]
- [273].Selvaraju RR, Cogswell M, Das A, Vedantam R, Parikh D, Batra D, Grad-cam: Visual explanations from deep networks via gradient-based localization, in: Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 618–626. [Google Scholar]
- [274].Lundberg SM, Lee S-I, A unified approach to interpreting model predictions, in: Advances in Neural Information Processing Systems, Vol. 30, 2017. [Google Scholar]
- [275].Ribeiro MT, Singh S, Guestrin C, “why should i trust you?”: Explaining the predictions of any classifier, in: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 1135–1144. [Google Scholar]
- [276].Finlayson SG, Subbaswamy A, Singh K, Bowers J, Kupke J, Zittrain J, Kohane IS, Beam AL, Clinician and algorithmic bias in clinical machine learning applications, New England Journal of Medicine 386 (9) (2022) 880–885.35235730 [Google Scholar]
- [277].Holzinger A, Carrington A, Müller H, Causability and explainability of artificial intelligence in medicine, Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery 9 (4) (2019) e1312. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [278].Mishra I, Kashyap V, Yadav N, Pahwa R, Harmonizing intelligence: A holistic approach to bias mitigation in artificial intelligence (ai), International Research Journal on Advanced Engineering Hub (IRJAEH) 2 (07) (2024) 1978–1985. [Google Scholar]
- [279].Gao W, Wang D, Huang Y, Designing a deep learning-driven resource-efficient diagnostic system for metastatic breast cancer: Reducing long delays of clinical diagnosis and improving patient survival in developing countries, Cancer Informatics 22 (2023) 11769351231214446. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [280].Zhou Z, Xie X, Zhou AL, Yang Z, Nabeel M, Deng Y, Feng Z, Zheng X, Fang Z, “dr.j”: An artificial intelligence powered ultrasonography breast cancer preliminary screening solution, International Journal of Advanced Computer Science and Applications 11 (7) (2020). [Google Scholar]
- [281].Hamid MTR, Mumin N, Hamid SA, Rahmat K, Application of artificial intelligence (ai) system in opportunistic screening and diagnostic population in a middle-income nation., Current Medical Imaging (2024). [DOI] [PubMed] [Google Scholar]
- [282].Xavier D, Miyawaki IA, Jorge CAC, Moreira MJB, Carvalho BM, Batalini F, Abstract p3-04-06: Artificial intelligence-based triaging of breast cancer screening mammograms and radiologist workload reduction: a systematic review and meta-analysis, Cancer Research 83 (5_Supplement) (2023) P3–04. [Google Scholar]
- [283].Zarcaro C, Clauser P, Artificial intelligence clinical applications in breast diagnostic imaging, Journal of Radiological Review 10 (3) (2023) 127–37. [Google Scholar]
- [284].Sharma R, Tiwari AK, Bridging racial and ethnic disparities in cancer research, Cancer Reports 6 (Suppl 1) (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [285].DeWitt JT, Oropeza E, Haricharan S, Investigating the relationships between race and socioeconomic status on the biology of er+ breast cancer, Cancer Research 84 (6_Supplement) (2024) 6161–6161. [Google Scholar]
- [286].Eshun RB, Islam AK, Bikdash M, A deep convolutional neural network for the classification of imbalanced breast cancer dataset, Healthcare Analytics (2024) 100330. [Google Scholar]
- [287].Subasree S, Sakthivel N, Shobana M, Tyagi AK, Deep learning based improved generative adversarial network for addressing class imbalance classification problem in breast cancer dataset, International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems 31 (03) (2023) 387–412. [Google Scholar]
- [288].Park JI, Bozkurt S, Park JW, Lee S, Evaluation of race/ethnicity-specific survival machine learning models for hispanic and black patients with breast cancer, BMJ Health & Care Informatics 30 (1) (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [289].Smith LA, Cahill JA, Graim K, Equitable machine learning counteracts ancestral bias in precision medicine, improving outcomes for all, Research Square (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [290].Frazer HM, Tang JS, Elliott MS, Kunicki KM, Hill B, Karthik R, Kwok CF, Peña-Solorzano CA, Chen Y, Wang C, et al. , Admani: Annotated digital mammograms and associated non-image datasets, Radiology: Artificial Intelligence 5 (2) (2022) e220072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [291].Maouche I, Terrissa LS, Benmohammed K, Zerhouni N, An explainable ai approach for breast cancer metastasis prediction based on clinicopathological data, IEEE Transactions on Biomedical Engineering 70 (12) (2023) 3321–3329. [DOI] [PubMed] [Google Scholar]
- [292].Chakraborty D, Ivan C, Amero P, Khan M, Rodriguez-Aguayo C, Başağaoğlu H, Lopez-Berestein G, Explainable artificial intelligence reveals novel insight into tumor microenvironment conditions linked with better prognosis in patients with breast cancer, Cancers 13 (14) (2021) 3450. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [293].Sobhana M, Palaketi AK, Nalabothu R, Breast cancer prediction by ensembling machine learning algorithms and explainable ai, in: 2024 3rd International Conference for Innovation in Technology (INOCON), IEEE, 2024, pp. 1–6. [Google Scholar]
- [294].Nakach F-Z, Idri A, Goceri E, A comprehensive investigation of multimodal deep learning fusion strategies for breast cancer classification, Artificial Intelligence Review 57 (12) (2024) 327. [Google Scholar]
- [295].Chia JLL, He GS, Ngiam KY, Hartman M, Ng QX, Goh SSN, Harnessing artificial intelligence to enhance global breast cancer care: A scoping review of applications, outcomes, and challenges, Cancers 17 (2) (2025) 197. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [296].Jain N, Revolutionizing patient care: The impact of chatgpt and generative ai in healthcare, International Journal of Scientific and Research Publications 14 (11) (2024) 10–29322. [Google Scholar]
- [297].Amponsah D, Thamman R, Brandt E, James C, Spector-Bagdady K, Yong CM, Artificial intelligence to promote racial and ethnic cardiovascular health equity, Current Cardiovascular Risk Reports 18 (11) (2024) 153–162. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [298].Lukac S, Dayan D, Fink V, Leinert E, Hartkopf A, Veselinovic K, Janni W, Rack B, Pfister K, Heitmeir B, et al. , Evaluating chatgpt as an adjunct for the multidisciplinary tumor board decision-making in primary breast cancer cases, Archives of Gynecology and Obstetrics 308 (6) (2023) 1831–1844. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [299].Prinzi F, Insalaco M, Orlando A, Gaglio S, Vitabile S, A yolo-based model for breast cancer detection in mammograms, Cognitive Computation 16 (1) (2024) 107–120. [Google Scholar]
- [300].Al-Jabbar M, Alshahrani M, Senan EM, Ahmed IA, Multi-method diagnosis of histopathological images for early detection of breast cancer based on hybrid and deep learning, Mathematics 11 (6) (2023) 1429. [Google Scholar]
- [301].Qian X, Pei J, Han C, Liang Z, Zhang G, Chen N, Zheng W, Meng F, Yu D, Chen Y, et al. , A multimodal machine learning model for the stratification of breast cancer risk, Nature Biomedical Engineering (2024) 1–15. [DOI] [PubMed] [Google Scholar]
- [302].Schaffter T, Buist DS, Lee CI, Nikpoor N, Ribli D, Guan X, Lotter W, Parvin B, Leung J, Zuckerman J, et al. , Evaluation of combined artificial intelligence and radiologist assessment to interpret screening mammograms, JAMA Network Open 3 (3) (2020) e200265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [303].Saldanha G, Khoo B, Chou G, Moy L, Evans A, Tardivon A, et al. , Multicentre, multivendor clinical evaluation of artificial intelligence for breast cancer screening: A retrospective cohort study, The Lancet Digital Health 4 (6) (2022) e435–e447. [Google Scholar]
- [304].Kim E-K, Kim HM, Yoon JH, Lee JM, Moon HJ, Lee S, Cho EJ, Kim MJ, Kwak JY, Kim SM, et al. , Impact of artificial intelligence-based computer-aided diagnosis on the diagnostic performance of radiologists in breast ultrasonography: A multicenter study, European Radiology 32 (2) (2022) 1043–1051. [Google Scholar]
- [305].Schelb P, Kohl S, Geyer S, Truhn D, Kaul M, Meyer HJ, Maintz D, Schlamann M, Kloth J, Persigehl T, et al. , Alternative deep learning architectures for liver lesion classification in contrast-enhanced ultrasound images, European Radiology 31 (7) (2021) 4757–4767. [Google Scholar]
- [306].Langlotz CP, Allen B, Erickson BJ, Kalpathy-Cramer J, Bigelow K, Cook TS, Flanders AE, Lungren MP, Mendelson D, Rudie J, et al. , A roadmap for foundational research on artificial intelligence in medical imaging: From the 2018 nih/rsna/acr/the academy workshop, Radiology: Artificial Intelligence 1 (1) (2019) e190021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [307].W. H. Organization, Ethics and governance of artificial intelligence for health, wHO Guidance Document; (2021). [Google Scholar]
- [308].O. for Economic Co-operation, Development, Oecd framework for the classification of ai systems, oECD Publications; (2023). [Google Scholar]
- [309].Busch F, Kather JN, Johner C, Moser M, Truhn D, Adams LC, Bressem KK, Navigating the european union artificial intelligence act for healthcare, npj Digital Medicine 7 (1) (2024) 210. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [310].Van Kolfschooten H, Van Oirschot J, The eu artificial intelligence act (2024): implications for healthcare, Health Policy 149 (2024) 105152. [DOI] [PubMed] [Google Scholar]
- [311].U. FDA, H. Canada, Good machine learning practice for medical device development: Guiding principles, available from FDA website; (2021). [Google Scholar]
- [312].Mathews A, Lee J, Good machine learning practice for medical device development: Fda guidance 2022, Medical Device and Diagnostics Industry (2022). [Google Scholar]
- [313].Yang Q, Liu Y, Chen T, Tong Y, Federated machine learning: Concept and applications, ACM Transactions on Intelligent Systems and Technology 10 (2) (2019) 12. [Google Scholar]
- [314].Sheller MJ, Edwards B, Reina GA, Martin J, Pati S, Kotrotsou A, Milchenko M, Xu W, Marcus D, Colen RR, Bakas S, Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data, Scientific Reports 10 (1) (2020) 12598. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [315].Rieke N, Hancox J, Li W, Milletarì F, Roth HR, Albarqouni S, Bakas S, Galtier M, Landman BA, Maier-Hein K, et al. , The future of digital health with federated learning, npj Digital Medicine 3 (1) (2020) 119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [316].Abedini M, Tizhoosh HR, Lane CA, Martel AL, Federated learning for breast cancer detection, Scientific Reports 11 (1) (2021) 15240.34315913 [Google Scholar]
- [317].Li Y, Liu X, Qian S, Li P, Liu L, Ding D, Li X, Zhang S, Wang T, Federated learning in medical imaging: facilitating multi-institutional collaborations for breast cancer prediction, Frontiers in Oncology 13 (2023) 1173691. [Google Scholar]
- [318].Kaissis GA, Makowski MR, Rückert D, Braren RF, Secure, privacy-preserving and federated machine learning in medical imaging, Nature Machine Intelligence 2 (6) (2020) 305–311. [Google Scholar]
- [319].Gaudioso M, Giallombardo G, Miglionico G, Vocaturo E, Classification in the multiple instance learning framework via spherical separation, Soft Computing 24 (7) (2020) 5071–5077. [Google Scholar]
- [320].Astorino A, Fuduli A, Gaudioso M, Vocaturo E, et al. , Multiple instance learning algorithm for medical image classification., in: SEBD, Vol. 2400, 2019, pp. 1–8. [Google Scholar]
- [321].Sorin V, Barash Y, Konen E, Klang E, Deep-learning natural language processing for oncological applications, The Lancet. Oncology 21 (12) (2020) 1553–1556. [DOI] [PubMed] [Google Scholar]
- [322].Haider SA, Pressman SM, Borna S, Gomez-Cabello CA, Sehgal A, Leibovich BC, Forte AJ, Evaluating large language model (llm) performance on established breast classification systems, Diagnostics 14 (14) (2024) 1491. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [323].Rydzewski NR, Dinakaran D, Zhao SG, Ruppin E, Turkbey B, Citrin DE, Patel KR, Comparative evaluation of llms in clinical oncology, Nejm Ai 1 (5) (2024) AIoa2300151. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [324].Chen X, Li Y, Hu M, Salari E, Chen X, Qiu RL, Zheng B, Yang X, Mammo-clip: Leveraging contrastive language-image pre-training (clip) for enhanced breast cancer diagnosis with multi-view mammography, arXiv preprint arXiv:2404.15946 (2024). [Google Scholar]
- [325].Cao X, Ye W, Moise K, Coffee M, Mpoxvlm: A vision-language model for diagnosing skin lesions from mpox virus infection, arXiv preprint arXiv:2411.10888 (2024). [Google Scholar]
- [326].Gupta R, Pandey G, Pal SK, Automating government report generation: A generative ai approach for efficient data extraction, analysis, and visualization, Digital Government: Research and Practice (2024). [Google Scholar]
- [327].Gao Y, Fischer L, Lintner A, Ebling S, Audio description generation in the era of llms and vlms: A review of transferable generative ai technologies, arXiv preprint arXiv:2410.08860 (2024). [Google Scholar]
- [328].Yildirim N, Richardson H, Wetscherek MT, Bajwa J, Jacob J, Pinnock MA, Harris S, Coelho De Castro D, Bannur S, Hyland S, et al. , Multimodal healthcare ai: identifying and designing clinically relevant vision-language applications for radiology, in: Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems, 2024, pp. 1–22. [Google Scholar]
- [329].Li X, Peng L, Wang Y, Zhang W, Open challenges and opportunities in federated foundation models towards biomedical healthcare, arXiv preprint arXiv:2405.06784 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [330].Liu L, Yang X, Lei J, Liu X, Shen Y, Zhang Z, Wei P, Gu J, Chu Z, Qin Z, et al. , A survey on medical large language models: Technology, application, trustworthiness, and future directions, arXiv preprint arXiv:2406.03712 (2024). [Google Scholar]
- [331].April A, Kanker payudara (Jan 2023).
- [332].Raaj RS, Breast cancer detection and diagnosis using hybrid deep learning architecture, Biomedical Signal Processing and Control 82 (2023) 104558. [Google Scholar]
- [333].Lin Q, Tan W-M, Ge J-Y, Huang Y, Xiao Q, Xu Y-Y, Jin Y-T, Shao Z-M, Gu Y-J, Yan B, et al. , Artificial intelligence-based diagnosis of breast cancer by mammography microcalcification, Fundamental Research; (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [334].Shen Y, Wu N, Phang J, Park J, Liu K, Tyagi S, Heacock L, Kim SG, Moy L, Cho K, et al. , An interpretable classifier for high-resolution breast cancer screening images utilizing weakly supervised localization, Medical Image Analysis (2020) 101908. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [335].Wu N, Phang J, Park J, Shen Y, Huang Z, Zorin M, Jastrzębski S, Févry T, Katsnelson J, Kim E, Wolfson S, Parikh U, Gaddam S, Lin LLY, Ho K, Weinstein JD, Reig B, Gao Y, Toth H, Pysarenko K, Lewin A, Lee J, Airola K, Mema E, Chung S, Hwang E,Samreen N, Kim SG, Heacock L, Moy L, Cho K, Geras KJ, Deep neural networks improve radiologists’ performance in breast cancer screening, IEEE Transactions on Medical Imaging (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [336].Al-Dhabyani W, Gomaa M, Khaled H, Fahmy A, Dataset of breast ultrasound images, Data in brief 28 (2020) 104863. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [337].Kwak D, Choi J, Lee S, Rethinking breast cancer diagnosis through deep learning based image recognition, Sensors 23 (4) (2023) 2307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [338].Janowczyk A, Madabhushi A, Deep learning for digital pathology image analysis: A comprehensive tutorial with selected use cases, Journal of pathology informatics 7 (1) (2016) 29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [339].Obayya M, Maashi MS, Nemri N, Mohsen H, Motwakel A, Osman AE, Alneil AA, Alsaid MI, Hyperparameter optimizer with deep learning-based decision-support systems for histopathological breast cancer diagnosis, Cancers 15 (3) (2023) 885. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [340].Li H, Zhu C, Zhang Y, Sun Y, Shui Z, Kuang W, Zheng S, Yang L, Task-specific fine-tuning via variational information bottleneck for weakly-supervised pathology whole slide image classification, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 7454–7463. [Google Scholar]
- [341].Bejnordi B, Veta M, Van Diest PJ, Van Ginneken B, Karssemeijer N, Litjens G, Van Der Laak JA, Hermsen M, Manson QF, Balkenhol M, Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer, Jama 318 (22) (2017) 2199–2210. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [342].Petrick NA, Akbar S, Cha KH, Nofech-Mozes S, Sahiner B, Gavrielides MA, Kalpathy-Cramer J, Drukker K, Martel AL, Spieaapm-nci breastpathq challenge: an image analysis challenge for quantitative tumor cellularity assessment in breast cancer histology images following neoadjuvant treatment, Journal of Medical Imaging 8 (3) (2021) 034501. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [343].Vasilev YA, Kolsanov AV, Arzamasov KM, Vladzymyrskyy AV, Omelyanskaya OV, Semenov SS, Axenova LE, Evaluating the performance of artificial intelligence-based software for digital mammography characterization, Digital Diagnostics 5 (4) (2024) 695–711. [Google Scholar]
- [344].Oyebanji OS, Apampa AR, Idoko P, Babalola A, Ijiga OM, Afolabi O, Michael CI, Enhancing breast cancer detection accuracy through transfer learning: A case study using efficient net, World Journal of Advanced Engineering Technology and Sciences 13 (01) (2024) 285–318. [Google Scholar]
- [345].KARAGÖZ MA, NALBANTOĞLU ÖU, KARABOĞA D, Akay B, BAŞTÜRK A, Ulutabanca H, DOĞAN S, COŞKUN D, Demir O, Deep learning-based breast cancer diagnosis with multiview of mammography screening to reduce false positive recall rate, Turkish journal of electrical engineering and computer sciences 32 (3) (2024) 382–402. [Google Scholar]
- [346].Hussain L, Ansari S, Shabir M, Qureshi SA, Aldweesh A, Omar A, Iqbal Z, Bukhari SAC, Deep convolutional neural networks accurately predict breast cancer using mammograms, Waves in Random and Complex Media (2023) 1–24. [Google Scholar]
- [347].Altameem A, Mahanty C, Poonia RC, Saudagar AKJ, Kumar R, Breast cancer detection in mammography images using deep convolutional neural networks and fuzzy ensemble modeling techniques, Diagnostics 12 (8) (2022) 1812. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [348].Tang W-J, Chen S-Y, Hu W-K, Li X-L, Zheng B-J, Wang Z-S, Ding H-J, Chen L-X, Zhang Q-Q, Yu X-M, et al. , Abbreviated versus full-protocol mri for breast cancer neoadjuvant chemotherapy response assessment: diagnostic performance by general and breast radiologists, American Journal of Roentgenology 220 (6) (2023) 817–825. [DOI] [PubMed] [Google Scholar]
- [349].Niu S, Wang X, Zhao N, Liu G, Kan Y, Dong Y, Cui E-N, Luo Y, Yu T, Jiang X, Radiomic evaluations of the diagnostic performance of dm, dbt, dce mri, dwi, and their combination for the diagnosisof breast cancer, Frontiers in Oncology 11 (2021) 725922. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [350].Sahu T, Subasi A, Chapter 8 - Deep learning approaches for breast cancer detection using breast MRI, in: Subasi A (Ed.), Applications of Artificial Intelligence in Healthcare and Biomedicine, Artificial Intelligence Applications in Healthcare and Medicine, Academic Press, 2024, pp. 205–242. doi: 10.1016/B978-0-443-22308-2.00012-3. URL https://www.sciencedirect.com/science/article/pii/B9780443223082000123 [DOI] [Google Scholar]
- [351].Alonazi B, Magnetic Resonance Imaging in Breast Cancer Screening and Diagnosis, International Journal of Biomedicine 12 (1) (2022) 89–94. doi: 10.21103/article12(1)_ra4. URL https://doi.org/10.21103/article12(1)_ra4 [DOI] [Google Scholar]
- [352].Zang H, Liu H.-l., Zhu L.-y., Wang X, Wei L.-m., Lou J.-j., Zou Q.-g., Wang S.-q., Wang S.-j., Jiang Y.-n., Diagnostic performance of dcemri, multiparametric mri and multimodality imaging for discrimination of breast non-mass-like enhancement lesions, The British Journal of Radiology 95 (1136) (2022) 20220211. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [353].Afrifa S, Varadarajan V, Appiahene P, Zhang T, A novel artificial intelligence techniques for women breast cancer classification using ultrasound images, Clinical and Experimental Obstetrics & Gynecology 50 (12) (2023) 271. [Google Scholar]
- [354].Khaleefah SH, Lojungin EC, Mostafa SA, Baharum Z, Aldulaimi MH, Ghazal TM, Alo SO, Hidayat R, Applying deep learning models to breast ultrasound images for automating breast cancer diagnosis, JOIV: International Journal on Informatics Visualization 8 (3–2) (2024) 1779–1783. [Google Scholar]
- [355].Rai HM, Dashkevych S, Yoo J, Next-generation diagnostics: the impact of synthetic data generation on the detection of breast cancer from ultrasound imaging, Mathematics 12 (18) (2024) 2808. [Google Scholar]
- [356].Morsy SE, Abd-Elsalam NM, Abduh Z, Kandil AH, El-Bialy A, Youssef AM, A model for classifying breast masses in ultrasound images, International Journal of Advances in Applied Sciences 13 (2024) 566. doi: 10.11591/ijaas.v13.i3.pp566-578. [DOI] [Google Scholar]
- [357].Zhang S, Liao M, Wang J, Zhu Y, Zhang Y, Zhang J, Zheng R, Lv L, Zhu D, Chen H, et al. , Fully automatic tumor segmentation of breast ultrasound images with deep learning, Journal of Applied Clinical Medical Physics 24 (1) (2023) e13863. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [358].Liao W-X, He P, Hao J, Wang X-Y, Yang R-L, An D, Cui L-G, Automatic identification of breast ultrasound image based on supervised block-based region segmentation algorithm and features combination migration deep learning model, IEEE journal of biomedical and health informatics 24 (4) (2019) 984–993. [DOI] [PubMed] [Google Scholar]
- [359].Liu H, Cui G, Luo Y, Guo Y, Zhao L, Wang Y, Subasi A, Dogan S, Tuncer T, Artificial intelligence-based breast cancer diagnosis using ultrasound images and grid-based deep feature generator, International Journal of General Medicine (2022) 2271–2282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [360].Sokouti M, Ramin RS, Pashazadeh S, Abadi SEH, Sokouti M, Ghojazadeh M, Sokouti B, Investigation of diagnostic value of artificial intelligence systems in the diagnosis of breast cancer based on histopathological images using meta-mums dta tool, Epidemiology, Biostatistics, and Public Health 17 (2) (2020). [Google Scholar]
- [361].Ramamoorthy P, Reddy BRR, Askar S, Abouhawwash M, Histopathology-based breast cancer prediction using deep learning methods for healthcare applications, Frontiers in Oncology 14 (2024) 1300997. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [362].El Agouri H, Azizi M, El Attar H, El Khannoussi M, Ibrahimi A, Kabbaj R, Kadiri H, BekarSabein S, EchCharif S, Mounjid C, et al. , Assessment of deep learning algorithms to predict histopathological diagnosis of breast cancer: first moroccan prospective study on a private dataset, BMC research notes 15 (1) (2022) 66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [363].Islam R, Tarique M, Artificial intelligence (ai) and nuclear features from the fine needle aspirated (fna) tissue samples to recognize breast cancer, Journal of Imaging 10 (8) (2024) 201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [364].Co M, Lau YCC, Qian YXY, Chan MCR, Wong D. K.-k., Lui KH, So NYH, Tso SWS, Lo YC, Lee WJ, et al. , Artificial intelligence in histologic diagnosis of ductal carcinoma in situ, Mayo Clinic Proceedings: Digital Health 1 (3) (2023) 267–275. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [365].Abdollahi J, Keshandehghan A, Gardaneh M, Panahi Y, Gardaneh M, Accurate detection of breast cancer metastasis using a hybrid model of artificial intelligence algorithm, Archives of Breast Cancer (2020) 22–28. [Google Scholar]
- [366].Kowsalya M, Sumathi P, Breast thermograms analysis using deep neural network, International Journal For Multidisciplinary Research (2023). doi: 10.36948/ijfmr.2023.v05i04.4772. [DOI] [Google Scholar]
- [367].Aidossov N, Mashekova A, Zhao Y, Zarikas V, Eddie-Yin-Kwee Ng MO, Mukhmetov O, Intelligent diagnosis of breast cancer with thermograms using convolutional neural networks., in: ICAART (2), 2022, pp. 598–604. [Google Scholar]
- [368].Dihmani H, Bousselham A, Bouattane O, A new computer-aided diagnosis system for breast cancer detection from thermograms using metaheuristic algorithms and explainable ai, Algorithms 17 (10) (2024) 462. [Google Scholar]
- [369].Chebbah NK, Ouslim M, Benabid S, New computer aided diagnostic system using deep neural network and svm to detect breast cancer in thermography, Quantitative InfraRed Thermography Journal 20 (2) (2023) 62–77. [Google Scholar]
- [370].Botlagunta M, Botlagunta MD, Myneni MB, Lakshmi D, Nayyar A, Gullapalli JS, Shah MA, Classification and diagnostic prediction of breast cancer metastasis on clinical data using machine learning algorithms, Scientific Reports 13 (1) (2023) 485. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [371].Gonzalez-Castro L, Chávez M, Duflot P, Bleret V, Martin AG, Zobel M, Nateqi J, Lin S, Pazos-Arias JJ, Del Fiol G, et al. , Machine learning algorithms to predict breast cancer recurrence using structured and unstructured sources from electronic health records, Cancers 15 (10) (2023) 2741. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [372].Jaiswal V, Saurabh P, Lilhore UK, Pathak M, Simaiya S, Dalal S, A breast cancer risk predication and classification model with ensemble learning and big data fusion, Decision Analytics Journal 8 (2023) 100298. [Google Scholar]
- [373].Sharma A, Kumar D, Classification with 2-d convolutional neural networks for breast cancer diagnosis, Scientific Reports 12 (1) (2022) 21857. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [374].Mihaylov I, Kańduła M, Krachunov M, Vassilev D, A novel framework for horizontal and vertical data integration in cancer studies with application to survival time prediction models, Biology direct 14 (1) (2019) 22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [375].Wani NA, Kumar R, Bedi J, Harnessing fusion modeling for enhanced breast cancer classification through interpretable artificial intelligence and in-depth explanations, Engineering Applications of Artificial Intelligence 136 (2024) 108939. [Google Scholar]
- [376].Hassan AM, Biaggi-Ondina A, Rajesh A, Asaad M, Nelson JA, Coert JH, Mehrara BJ, Butler CE, Predicting patient-reported outcomes following surgery using machine learning, The American Surgeon 89 (1) (2023) 31–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [377].Kang D, Kim H, Cho J, Kim Z, Chung M, Lee JE, Nam SJ, Kim SW, Yu J, Chae BJ, et al. , Prediction model for postoperative quality of life among breast cancer survivors along the survivorship trajectory from pretreatment to 5 years: Machine learning–based analysis, JMIR public health and surveillance 9 (1) (2023) e45212. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [378].Pfob A, Mehrara BJ, Nelson JA, Wilkins EG, Pusic AL, Sidey-Gibbons C, Towards patient-centered decision-making in breast cancer surgery: machine learning to predict individual patient-reported outcomes at 1-year follow-up, Annals of surgery 277 (1) (2023) e144–e152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [379].Pfob A, Mehrara B, Nelson J, Wilkins EG, Pusic A, Sidey-Gibbons C, Towards data-driven decision-making for breast cancer patients undergoing mastectomy and reconstruction: Prediction of individual patient-reported outcomes at two-year follow-up using machine learning. (2020).
- [380].Iivanainen S, Ekstrom J, Virtanen H, Kataja VV, Koivunen JP, Electronic patient-reported outcomes and machine learning in predicting immune-related adverse events of immune checkpoint inhibitor therapies, BMC Medical Informatics and Decision Making 21 (1) (2021) 205. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [381].Hussain S, Ali M, Naseem U, Avalos DBA, Cardona-Huerta S, Tamez-Pena JG, Multiview multimodal feature fusion for breast cancer classification using deep learning, IEEE Access; (2024). [Google Scholar]
- [382].Arya N, Saha S, Mathur A, Saha S, Improving the robustness and stability of a machine learning model for breast cancer prognosis through the use of multi-modal classifiers, Scientific Reports 13 (1) (2023) 4079. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [383].Maigari A, XinYing C, Zainol Z, Multimodal deep learning breast cancer prognosis models: narrative review on multimodal architectures and concatenation approaches, Journal of Medical Artificial Intelligence 8 (2025). [Google Scholar]
- [384].Yao Y, Lv Y, Tong L, Liang Y, Xi S, Ji B, Zhang G, Li L, Tian G, Tang M, et al. , Icsda: a multi-modal deep learning model to predict breast cancer recurrence and metastasis risk by integrating pathological, clinical and gene expression data, Briefings in bioinformatics 23 (6) (2022) bbac448. [DOI] [PubMed] [Google Scholar]
- [385].Kanwal S, Khan F, Alamri S, A multimodal deep learning infused with artificial algae algorithm–an architecture of advanced e-health system for cancer prognosis prediction, Journal of King Saud University-Computer and Information Sciences 34 (6) (2022) 2707–2719. [Google Scholar]
- [386].Doi K, Computer-aided diagnosis in medical imaging: Historical review, current status and future potential, Computerized Medical Imaging and Graphics 31 (4–5) (2007) 198–211. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.