Artificial intelligence-driven transformative applications in disease diagnosis technology

Junyu Zhou; Sunmin Park; Sihan Dong; Xiaoying Tang; Xunbin Wei

doi:10.1515/mr-2024-0097

. 2025 Apr 11;5(5):353–377. doi: 10.1515/mr-2024-0097

Artificial intelligence-driven transformative applications in disease diagnosis technology

Junyu Zhou ^1,², Sunmin Park ³, Sihan Dong ¹, Xiaoying Tang ⁴, Xunbin Wei ^1,^5,^6,^7,^✉

PMCID: PMC12558038 PMID: 41158286

Abstract

The integration of artificial intelligence (AI) in medical diagnostics represents a transformative advancement in healthcare, with projected market growth reaching $188 billion by 2030. This comprehensive review examines the latest developments in AI-driven diagnostic technologies across multiple disease domains, particularly focusing on cancer, Alzheimer’s disease (AD), and diabetes. Through systematic bibliometric analysis using GraphRAG methodology, we analyzed research publications from 2022 to 2024, revealing the distribution and impact of AI applications across various medical fields. In cancer diagnostics, AI systems have achieved breakthrough performances in analyzing medical imaging and molecular data, with notable advances in early detection capabilities across 19 different cancer types. For AD diagnosis, AI-powered tools have demonstrated up to 90 % accuracy in risk detection through non-invasive methods, including speech pattern analysis and blood-based biomarkers. In diabetes care, AI-integrated systems incorporating deep neural networks and electronic nose technology have shown remarkable accuracy in predicting disease onset before clinical manifestation. These developments collectively indicate a paradigm shift toward more precise, efficient, and accessible diagnostic approaches. However, challenges remain in standardization, data quality, and clinical implementation. This review synthesizes current progress while highlighting the potential for AI to revolutionize medical diagnostics through enhanced accuracy, early detection, and personalized patient care.

Keywords: artificial intelligence, disease diagnosis, deep learning, medical imaging, electronic health records

Introduction

The dawn of artificial intelligence (AI) in healthcare marks one of the most profound technological revolutions in medical history, poised to fundamentally reshape how we diagnose, treat, and prevent diseases [1]. This transformation is not merely theoretical, market analyses project the healthcare AI industry to reach an unprecedented $188 billion by 2030, reflecting both its current impact and enormous future potential [2]. At the heart of this revolution lies AI’s extraordinary ability to process and analyze vast amounts of medical data with unprecedented speed and accuracy, uncovering subtle patterns that often elude human detection [3]. The convergence of sophisticated deep learning (DL) algorithms and advanced GPU processing power has catalyzed a new era in medical diagnostics [4], where AI systems consistently demonstrate capabilities that complement and sometimes surpass human expertise [5]. This technological leap is particularly evident in clinical diagnostics, where AI-powered systems are achieving breakthrough levels of accuracy across multiple modalities, from medical imaging to genomic analysis [6]. For instance, applications like photoacoustic effect technology, enhanced by DL frameworks, now enable non-invasive diagnosis of conditions such as circulating melanoma, showcasing the transformative potential of AI in diagnostic medicine [7]. Among the vast landscape of medical conditions, our review specifically focuses on three critical areas, cancer, Alzheimer’s disease (AD), and diabetes. These diseases represent some of the most significant global health challenges of our time and these diseases exemplify the diverse applications and capabilities of AI in medical diagnostics. Cancer diagnostics showcase AI’s prowess in image analysis and molecular data interpretation, from mammography to histopathology. AD diagnosis demonstrates AI’s ability to process complex multimodal data, including brain imaging, cognitive assessments, and genetic markers. Diabetes care illustrates AI’s capacity for continuous monitoring and predictive analytics, integrating various physiological parameters and lifestyle data. The impact of AI extends far beyond individual diagnostic tools. In clinical decision support, AI algorithms have demonstrated superior performance compared to traditional assessment methods like the modified early warning score (MEWS) [8], while advanced natural language processing (NLP) capabilities are revolutionizing how we extract and utilize information from EHRs [9]. These developments are complemented by traditional clinical decision support (CDS) systems [10], which have evolved to incorporate sophisticated machine learning (ML) techniques [11]. Integrating AI across multiple healthcare domains, including ultrasound, magnetic resonance imaging, genomics, and computed tomography (CT), represents a comprehensive transformation in medical practice [12]. Particularly in image analysis, combining ML for feature extraction and deep neural networks (DNN) for classification has enabled unprecedented advances in lesion detection, segmentation, and diagnostic accuracy [13]. This technological convergence is creating a new paradigm in healthcare delivery, where AI augments human expertise to enhance diagnostic precision and streamline clinical workflows across the entire healthcare spectrum [6].

While AI’s revolutionary impact on healthcare diagnostics has demonstrated remarkable achievements, it simultaneously presents a complex landscape of both unprecedented opportunities and significant challenges [14]. The transformation is particularly evident in how AI-driven CDS Systems (CDSSs) are reshaping medical practice, offering capabilities that extend far beyond traditional diagnostic tools. These systems are not merely enhancing accuracy, they are fundamentally redefining the speed, scope, and sophistication of medical diagnosis [15]. The journey toward AI integration in healthcare reveals a dual narrative of breakthrough innovations and practical hurdles. On one front, AI systems are achieving remarkable success in drug development and complex disease diagnosis [16], 17], pushing the boundaries of what’s possible in personalized medicine. For instance, advanced ML algorithms are now capable of identifying subtle patterns in patient data that can predict disease progression or treatment responses with unprecedented accuracy. However, these same systems face significant real-world challenges, particularly in navigating the inconsistencies and quality variations inherent in electronic health records (EHRs) [15]. These challenges, rather than serving as roadblocks, are catalyzing innovative solutions that promise to revolutionize healthcare delivery [14]. As AI systems evolve to address these complexities, they are simultaneously driving improvements in data quality, standardization, and interoperability. This evolution is creating a more robust foundation for future innovations, suggesting that current challenges are not merely obstacles but stepping stones toward a more sophisticated and integrated healthcare ecosystem [16], 17]. This dynamic interplay between progress and challenges is shaping a future where medical diagnostics becomes increasingly precise, efficient, and accessible, while maintaining the critical balance between technological innovation and practical implementation.

Bibliometric analysis of AI applications in medical diagnostics

The research methodology for this review on AI applications in medical diagnostics followed a systematic and structured approach to synthesize and analyze existing literature. The methodology was designed to address the complex research questions surrounding the role of AI in medical diagnosis across a wide range of disease domains. A comprehensive, multi-stage literature search strategy was employed to ensure the inclusion of relevant studies. The primary method for literature selection, categorization, and analysis was Graph-based Retrieval Augmented Generation (GraphRAG), an advanced tool for managing and synthesizing research data [18]. The search process involved several key steps. First, a range of prominent academic databases were selected for the literature search, including PubMed (https://pubmed.ncbi.nlm.nih.gov), Web of Science (clarivate.com/products/web-of-science), Scopus (https://www.scopus.com), IEEE Xplore (https://ieeexplore.ieee.org/Xplore/home.jsp), and the ACM Digital Library (https://dl.acm.org/). The search was limited to publications from the years 2022–2024, and only articles in English were included. It enabled semantic clustering of research papers (Supplementary file), which allowed for more nuanced filtering and classification of literature. It helped identify interconnections between AI technologies and different medical diagnostic domains, contributing to a deeper understanding of the trends in AI-driven healthcare advancements.

The inclusion and exclusion criteria for selecting studies were clearly defined. The review focused on peer-reviewed research articles that provided empirical evidence of AI’s diagnostic performance in medical diagnosis. Specifically, the studies included in this review were categorized based on the community classification of AI applications in medical diagnosis. The research covered a broad spectrum of disease fields, including cancer, AD, diabetes, cardiovascular diseases, ophthalmological conditions, respiratory diseases, and other relevant medical diagnosis. Articles without full-text availability, those published in non-English languages, and studies that lacked a clear AI diagnostic methodology were excluded from the review. Additionally, purely theoretical research or studies with insufficient methodological details were not considered. For data extraction and analysis, a structured process was employed. The challenges of AI integration into medical diagnostics were also explored, along with the assessment of diagnostic accuracy, efficiency, and potential limitations of the technologies discussed in the reviewed literature.

We conducted a systematic bibliometric network analysis of AI applications in medical diagnostics (Figure 1). The citation network analysis (Figure 1A) demonstrated complex inter-publication citation patterns. Significant knowledge associations and cross-referencing were observed across different research domains, highlighting the interconnected nature of AI-driven medical diagnostic research. The community-colored cluster citation network (Figure 1B) revealed the multidisciplinary nature of AI medical diagnostic research. Color-coded clusters illustrate the diverse and interconnected research communities within the field. The proportional representation of research applications (Figure 1C) showed cancer diagnostic domains comprised the largest proportion of research applications neurodegenerative disease and cardiovascular research followed closely in research intensity. The bar graph of research application quantities (Figure 1D) further validated the distribution patterns, quantitatively measuring research efforts across different medical diagnostic domains. The Sankey diagram of citation volume distribution (Figure 1E) provided insights into research impact gradients. High-impact research (≥50 citations) was primarily concentrated in cancer and neurological disease domains, medium-impact research (≥20 citations) was broadly distributed across diagnostic areas, low-impact research was most numerous, reflecting the field’s diversity and developmental stage.

AI-powered non-invasive cancer diagnostics and specialized applications

The intersection of AI and cancer diagnostics represents one of the most promising frontiers in modern medicine, where ML is fundamentally transforming our approach to oncology [19]. This transformation spans the entire spectrum of cancer care, from early detection to treatment planning, with particularly remarkable advances in analyzing complex medical imaging and molecular data from both liquid and solid tumor biopsies. The sophistication of these AI systems extends beyond simple pattern recognition, offering unprecedented capabilities in integrating and analyzing diverse data sources, including EHRs, diagnostic images, pathology slides, and blood tests [20]. In the realm of medical imaging, DM algorithms have achieved breakthrough performances, particularly in the analysis of head CT scans [20]. These systems demonstrate exceptional accuracy in identifying subtle abnormalities that might escape human detection, revolutionizing the early detection of various cancers and related conditions [21]. The impact of these advances is particularly evident in how they’re reshaping diagnostic workflows, significantly reducing the time from initial screening to diagnosis while maintaining or exceeding human-level accuracy.

A landmark development in this field emerged from Harvard Medical School, where researchers created a versatile AI model reminiscent of ChatGPT’s flexibility but specifically tailored for cancer diagnostics [22]. This model represents a quantum leap in capability, successfully handling 19 different cancer types, a dramatic improvement over previous systems limited to single cancer types or specific tasks. Its ability to predict patient outcomes and validate results across international patient populations marks a new era in global cancer diagnostics. The evolution of AI in cancer diagnosis has also seen remarkable innovations in specialized applications. AMIE, a sophisticated language model-based system, has demonstrated superior diagnostic accuracy compared to primary care physicians in simulated consultations [23]. This advancement is complemented by groundbreaking work in non-invasive diagnosis methods, such as the development of surface-enhanced Raman spectroscopy utilizing plasma metal organic framework nanoparticle film technology [24], offering new possibilities for early and accurate cancer detection without invasive procedures.

Recent breakthroughs in specific cancer types showcase the diverse applications of AI in oncology. DeepXplainer’s interpretable DM approach has revolutionized lung cancer detection [25], while AI-powered MRI analysis has transformed the landscape of prostate cancer diagnosis [26]. MIT’s Sybil project represents another significant milestone, introducing sophisticated risk assessment capabilities for lung cancer [27]. In the realm of blood cancers, the Dana-Farber Cancer Institute’s CRISPR-based diagnostic tool has achieved remarkable accuracy in detecting gene fusions in acute promyelocytic leukemia and chronic myeloid leukemia [28]. One of the most promising developments comes from a collaborative effort between Harvard and the University of Copenhagen, where researchers have developed an AI tool capable of predicting pancreatic cancer up to three years before conventional diagnosis [29]. This predictive capability, achieved through sophisticated analysis of medical records, represents a paradigm shift in cancer diagnostics, moving from reactive to proactive disease detection. These advancements collectively demonstrate how AI is not just enhancing existing diagnostic methods but fundamentally reimagining the possibilities in cancer detection and treatment. The integration of these various technologies and approaches is creating a more comprehensive and nuanced understanding of cancer diagnosis, promising earlier detection, more accurate prognosis, and ultimately, better patient outcomes.

AI-driven cancer diagnostic tools are transforming the diagnostic workflow by integrating and analyzing diverse data sources such as medical imaging, EHRs, and blood tests. Compared to traditional methods, AI offers superior accuracy and faster diagnosis, particularly in the early detection of cancers where human detection may fail (Figure 2A). The ability of AI to process complex datasets and detect subtle patterns ensures a more comprehensive approach to diagnosis across a range of cancer types. AI technologies are being applied across a wide range of cancer types, from lung and prostate cancer to blood cancers and pancreatic cancer. For example, DeepXplainer uses an interpretable DM method for lung cancer detection, while AI-powered MRI analysis improves the accuracy of prostate cancer diagnoses. Moreover, AI systems like the CRISPR-based diagnostic tool can detect genetic mutations in blood cancers, such as leukemia, marking a significant breakthrough in non-invasive diagnostic capabilities (Figure 2B). These AI applications offer a higher degree of precision and early detection, revolutionizing the landscape of cancer diagnostics.

The pairwise comparisons reveal notable differences in the performance of AI models across various cancer diagnostic tasks (Figure 2C). DeepXplainer excels in lung cancer detection by identifying subtle patterns in CT scans, while Sybil offers superior lung cancer risk prediction through a more comprehensive risk assessment model. AI-powered MRI tools show a clear advantage in the diagnosis of prostate cancer due to their ability to analyze detailed imaging data, whereas CRISPR-based AI tools are highly effective in molecular diagnostics for blood cancers like leukemia, detecting gene mutations with high accuracy. The comparison between AMIE system and primary care physicians demonstrates that AI models can provide higher diagnostic accuracy, particularly in complex consultations or in cases involving rare cancer types. Additionally, the Harvard AI tool’s ability to handle multiple cancer types with international validation underscores the growing versatility and global applicability of AI models in clinical practice (Figure 2C).

A paradigm shift in Alzheimer’s patient-friendly diagnostics

Beyond cancer, AI technologies are making significant strides in diagnosing various diseases, particularly in three major areas: AD, diabetes, and cardiovascular diseases. The application of AI in AD diagnosis represents a paradigm shift in neurodegenerative disease detection, moving from traditional cognitive assessments to sophisticated, multi-modal diagnostic approaches that promise earlier and more accurate detection [30]. This transformation is particularly significant given the complex nature of AD and the critical importance of early intervention in managing its progression.

A landmark development in this field came with the AD Association’s 2024 updated diagnostic criteria, which notably incorporates blood-based biomarkers (BBMs) [30]. This inclusion marks a revolutionary step forward, offering a less invasive and more accessible method for detecting disease-related brain changes before clinical symptoms become apparent. The integration of BBMs with AI analysis capabilities has opened new frontiers in early diagnosis, potentially allowing for intervention at stages where therapeutic interventions might be most effective. The evolution of AI-driven diagnostic tools has been particularly remarkable in the analysis of speech patterns and cognitive function. Columbia Nursing’s groundbreaking ADscreen represents a significant advancement in this area [31], utilizing sophisticated speech processing algorithms that enable clinicians to detect subtle changes in cognitive function through natural conversation. This innovation is complemented by the University of Sheffield’s CognoSpeak system [32], which has achieved a remarkable breakthrough by matching the accuracy of traditional cognitive assessments through automated speech pattern analysis. These tools are transforming the diagnostic landscape by offering non-invasive, easily administered tests that can be conducted in various clinical settings. Perhaps one of the most significant breakthroughs comes from Massachusetts General Hospital, where researchers have achieved an impressive 90 % accuracy rate in detecting AD risk [33]. Their DM approach, applied to routine brain images, demonstrates how AI can extract meaningful diagnostic information from standard medical data, potentially eliminating the need for specialized and expensive imaging procedures. This advancement is particularly noteworthy as it makes high-accuracy screening more accessible and cost-effective. Boston University’s contribution to this field has been equally revolutionary, with their AI program’s ability to predict the progression from mild cognitive impairment (MCI) to AD dementia within a six-year window [34]. This predictive capability represents a crucial advancement in disease management, allowing healthcare providers to initiate interventions earlier and potentially slow disease progression. The program’s ability to identify those at highest risk for progression provides invaluable information for both clinical care and research purposes.

These developments collectively represent a significant departure from traditional diagnostic methods that relied heavily on invasive procedures like spinal taps or expensive PET scans. The emergence of these AI-driven tools marks a transformation in how we approach AD diagnosis, offering a more comprehensive, accessible, and patient-friendly approach to disease detection and monitoring. The integration of multiple diagnostic modalities, from blood biomarkers to speech analysis and brain imaging, creates a more robust and nuanced understanding of disease progression, potentially leading to more personalized and effective treatment strategies. The rapid advancement in AI-driven diagnostic tools for AD not only improves the accuracy and accessibility of diagnosis but also opens new possibilities for research and understanding of the disease mechanism itself. These technologies are not just diagnostic tools; they are becoming integral components in the broader ecosystem of AD research and treatment, potentially accelerating the path to more effective therapies.

The comparison of brain slices from different stages of AD reveals significant structural changes, with the most pronounced brain atrophy seen in Severe AD cases. In contrast to the normal brain, Mild AD cases show early signs of degeneration, including slight shrinkage in key areas such as the hippocampus and temporal cortex (Figure 3A). These differences in brain morphology are increasingly detectable through AI-enhanced imaging analysis, which provides greater sensitivity in identifying early-stage AD changes that are not immediately visible through traditional methods. AI diagnostic tools allow for more accurate, automated analysis of brain imaging, offering a non-invasive way to assess disease progression and potentially detect AD earlier than traditional diagnostic methods. The integration of brain imaging analysis with DM algorithms enables accurate detection of AD-related structural changes, even in its early stages. However, while methods like MRI and PET scans offer high accuracy, they require expensive equipment and specialized training, which can limit accessibility (Figure 3B). AI-driven speech pattern analysis (e.g., ADscreen and CognoSpeak) provides a non-invasive and easily administered tool, but may still face challenges in distinguishing subtle cognitive changes in the early stages. Blood-based biomarkers (BBMs) offer an exciting alternative, providing a less invasive and more accessible diagnostic approach, although the technology is still evolving and may not yet capture all the nuances of disease progression (Figure 3B). Additionally, some AI models, such as those predicting the progression from MCI to AD dementia, show great promise in predicting future disease development, which can help clinicians intervene earlier.

Figure 3: — AI-enhanced diagnostic approaches in Alzheimer’s disease. (A) Comparison of normal brain and Alzheimer’s disease (mild and severe stages). The top panel shows a horizontal cross-sectional view of a healthy brain, with labeled regions of interest. The lower panels display brain slices from patients diagnosed with mild AD and severe AD, respectively, highlighting the progressive changes in brain structure. These images emphasize the atrophy and shrinkage of brain tissue in AD, particularly in areas associated with cognitive function such as the hippocampus and cortical regions. The images are analyzed using AI-based tools to detect subtle changes that may not be visible to the naked eye, illustrating the potential of AI in diagnosing AD at earlier stages. (B) AI-driven approaches in Alzheimer’s disease diagnostic assessment. This figure summarizes the various AI-based diagnostic methods used in Alzheimer’s disease (AD) detection, outlining their respective advantages and limitations. The methods include AI-driven analysis of brain imaging (MRI, PET scans), speech pattern analysis, blood-based biomarkers (BBMs), and cognitive function assessments. For each method, the figure highlights key benefits such as early detection, non-invasiveness, and cost-effectiveness, as well as limitations such as accuracy in detecting early-stage disease or the need for specialized equipment.

AI integrates multi-modal data for comprehensive diabetes diagnostics

The integration of AI in diabetes diagnostics marks a revolutionary advancement in how we detect, predict, and manage one of the world’s most prevalent chronic conditions [35]. This transformation is particularly significant given the complex nature of diabetes and its various manifestations, from Type 2 Diabetes Mellitus (T2DM) to gestational diabetes and its numerous complications. The sophistication of AI-based diagnostic methods has evolved dramatically, particularly in their ability to analyze vast amounts of EHRs data [36]. These systems demonstrate remarkable capabilities in not only identifying current diabetic conditions but also predicting future onset and complications. The scope of AI applications in diabetes care has expanded to encompass a comprehensive range of related conditions, including hypoglycemia, diabetic retinopathy, foot ulcers, peripheral neuropathy, and nephropathy. This multi-faceted approach allows for earlier intervention and more personalized treatment strategies.

A groundbreaking innovation in this field is the development of an electronic nose system incorporating metal oxide semiconductor (MOS) sensors integrated with DNN [37]. This technology represents a significant leap forward in real-time diabetes detection, offering a non-invasive alternative to traditional blood glucose monitoring methods. The system’s ability to detect subtle metabolic changes through breath analysis demonstrates the potential for more patient-friendly diagnostic approaches. The reliability and trustworthiness of AI diagnostics have been significantly enhanced through the implementation of sophisticated convolutional and recurrent neural networks (CNN and RNN) [38]. These advanced algorithms not only improve diagnostic accuracy but also provide transparent decision-making processes, building trust among healthcare providers and patients alike. This aspect is crucial for the widespread adoption of AI-based diagnostic tools in clinical settings. Another significant advancement comes in the form of microstrip isoelectric focusing combined with DM algorithms [39]. This innovative approach offers highly precise molecular-level analysis, enabling earlier detection of diabetes-related biomarkers and potentially identifying subtypes of the disease that might respond differently to various treatments.

In examining the stability and generalization capabilities of AI systems in diabetes diagnostics, recent studies have demonstrated significant progress while highlighting important considerations for real-world implementation. Huang et al. [36] developed a robust model for predicting diabetic complications, achieving impressive accuracy (89 %) and AUC (0.91) (Figure 4A). However, their findings emphasize the critical need for extensive validation across diverse healthcare settings to ensure consistent performance in clinical applications. Significant advancements in model stability have been demonstrated by Gudiño-Ochoa et al. [37], who implemented a TinyML approach for non-invasive breath analysis. Their system achieved 85 % accuracy while maintaining performance stability across varying environmental conditions, including different humidity levels and temperatures. This robustness to environmental factors represents a crucial step toward reliable deployment in real-world clinical settings. Further evidence of generalization capability comes from Eben and Bolanle [38], whose deep learning approach demonstrated remarkable consistency with 92 % accuracy across multiple hospital systems. Their cross-institutional validation methodology provides strong evidence for the model’s ability to handle data distribution shifts while maintaining high performance standards. This work particularly emphasizes the importance of diverse training data and rigorous validation protocols in ensuring model generalization. Fu et al. [39] contributed valuable insights into multi-condition screening capabilities, demonstrating stable performance across diverse demographic groups. Their model’s ability to maintain consistent accuracy levels while screening for multiple conditions simultaneously suggests robust feature extraction and classification capabilities (Figure 4A). This work particularly highlights the importance of comprehensive validation across different patient populations to ensure reliable clinical application.

Figure 4: — AI innovations in diabetes diagnosis and their impact on early detection and management. (A) Performance analysis of AI diagnostic models in diabetes care across clinical and environmental variables. This figure compares the stability and generalization capabilities of recent AI models for diabetes diagnostics under varying real-world conditions. (B) Evaluation of AI-enabled diagnostic approaches in diabetes mellitus. Overview of various AI-based diagnostic methods used in diabetes detection and management, highlighting how these AI-driven technologies are transforming diabetes diagnostics by offering more personalized, accurate, and efficient diagnostic tools compared to traditional methods. (C) Key advantages of AI-based diabetes diagnosis. The figure highlights the major advantages of AI over traditional methods, such as non-invasive nature, early and more accurate detection, the ability to predict diabetes complications, and real-time monitoring. It also addresses the advantages of AI systems in predicting disease onset before symptoms appear, offering the potential for proactive intervention.

The clinical validation of these AI systems has yielded remarkable results, with some models achieving unprecedented accuracy levels of up to 90 % in predicting diabetes onset before clinical symptoms become apparent [40]. This predictive capability represents a paradigm shift in diabetes care, moving from reactive treatment to proactive prevention. The ability to identify high-risk individuals before the onset of clinical symptoms opens new possibilities for early intervention and lifestyle modifications that could potentially prevent or delay the development of diabetes. These technological advances collectively represent a transformation in diabetes diagnostics, offering multiple advantages over traditional methods: (1) earlier detection and intervention opportunities; (2) more accurate prediction of disease onset and complications; (3) non-invasive monitoring options; (4) comprehensive analysis of multiple risk factors and biomarkers; (5) personalized risk assessment and treatment planning; (6) real-time monitoring capabilities; (7) cost-effective screening methods. The integration of these various AI technologies is creating a more comprehensive approach to diabetes diagnosis and management, where multiple data points and diagnostic methods work in concert to provide a more complete picture of patient health. This holistic approach not only improves diagnostic accuracy but also enables more personalized treatment strategies, potentially leading to better patient outcomes and quality of life. The continued evolution of AI in diabetes diagnostics suggests we are only beginning to realize the full potential of these technologies. As ML algorithms become more sophisticated and our understanding of diabetes pathophysiology deepens, we can expect even more innovative and effective diagnostic tools to emerge, further transforming our approach to diabetes care and management.

AI-based diagnostic methods for diabetes offer significant advantages, including earlier detection of diabetes onset, non-invasive monitoring options (e.g., electronic nose system), and high predictive accuracy through ML models like CNNs and RNNs (Figure 4B). These methods have achieved exceptional accuracy rates, with some models capable of predicting diabetes onset before clinical symptoms appear. The AI systems provide a more personalized, data-driven approach, offering precise risk assessments based on EHR data, biomarkers, and patient history. However, challenges remain, such as the need for extensive clinical validation, the requirement for specialized technology (e.g., MOS sensors or microstrip isoelectric focusing), and the complexity of integrating these systems into routine clinical practice (Figure 4B). Despite these limitations, the potential of AI to revolutionize diabetes diagnostics and management is clear, providing more cost-effective and comprehensive screening options. AI-based diabetes diagnostic methods represent a major advancement over traditional approaches. Unlike traditional methods such as blood glucose testing and OGTT, AI-driven tools can predict diabetes onset and complications earlier, offering a proactive approach to disease management. AI technologies such as breath analysis with electronic noses and real-time retinal screening are non-invasive, patient-friendly alternatives that reduce the need for costly and invasive procedures like blood draws or glucose tolerance tests. Furthermore, AI models that analyze EHRs and other patient data enable more personalized, precise risk assessments, making diabetes care more tailored to individual needs (Figure 4C). The ability of AI to offer continuous monitoring and to predict complications such as diabetic retinopathy or hypoglycemia represents a significant leap forward, enhancing long-term management and potentially improving patient outcomes.

Comprehensive approaches in imaging precision and early diagnosis to heart disease

The landscape of cardiovascular medicine is undergoing a transformative revolution, driven by the unprecedented capabilities of AI [41]. This paradigm shift could not be more timely, as cardiovascular diseases continue to be the leading cause of mortality globally, demanding innovative approaches to early detection and precise intervention. At the forefront of this medical revolution is the groundbreaking AI-based super stethoscope system [42], a technological marvel that fundamentally challenges centuries-old diagnostic limitations. Unlike traditional auscultation methods, this cutting-edge technology harnesses advanced acoustic analysis and sophisticated ML algorithms to unveil cardiac abnormalities that would remain hidden from the human ear. It represents a quantum leap in primary cardiac screening, promising to detect the most subtle heart sound variations with unprecedented sensitivity.

The diagnostic prowess of AI extends far beyond traditional boundaries. Multiple rigorous studies have demonstrated the remarkable superiority of AI-driven tools in detecting heart valve diseases [43], [44], [45], consistently outperforming conventional diagnostic approaches. By deciphering complex acoustic patterns and correlating them with specific pathological signatures, these intelligent systems are redefining the precision and accessibility of cardiac diagnostics. In the realm of electrocardiogram (ECG) analysis, AI’s capabilities are nothing short of revolutionary [41]. These intelligent systems excel at comprehensively examining cardiac conditions, demonstrating exceptional skill in identifying reduced ejection fraction, exploring various forms of cardiomyopathy, detecting complex arrhythmias, and pinpointing structural heart abnormalities. Remarkably, these AI technologies not only match but frequently surpass human diagnostic expertise in both accuracy and diagnostic speed. Most exciting is the innovative research utilizing the UK Biobank data, which demonstrates AI’s potential to transform cardiovascular risk assessment through retinal imaging [46]. By analyzing microvasculature in eye images – particularly for individuals with prediabetes and T2DM, this non-invasive approach offers a more comprehensive risk stratification method that could potentially revolutionize cardiac screening.

The integration of AI in cardiac CT further amplifies these diagnostic capabilities [47]. These advanced systems have dramatically enhanced our ability to diagnose critical cardiac conditions, including heart failure, atrial fibrillation, valvular heart disease, and coronary artery disease. By providing unprecedented detail and accuracy in image analysis, AI is not just improving diagnostic precision but fundamentally restructuring cardiac imaging workflows. This advancement has particular significance in earlier detection of cardiac abnormalities, more accurate risk stratification, better treatment planning, and enhanced monitoring of disease progression. The broader implications are profound. We are witnessing the emergence of a new medical paradigm marked by earlier disease detection, more personalized risk assessment, better-informed treatment decisions, potentially improved patient outcomes, and more efficient healthcare resource utilization.

The collective impact of these AI-driven innovations extends beyond mere diagnosis. They are creating a new paradigm in cardiovascular medicine where detection occurs earlier in the disease process, risk assessment becomes more personalized, treatment decisions are better informed, patient outcomes are potentially improved, and healthcare resources are used more efficiently.

The integration of multiple AI technologies in cardiovascular diagnosis represents a comprehensive approach to heart disease detection and management. By combining various diagnostic modalities, from advanced stethoscopes to retinal imaging and cardiac CT, AI is enabling a more nuanced and complete understanding of cardiovascular health. This multi-modal approach not only improves diagnostic accuracy but also provides clinicians with richer, more detailed information for treatment planning. These developments suggest we are entering a new era in cardiovascular medicine, where AI-enhanced diagnostics will play an increasingly central role in patient care.

Figure 5 demonstrates how core CNN technology is applied across various AI-based diagnostic methods for cardiovascular diseases, significantly enhancing diagnostic capabilities. The super stethoscope system represents a groundbreaking approach to primary cardiac screening by detecting subtle heart sounds with remarkable sensitivity, improving early detection of heart abnormalities that would otherwise go unnoticed. ECG analysis powered by AI excels in detecting a range of conditions, including arrhythmias and structural heart abnormalities, and has shown the ability to outperform human diagnostic expertise in both speed and accuracy. AI’s integration with retinal imaging offers a novel, non-invasive method for assessing cardiovascular risk, particularly in individuals with diabetes, marking a transformative step in preventive cardiovascular care.

AI-enhanced cardiac CT imaging also plays a crucial role in diagnosing critical heart conditions, providing detailed, accurate insights into coronary artery disease, heart failure, and atrial fibrillation. This multi-modal approach, combining acoustic analysis, ECG, retinal imaging, and cardiac CT, enables a comprehensive, more nuanced understanding of cardiovascular health. AI systems offer substantial advantages over traditional methods, including faster diagnosis, earlier detection, and more personalized risk assessments. By streamlining diagnostic workflows and improving accuracy, AI is poised to redefine cardiovascular medicine, leading to better-informed treatment decisions, potentially improved patient outcomes, and more efficient healthcare resource utilization.

AI’s transformative role in making ophthalmology more accessible and efficient

The integration of AI in ophthalmology heralds a transformative era in eye care, where advanced algorithms are fundamentally revolutionizing the detection, diagnosis, and monitoring of vision-threatening conditions [48]. This technological evolution is particularly momentous, given the intricate complexity of eye diseases and the paramount importance of early detection in preserving human vision. Professor Adam Dubis’s pioneering research stands as a watershed moment in ophthalmic diagnostics. His innovative project harnesses the extraordinary power of AI to analyze comprehensive repositories of clinical data and optical coherence tomography (OCT) scans from patients with macular disease. This groundbreaking approach transcends traditional diagnostic methods by identifying subtle disease patterns before they become clinically apparent, tracking disease progression with unprecedented precision, predicting treatment responses through complex data analysis, and enabling deeply personalized treatment strategies through advanced pattern recognition. The project’s significance extends far beyond mere diagnostic capabilities, promising to fundamentally transform early intervention strategies for macular diseases.

A truly groundbreaking advancement emerged from the collaborative efforts of Moorfields Eye Hospital and the UCL Institute of Ophthalmology with their development of RETFound [49]. This sophisticated AI system represents a quantum leap in ophthalmic diagnostics, demonstrating exceptional capabilities in rapidly identifying sight-threatening conditions, analyzing complex retinal imaging data, seamlessly integrating with existing clinical workflows, and enabling accessible screening across diverse healthcare settings. The system’s remarkable ability to process and analyze intricate eye imaging data has established new benchmarks in diagnostic accuracy and operational efficiency.

The validation of these AI systems through competitive diagnostic challenges, particularly in diagnosing conditions like myopic maculopathy [50], has conclusively demonstrated their robust clinical applicability [50]. These rigorous evaluations have revealed that AI systems can match or exceed expert-level diagnosis, process extensive case volumes with remarkable efficiency, maintain consistent diagnostic accuracy, and adeptly adapt to various imaging modalities and clinical environments. The impact of these technological developments extends far beyond individual conditions, suggesting transformative potential across hundreds of sight-threatening eye diseases. The technology shows particular promise in comprehensive screening and diagnostic capabilities, including diabetic retinopathy detection, glaucoma monitoring, age-related macular degeneration assessment, retinal vascular disease diagnosis, and pediatric eye disorder identification.

The integration of AI in ophthalmology is revolutionizing global eye care through multiple critical dimensions. Enhanced accessibility is achieved through remote screening capabilities, reduced dependency on specialist availability, and cost-effective diagnostic solutions. Improved accuracy manifests in consistent and objective assessments, early detection of subtle physiological changes, and significant reduction of diagnostic errors. Efficient workflow optimization enables rapid processing of imaging data, automated screening protocols, and intelligent prioritization of urgent cases. Moreover, advanced predictive capabilities now allow for sophisticated disease progression forecasting, treatment response prediction, and optimized risk assessment. These advancements are particularly significant in addressing global eye health challenges, offering solutions that are inherently scalable across different healthcare systems, adaptable to various resource settings, supportive of preventive eye care initiatives, and effective in alleviating the burden on healthcare providers.

The evolution of AI in ophthalmology represents more than a mere technological advancement; it signifies a fundamental paradigm shift in our approach to eye care. By ingeniously combining sophisticated image analysis with advanced ML capabilities, these systems are creating unprecedented possibilities for earlier disease detection, more accurate diagnosis, improved treatment planning, enhanced patient monitoring, and more intelligent healthcare resource allocation. As these technologies continue to evolve, we can anticipate even more sophisticated applications that will further transform ophthalmic care. The integration of AI in ophthalmology is not simply enhancing current diagnostic capabilities but is actively paving the way for a future where vision-threatening conditions can be identified and treated earlier, potentially preventing vision loss for millions of people worldwide.

AI systems are capable of diagnosing a variety of sight-threatening eye diseases with unprecedented accuracy and efficiency. These include conditions such as diabetic retinopathy, glaucoma, and age-related macular degeneration (AMD), all of which are major causes of vision loss globally. AI tools like the RETFound system have shown exceptional performance in analyzing complex retinal imaging data, identifying subtle pathological changes before they are clinically detectable. The ability to process and analyze large volumes of eye imaging data efficiently allows for early intervention, reducing the risk of severe vision impairment (Figure 6A). AI’s role in diagnosing pediatric eye disorders also opens new possibilities for early screening and preventive care in children, which is critical for long-term vision preservation. AI technologies in ophthalmology have brought significant improvements to diagnostic accuracy and clinical outcomes. OCT analysis and retinal imaging systems like RETFound offer precise, non-invasive assessments of retinal conditions, detecting subtle changes in the eye before symptoms appear. The ML algorithms used in these systems enhance diagnostic capabilities by identifying patterns that are not easily detectable by human clinicians. These systems not only improve the accuracy of diagnoses but also provide personalized treatment plans based on individual patient data, allowing for more targeted and effective interventions (Figure 6B). The clinical impact is profound, leading to earlier disease detection, more effective monitoring of disease progression, and optimized treatment strategies that can help prevent vision loss. AI technologies in ophthalmology provide significant benefits in accessibility, particularly for remote and underserved populations. AI-driven systems make it possible to screen for eye diseases without the need for highly specialized equipment or specialists, opening up new avenues for remote screening and improving global healthcare equity. By automating complex image analyses and decision-making processes, AI systems drastically improve diagnostic accuracy, offering consistent, objective, and reliable results across different healthcare settings. The integration of AI also enhances workflow efficiency by reducing the time required to process and analyze imaging data, enabling faster diagnoses and prioritization of critical cases (Figure 6C). These advantages not only streamline clinical operations but also enable early detection and intervention, improving patient outcomes and optimizing resource allocation across diverse healthcare systems.

Figure 6: — The role of AI in ophthalmology early detection, personalized diagnostics. (A) AI diagnostics for specific ophthalmic diseases. This figure illustrates the specific eye diseases that can be diagnosed using AI technologies in ophthalmology, highlighting how AI models, using imaging data such as optical coherence tomography (OCT) and retinal scans, can accurately detect early-stage disease progression, identify key biomarkers, and offer personalized diagnostic insights that were previously difficult to achieve with traditional methods. (B) Functional advantages, personalized diagnostics, and clinical impact of AI-based ophthalmology diagnosis. This figure outlines the core AI technologies used in the diagnosis of eye diseases and highlights their key functionalities, advantages, and the clinical impact. (C) Key advantages of integrating AI into ophthalmology. Enhancing accessibility, diagnostic accuracy, workflow efficiency, and enabling remote screening.

Advanced sound analysis and machine learning enhancing respiratory disease diagnosis

The integration of AI into respiratory medicine has ushered in a transformative shift in how we diagnose and monitor pulmonary diseases. AI has leveraged the power of bioacoustic analysis, particularly the detection of abnormal sounds such as coughs and breath patterns, which can now be captured with the ubiquitous smartphone microphones. This innovation has proven especially beneficial for the screening and management of diseases like tuberculosis (TB) and chronic obstructive pulmonary disease (COPD), where timely and accurate diagnosis is crucial for effective treatment. One of the most promising advancements is the ability of AI to discern unique spectral signatures in cough sounds. Researchers have identified these patterns as potentially indicative of specific respiratory conditions, allowing for a non-invasive, low-cost diagnostic method. For instance, AI algorithms can classify coughs based on subtle variations in pitch, rhythm, and intensity that correspond to disease markers [51], 52]. In fact, studies have demonstrated that AI-powered cough sound analysis could potentially serve as a powerful screening tool for TB and COPD, detecting these conditions early, often before traditional symptoms manifest.

Further innovations have been made with the development of user-friendly, AI-based cough audio classifiers. These classifiers allow for easy collection and analysis of cough sounds through consumer-grade devices such as smartphones, which make the technology both affordable and widely accessible. Such systems not only offer a potential diagnostic advantage in resource-limited settings but also facilitate continuous monitoring of patients, enabling remote healthcare and telemedicine [53]. Recent advances have focused on refining these classifiers, making them capable of diagnosing respiratory diseases with remarkable accuracy and reliability, independent of clinician bias [54]. Alongside this, comprehensive sound analysis systems have been developed to process and interpret more complex acoustic patterns, offering a deeper understanding of the relationship between bioacoustic features and various lung conditions. These systems leverage ML algorithms to analyze large datasets of cough and breath sounds, improving their diagnostic capabilities with each iteration. Some systems can now differentiate between the subtle audio signatures of conditions like pneumonia, asthma, and bronchitis, even in noisy environments, making them robust tools for widespread clinical application [55]. The field has continued to evolve with the advent of multimodal DM techniques that combine chest medical imaging (such as X-rays and CT scans) with bioacoustic data. By integrating cough sound features with visual diagnostic tools, AI models can offer a more holistic view of a patient’s respiratory health. This approach not only enhances diagnostic precision but also enables more personalized treatment plans. For instance, AI algorithms can use chest X-rays to detect structural changes in the lungs, while simultaneously analyzing cough sounds for inflammation or infection markers, offering a comprehensive diagnostic toolkit that is greater than the sum of its parts [56].

One particularly exciting development in this area is the creation of hybrid AI frameworks that combine quantum feature extraction techniques with advanced classification algorithms for chest X-ray analysis. These hybrid models leverage the power of quantum computing to process and analyze large volumes of medical data with unprecedented speed and accuracy. By applying these sophisticated methodologies, AI has the potential to identify previously undetectable patterns in chest X-rays, improving early detection rates for a variety of respiratory conditions, including lung cancer, fibrosis, and pneumonia [57]. This fusion of cutting-edge technologies positions AI as a critical tool in the early diagnosis and management of respiratory diseases, not only enhancing the speed of diagnosis but also ensuring a higher degree of precision in clinical settings. As AI continues to reshape the landscape of respiratory medicine, these innovative technologies are laying the foundation for a new era of disease detection that is more efficient, accessible, and accurate than ever before. The combination of bioacoustic sound analysis with multimodal imaging and quantum-powered classification systems promises to revolutionize how we approach respiratory diagnostics, offering significant benefits for both patients and healthcare providers alike.

The integration of AI technologies, particularly those involving bioacoustic analysis and multimodal data, is revolutionizing the diagnosis of pulmonary diseases. Figure 7A illustrates how hybrid AI frameworks combining cough sound analysis with chest imaging (X-rays and CT scans) can provide a more comprehensive understanding of a patient’s respiratory health. These systems leverage the strengths of both acoustic and visual data, improving diagnostic accuracy for diseases such as tuberculosis, COPD, and lung cancer. In Figure 7B, the workflow for an AI-based audio classifier is detailed. This system, which utilizes cough and breath sound recordings collected via smartphones, undergoes preprocessing and ML-based analysis to identify disease-specific features. Such audio classifiers have demonstrated remarkable potential for real-time diagnostic support and remote monitoring, making them particularly useful in resource-limited settings. Lastly, Figure 7C categorizes the various data types used in AI-driven pulmonary disease diagnosis, including bioacoustic signals and chest imaging. The figure underscores the broad range of diseases, ranging from TB and COPD to pneumonia and lung cancer, that can be diagnosed using these AI-powered tools. The combination of different data modalities enhances the precision of AI models, facilitating earlier detection and more personalized treatment strategies for respiratory conditions. These advancements position AI as a transformative tool in the future of respiratory medicine.

Figure 7: — Multimodal AI for pulmonary disease diagnosis integrating bioacoustics and imaging. (A) Hybrid AI frameworks and multimodal deep learning for pulmonary disease diagnosis. The applications enhance the diagnosis of pulmonary diseases by integrating bioacoustic analysis (cough and breath sounds) with chest imaging (X-rays and CT scans). The fusion of these technologies enables more personalized and accurate diagnostic methods, leveraging machine learning algorithms to extract patterns from both audio and visual inputs. (B) Audio classifier for pulmonary disease diagnosis. The process involved in the development and use of an AI-based audio classifier for pulmonary disease diagnosis. The workflow begins with the collection of bioacoustic data, particularly cough and breath sounds. The system illustrates how these classifiers can be integrated into mobile and telemedicine platforms, facilitating remote monitoring and early screening. (C) Data types and target diseases for AI in pulmonary diagnosis. The various data types used by AI models to diagnose pulmonary diseases and the specific diseases that can be identified using these AI-driven approaches, emphasizing the diverse range of diseases that AI tools can diagnose, showcasing how multimodal data enhances diagnostic accuracy and enables AI systems to identify disease markers across different data modalities.

The impact of AI across medical diagnosis in other disease

AI is making remarkable strides across a wide array of medical disciplines, revolutionizing diagnostics and patient care in ways that were once thought unimaginable. From andrology to psychiatry, AI is not only enhancing clinical decision-making but also offering new insights and capabilities that are driving the future of healthcare. These advancements underscore the power of AI to process large datasets, identify subtle patterns, and generate insights that are often overlooked by traditional diagnostic methods. In andrology and reproductive medicine, AI is proving to be a game-changer. The technology is now being used to optimize sperm, oocyte, and embryo selection processes, crucial elements in assisted reproductive technologies such as in vitro fertilization (IVF). AI algorithms can assess thousands of microscopic features and predict the viability of sperm and embryos with remarkable accuracy. This objective, data-driven approach is enhancing treatment outcomes by reducing the subjectivity of human assessment and enabling more personalized care. Beyond selection, AI is also being applied to predict surgical outcomes in reproductive surgeries, assess the cost-effectiveness of various treatments, and even assist in the development of robotic surgery systems [58]. This broad range of applications demonstrates AI’s versatile role in improving both the precision and efficiency of reproductive health interventions.

In the field of pediatric otolaryngology, AI is setting new standards for diagnosing ear conditions such as otitis media. Traditionally, otitis media has been diagnosed using handheld otoscopes, with pediatricians manually examining the eardrum for signs of infection. However, ML models have demonstrated superior diagnostic accuracy when applied to otoscopic images, often outperforming experienced clinicians in terms of both speed and precision [59]. By training AI on large datasets of otoscopic images, these systems are capable of detecting subtle features that may be missed in routine clinical examinations. This advancement not only promises to improve diagnostic accuracy but also to make otitis media diagnosis more accessible, especially in resource-limited settings where specialists may not be readily available.

Sleep medicine has also experienced a paradigm shift with AI integration. One of the key developments has been in the detection of night-time breathing disorders, particularly snoring and obstructive sleep apnea (OSA). DM-based analysis of snore sounds has emerged as a non-invasive, cost-effective method for identifying sleep apnea. AI models trained to analyze these sounds can detect patterns that correlate with the severity of OSA, offering a promising alternative to traditional diagnostic approaches such as polysomnography, which is expensive and labor-intensive [60]. Additionally, more innovative approaches have combined multiple data modalities, such as nasal airflow and blood oxygen signals, to improve the detection of OSA. Using dual-mode feature fusion CNNs, AI can analyze both airflow dynamics and oxygen saturation levels in real-time, providing a comprehensive picture of the patient’s sleep health [61]. This fusion of data sources enhances diagnostic accuracy, especially in the early stages of the disorder when symptoms may be less obvious.

AI’s diagnostic capabilities extend beyond traditional clinical disciplines into areas such as obstetrics, where it is being used to identify preeclampsia with remarkable accuracy. By analyzing ECG data, AI models can identify subtle electrocardiac patterns associated with preeclampsia, an often hard-to-diagnose condition that poses risks for both mothers and babies. The ability to detect preeclampsia early using AI not only improves maternal and fetal outcomes but also demonstrates the potential of AI to analyze complex biomedical signals in novel ways [62].

In psychiatry, AI is making strides in the diagnosis and treatment of mental health disorders. Traditional psychiatric diagnoses, such as those for depression, schizophrenia, and autism spectrum disorder, are often based on subjective assessments and patient interviews. AI, however, is capable of analyzing vast amounts of data from multiple sources, including clinical notes, brain imaging, and even behavioral patterns, to make more objective and accurate diagnoses. This approach is particularly beneficial in conditions like ADHD, addiction, and sleep disorders, where the boundaries between normal and abnormal behavior can be difficult to define. The continuous improvement of algorithmic models, fueled by large datasets and enhanced computational power, is driving more precise and tailored psychiatric care [63].

Despite these promising developments, the widespread implementation of AI in healthcare remains hampered by a lack of standardized validation procedures. For AI to be seamlessly integrated into clinical practice, there is a pressing need for robust, standardized datasets across institutions. Without such standards, AI models trained in one healthcare setting may fail to generalize effectively to others, compromising their diagnostic accuracy and clinical utility. The establishment of these datasets is critical to ensuring that AI technologies are not only effective but also reliable and transferable across diverse healthcare environments [64].

Nonetheless, the potential of AI to enhance diagnostic accuracy and streamline clinical workflows is undeniable. In every field where AI is being applied, from reproductive medicine to psychiatry, its ability to process vast amounts of patient data, detect patterns invisible to the human eye, and assist healthcare professionals in making more informed decisions is increasingly transforming patient care. As AI technologies continue to evolve, their impact on medicine will only grow, offering new possibilities for early diagnosis, personalized treatment, and improved patient outcomes across a wide range of medical specialties.

AI is playing a transformative role across various medical specialties by improving diagnostic accuracy, decision-making, and patient care. Figure 8A shows how AI’s data analysis and pattern recognition capabilities help identify disease markers from diverse data sources, such as medical images, bioacoustic signals, and clinical histories. In Figure 8B, AI provides objective decision support by assisting clinicians in interpreting complex patient data, which is particularly beneficial in specialties like psychiatry and obstetrics where subjective assessments are common. Figure 8C demonstrates the power of multimodal data integration, combining different diagnostic inputs (e.g., imaging and bioacoustic data) to provide a more comprehensive and accurate diagnosis. AI’s potential for early detection is showcased in Figure 8D, where it identifies subtle markers of diseases like sleep apnea and preeclampsia, enabling earlier intervention and better patient outcomes. Figure 8E highlights the automation of diagnostic processes, reducing clinician workload and ensuring quicker, more efficient diagnoses. Finally, Figure 8F illustrates how AI is revolutionizing personalized care by analyzing patient-specific data to recommend tailored treatment options, ensuring more precise and effective interventions in fields such as reproductive medicine and psychiatry. These advancements underline the transformative potential of AI in reshaping diagnostics and patient care across medical disciplines.

Figure 8: — AI’s role in enhancing diagnostics and patient care across medical disciplines. (A) Data analysis and pattern recognition in AI diagnostics for other diseases. The different types of data analysis and pattern recognition methods used by AI to diagnose various diseases. (B) Objective decision support in AI diagnostics for other Diseases. Objective decision support in the diagnostic process by automating and enhancing clinical decision-making. This system is particularly useful in specialties like psychiatry and obstetrics, where decision-making often involves complex, multifactorial inputs. (C) Multi-modal data integration in AI diagnostics for other diseases. The integration of multiple data modalities (e.g., clinical notes, imaging, speech, and bioacoustic signals) by AI systems to provide a holistic and more accurate diagnosis. This figure emphasizes how multimodal approaches improve the robustness and precision of AI diagnostic systems in various fields, including pediatric otolaryngology, sleep medicine, and obstetrics. (D) Early detection in AI diagnostics for other diseases. The different approaches to early detection across various disciplines, demonstrating how AI enables healthcare providers to act sooner, improving the prognosis for patients. (E) Diagnostic automation in AI for other diseases. It includes automated systems for image analysis, such as AI models that analyze otoscopic images for otitis media or chest X-rays for lung conditions. (F) Personalized care through AI diagnostics for other diseases. AI enhances personalized care by tailoring treatment plans based on individual patient data. By analyzing large datasets of medical histories, genomics, imaging, and even behavioral patterns, AI can recommend personalized treatment options that are specifically suited to a patient’s condition, and preferences.

The future is now

The integration of AI-driven diagnostic tools, particularly through wearable sensors and intelligent point-of-care (POC) tests, is dramatically reshaping the landscape of personalized healthcare. These technologies enable continuous, real-time physiological monitoring and automated sample analysis, laying the groundwork for a more proactive, patient-centered approach to diagnostics. By providing constant, non-invasive data collection, AI-enhanced wearables can detect early signs of health changes, alerting clinicians and patients to potential issues before symptoms appear. This shift from reactive to proactive healthcare is especially crucial as hospitals and health systems work to build the infrastructure necessary to support AI’s full integration into clinical practice. The widespread adoption of these AI-powered diagnostic systems is poised to revolutionize medical care, ushering in a new era of precision medicine where diagnoses are tailored to the individual, rather than relying on generalized, one-size-fits-all approaches.

At the American Association for Cancer Research (AACR) Annual Meeting 2024, experts highlighted the profound impact AI is having on patient care. During a plenary session, AI’s role in accelerating evidence generation, advancing disparities research, and improving clinical trial designs was a major focal point. Dr. Vivek Subbiah underscored that “the future is now”, signifying that the integration of AI into clinical practice is no longer a futuristic aspiration but an ongoing, tangible reality. AI is already being utilized to streamline trial designs, enhance recruitment, and predict patient responses to therapies, ultimately making clinical trials more efficient and accessible. Furthermore, AI’s capacity to analyze vast and diverse datasets, from imaging to molecular genomics, allows for better-informed clinical decisions that consider not only the patient’s current condition but also their unique genetic makeup and environmental factors. However, despite these groundbreaking advancements, significant challenges remain in translating AI models from research and development into everyday clinical practice. One of the most pressing issues is the complexity of integrating multi-disease diagnostic AI models into established clinical workflows. While ML models have shown promise in individual diagnostic areas, the broader adoption of AI in clinical settings faces substantial hurdles. The challenge is not only technological but also operational, as healthcare providers must adapt to new methods of data collection, analysis, and interpretation. Furthermore, the risk of AI models being perceived as black boxes where clinicians cannot easily interpret the rationale behind AI-driven decisions complicates their acceptance in clinical settings. To fully integrate AI into medical practice, the tools must be transparent, explainable, and trusted by clinicians, who must feel confident in their ability to interpret and act on the AI’s recommendations.

This is particularly evident in the field of cancer diagnostics and treatment planning, where the disease’s complexity requires AI models capable of analyzing large, heterogeneous datasets. These datasets can include everything from imaging and genomics to clinical data, each containing unique insights that must be synthesized for a comprehensive understanding of the disease. While AI models have demonstrated great potential in enhancing diagnostic accuracy and treatment strategies, they often require a level of sophistication that can be difficult to implement in clinical practice. The complexity of cancer itself, its various subtypes, the intricacies of tumor heterogeneity, and the variability in patient responses, require AI systems that can integrate multiple forms of data and adapt to evolving medical knowledge. Additionally, validation and evaluation of these AI models remain key challenges. Ensuring that AI diagnostic systems perform accurately across diverse patient populations and healthcare settings requires rigorous clinical trials and extensive real-world testing. While AI models have demonstrated success in specific applications, such as predicting sepsis shock, identifying high-risk postpartum depression patients, and screening for SARS-CoV-2 infection through blood tests, their reliability across varied patient groups, treatment protocols, and geographic regions is still under scrutiny [65], 66]. The diversity of patient demographics, as well as the differences in healthcare infrastructure across institutions, presents a challenge to the generalization of these models. Models that perform well in one healthcare setting or for one demographic may not necessarily yield the same results in another. To address this, a concerted effort must be made to create standardized datasets and establish multi-center trials that ensure AI tools can be applied universally and effectively. Furthermore, the complexity of integrating AI into the diagnostic workflow also raises concerns about healthcare equity. While AI has the potential to improve access to high-quality diagnostics, especially in underserved areas, its implementation may inadvertently deepen disparities if it’s not accessible to all populations. AI tools require significant infrastructure, including high-speed internet, access to advanced computing resources, and skilled personnel to operate the technology. In low-resource settings, where these prerequisites may not be available, the potential benefits of AI could be limited, potentially exacerbating existing health disparities.

Technical and ethical considerations

The integration of AI into medical diagnostics brings significant opportunities but also presents major technical and ethical challenges. As AI systems become increasingly complex and widespread in clinical settings, their technical design and ethical implications will determine their actual impact on patient care.

Explainability is one of the core technical challenges AI faces in healthcare. Many DM algorithms, such as “black-box” systems, can generate highly accurate predictions but lack transparency in their decision-making processes. This opacity poses a major challenge for clinicians, who need to understand the reasoning behind AI-generated conclusions. Developing explainable AI (XAI) systems is therefore crucial. Such systems must provide explanations that are clinically understandable, reliable, and consistent with medical knowledge [67]. In terms of regulation, medical AI faces complex approval processes. The U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMA) are working to establish frameworks to assess the safety and effectiveness of AI diagnostic tools. A particular challenge is the dynamic nature of AI systems, which continuously learn from new data. This is in stark contrast to traditional, static medical devices and calls for new regulatory models to balance innovation with patient safety [68]. The most critical ethical concern is bias. AI models are typically trained on historical data, which may reflect systemic biases present in past medical practices. If training data inadequately represent different racial, gender, or socioeconomic groups, AI may unintentionally exacerbate existing healthcare inequalities. For example, in dermatology, AI systems trained only on data from individuals with lighter skin tones may struggle to accurately diagnose conditions in individuals with darker skin tones. Addressing these challenges requires multi-faceted collaboration, including the development of more representative and inclusive training datasets, the design of fairness-sensitive algorithms, the assurance that AI tools benefit all patient groups, and the establishment of transparent and traceable AI decision-making mechanisms.

Current limitations and regulatory framework of diagnostic AI

The key limitation of medical diagnostic AI is its generalization ability. While AI performs exceptionally well in specific, well-defined tasks, such as cancer imaging detection or sepsis prediction, it often struggles with the broad range of medical conditions that clinicians encounter on a daily basis. Research by the National Institutes of Health (NIH) has shown that even in areas with high accuracy, AI still frequently makes errors when interpreting images or providing decision explanations [69]. Another foundational challenge is data quality. Medical data is often fragmented, inconsistent, and incomplete. There are significant differences in data collection standards across healthcare institutions, and these inconsistencies can severely impact the accuracy and reliability of AI models. Furthermore, strict privacy regulations, such as the U.S. HIPAA and the EU’s GDPR, further complicate the acquisition and use of data [70]. This includes the development of new testing paradigms that allow AI tools to be evaluated in more realistic, dynamic clinical settings. A key focus for the FDA is understanding that different AI applications require distinct regulatory approaches. For example, AI models designed for rule-out or triage purposes, which help clinicians prioritize patients based on the severity of their condition, face different regulatory scrutiny compared to AI tools intended to improve diagnostic accuracy or support complex treatment decisions. This nuanced regulatory approach is necessary because AI models used for initial assessments might have lower thresholds for error than those used for definitive diagnostic conclusions [71]. Additionally, the FDA has begun exploring the potential of synthetic data to address the challenges of data scarcity and variability. Synthetic data, generated through computer simulations or algorithms, can supplement real-world datasets, providing AI models with a more robust foundation for learning while maintaining privacy and confidentiality [72]. This represents a significant step toward mitigating the data limitations currently hindering AI’s full potential.

Regulatory bodies are actively addressing these challenges. In March 2024, the European Union’s AI Act (AIA) mandated that medical AI decision-support systems must include explainability features to ensure that healthcare providers can understand the reasoning behind AI decisions [73]. The EU’s regulatory framework sets a global precedent, mandating that all high-risk AI systems, particularly those used in medical diagnostics, adhere to these standards of transparency. Building on the AIA, the FDA’s CDRH issued guiding principles for ML-enabled medical device transparency in June 2024. These principles emphasize the importance of ensuring that AI systems are explainable, traceable, and accountable, especially as their use becomes more widespread in clinical decision-making [74]. This growing focus on explainability and transparency in AI design is essential for fostering clinician confidence and ensuring that AI tools can be integrated safely and effectively into clinical workflows.

Ultimately, successfully integrating AI into medical diagnostics requires interdisciplinary collaboration. Developers, clinicians, administrators, and policymakers must work together to maximize the potential of AI technology while ensuring patient safety, equity, and ethical integrity. This means not only developing advanced technologies but also building trust, ensuring fairness, and creating a more personalized, efficient, and equitable healthcare system.

A new era for diagnostic AI

The integration of AI with modern technologies, such as 5G, big data, and the Internet of Things (IoT), is reshaping the field of medical diagnostics in profound ways. These technologies are not only enabling faster, more accurate diagnostic capabilities but are also laying the foundation for a new era of cloud-based medical AI. With the rapid adoption of 5G, healthcare systems are gaining access to low-latency, high-speed networks that enable real-time transmission of medical data from patients, medical devices, and imaging systems. This connectivity improves the efficiency of AI-based diagnostic tools, ensuring quicker results and more timely interventions. Moreover, the expansion of big data is empowering AI algorithms to analyze vast quantities of medical information. From patient histories and imaging data to real-time physiological monitoring, AI can now process and interpret a broader spectrum of patient data to uncover patterns that may not be immediately evident to human clinicians. As cloud computing infrastructure becomes increasingly robust, offering scalable, reliable, and secure resources for computing, storage, and networking, medical AI systems can access virtually unlimited processing power, transforming diagnostics and enabling personalized care at a global scale [75]. These technological advances are particularly vital for fields such as medical imaging, where AI’s ability to analyze and interpret complex images – whether from MRIs, CT scans, or X-rays, is being enhanced by faster data processing speeds. This convergence is helping AI move from niche applications to mainstream clinical tools, making it feasible to deploy AI across a range of diagnostic tasks, from routine screenings to emergency care.

One of the most exciting developments in the AI field is the emergence of federated learning (FL) for decentralized AI. Unlike traditional centralized AI models, which require collecting and storing large volumes of patient data in a single location for training purposes, FL enables AI models to be trained across distributed datasets that remain at their original sources, such as hospitals or medical research centers. This approach minimizes the need for data centralization while ensuring that sensitive patient data never leaves its local institution, a key consideration in maintaining privacy and compliance with regulations like GDPR or HIPAA. FL is particularly impactful in medical diagnostics because it allows for collaboration and data-sharing across multiple institutions without the risk of compromising patient confidentiality. International collaboration in medical AI development becomes more feasible as institutions can contribute to training robust AI models without violating data sovereignty rules or exposing sensitive patient information. Research has shown that FL can achieve performance comparable to traditional centralized methods, even with relatively small datasets or under stringent data protection rules [76]. This advancement could pave the way for global AI-driven healthcare solutions, enabling researchers and clinicians worldwide to collaborate on improving diagnostic models, accelerating breakthroughs, and addressing global health disparities by ensuring that AI tools are trained on diverse, representative datasets. In practice, FL can help improve diagnostic accuracy for rare or emerging diseases by aggregating knowledge from multiple sources without compromising local data security. For example, AI models trained across a network of hospitals could enhance their ability to diagnose conditions like rare cancers or neurological disorders, where data from different regions or institutions may be sparse but valuable (Figure 9).

Figure 9: — The convergence of cutting-edge technologies: a new era for diagnostic AI. At the top of the figure, cloud computing provides scalable infrastructure for AI model development. Below, 5G enables real-time data transmission, and big data enhances AI’s ability to analyze vast, complex datasets. FL enables decentralized training, ensuring patient data privacy while promoting collaboration across institutions.

As AI becomes increasingly integrated into healthcare systems, it is critical to strike a careful balance between innovation and patient safety. While AI promises to revolutionize diagnostics by providing quicker, more accurate results, human clinical expertise remains essential for interpreting complex, context-dependent medical data. AI systems, though powerful, are not infallible, they may misinterpret data or overlook subtle nuances that human clinicians would catch. Therefore, the role of AI should be seen as complementary, rather than substituting for healthcare professionals. The ideal scenario is one where AI and clinicians work in tandem, with AI tools providing faster insights, freeing up clinicians to focus on patient care and decision-making. The path to successful AI integration involves ensuring that AI tools are not only effective but also reliable throughout their life-cycle. As AI systems are continuously updated, they must undergo rigorous post-deployment evaluations to ensure they maintain their diagnostic accuracy over time. This includes evaluating AI performance in diverse healthcare settings, where real-world conditions may vary from the controlled environments in which the algorithms were originally trained. Furthermore, AI algorithms must be able to adapt to changes in medical practice, new treatment protocols, and the emergence of new diseases.

The advent of DeepSeek represents a significant advancement in AI capabilities, with particularly promising implications for medical diagnostics [77]. DeepSeek’s advanced language modeling capabilities introduce several key advantages in medical diagnostics [78]. These include improved interpretation of complex medical data through enhanced pattern recognition, more sophisticated analysis of medical literature and clinical notes, better integration of multimodal medical data (including imaging and textual information), and an enhanced ability to identify subtle diagnostic indicators that might be missed by traditional methods [77]. The integration of DeepSeek with current medical AI frameworks also shows promising developments. It augments existing diagnostic algorithms with more sophisticated language understanding, enhances the ability to process and analyze unstructured medical data, improves accuracy in medical report generation and interpretation, and provides better context awareness in CDS systems [79]. DeepSeek’s capabilities particularly benefit precision medicine by enabling more accurate patient-specific risk assessment, better prediction of treatment outcomes based on comprehensive data analysis, enhanced understanding of disease progression patterns, and improved identification of potential treatment options based on individual patient characteristics [80].

Looking ahead, AI in medical diagnostics will likely play a central role in advancing personalized, predictive, and preventive medicine. AI’s ability to analyze massive amounts of data from patient histories, genomic information, and environmental factors positions it to predict individual health risks with a level of precision that was previously unattainable. AI could help identify patients at high risk for diseases before symptoms even appear, allowing for earlier, more targeted interventions. This predictive capability could revolutionize fields like oncology, where AI systems could identify early-stage cancers through analysis of subtle changes in imaging or biomarkers, even in the absence of overt symptoms. Furthermore, AI-driven precision medicine could tailor treatment plans based on a patient’s unique genetic makeup, lifestyle factors, and previous medical history. For instance, AI models could suggest personalized drug regimens or identify the most effective therapeutic approaches for individuals, minimizing trial-and-error in treatment decisions and improving overall patient outcomes. The future also holds promise for AI in preventive healthcare. With continuous monitoring enabled by wearables and connected medical devices, AI can track patients’ health metrics in real time, identifying early warning signs of chronic diseases like diabetes, heart disease, or hypertension. By providing ongoing feedback to patients and healthcare providers, AI can empower individuals to take a more proactive role in managing their health, potentially reducing the burden of preventable diseases on healthcare systems.

Contributions and limitations

Our review makes several significant contributions to the field of AI-driven medical diagnostics. We employ the innovative GraphRAG methodology to conduct a comprehensive analysis of research publications from 2022 to 2024. This represents the first systematic application of this approach in medical AI literature review, enabling us to map complex relationships between different research domains, identify emerging technological trends, quantify research impact across various medical fields, and track the evolution of AI applications in real-time. Our cross-domain integration approach provides unique insights by examining AI applications across cancer, AD, and diabetes. This comprehensive perspective reveals common technological patterns and methodologies, transferable solutions between different disease domains, shared challenges and successful implementation strategies, and opportunities for cross-pollination of ideas.

Despite these contributions, we acknowledge several limitations in our review. The GraphRAG methodology, while innovative, may not capture all relevant publications due to database limitations. The rapid pace of AI development means some very recent advances may not be included. Additionally, our focus on English-language publications may exclude valuable research from other languages. While our review is comprehensive within the chosen disease domains, other important medical fields are not covered. Economic and implementation aspects of AI diagnostics require deeper exploration. We also provide a limited discussion of regulatory frameworks and approval processes, and there is a need for more detailed analysis of AI’s impact on healthcare workforce dynamics. These limitations suggest several directions for future research, including expansion of the analysis to include additional disease domains, integration of economic and regulatory perspectives, investigation of implementation challenges in different healthcare systems, and examination of AI’s impact on healthcare provider roles and responsibilities. Our review highlights several critical areas for future investigation. These include the development of standardized validation protocols for AI diagnostic tools, integration of diverse data sources for more comprehensive diagnostic approaches, implementation strategies for different healthcare settings, and methods to ensure equitable access to AI-driven diagnostic technologies.

This comprehensive review provides a foundation for understanding both the current state and future trajectory of AI in medical diagnostics. By acknowledging both our contributions and limitations, we aim to provide valuable insights while encouraging further research in this rapidly evolving field. As AI continues to transform medical diagnostics, such critical analysis becomes increasingly important for guiding future developments and ensuring optimal implementation of these transformative technologies. This synthesis of current progress emphasizes AI’s potential to revolutionize medical diagnostics through enhanced accuracy, early detection, and personalized patient care, while maintaining a realistic perspective on the challenges and limitations that must be addressed for successful implementation.

Supplementary Material

j_mr-2024-0097_suppl_001.csv^{(477.1KB, csv)}

Acknowledgments

We gratefully acknowledge the contributions of Peking University. Special thanks to Xunbin Wei for invaluable assistance.

Supplementary Material

This article contains supplementary material (https://doi.org/10.1515/mr-2024-0097).

Footnotes

Research ethics: Not applicable.

Informed consent: Not applicable.

Author contributions: Junyu Zhou: Conceptualization, methodology, writing – original draft, project administration, visualization; Sunmin Park: Data curation, formal analysis; Sihan Dong: Investigation, resources; Xiaoying Tang: writing – review & editing; Xunbin Wei: Supervision, funding acquisition, validation.

Use of Large Language Models, AI and Machine Learning Tools: All AI-generated content was critically reviewed and validated by human researchers.

Conflict of interest: There are no conflicts of interest.

Research funding: This work was supported by the National Key Research and Development Program of China (Grant No. 2021YFF0502900) and the Special Fund for Research on National Major Research Instruments of China (Grant No. 62027824).

Data availability: All data generated or analyzed during this study are included in the manuscript and supplementary information files.

References

1.Alhur A, Alhur AA, Hamdan Alghamdi SJ, Asiri ZI, Saeed Alzamil SK, Alfateih S, et al. Advancing the frontiers of artificial intelligence in transforming healthcare: a comprehensive literature review. J Popul Ther Clin Pharmacol. 2023;30:2107–12. doi: 10.53555/jptcp.v30i19.4744. [DOI] [Google Scholar]
2.Zeb S, Nizamullah F, Abbasi N, Fahad M. AI in healthcare: revolutionizing diagnosis and therapy. Int J Multidiscip Sci Arts. 2024;3:118–28. [Google Scholar]
3.Nazi ZA, Peng W. Large language models in healthcare and medical domain: a review. Informatics. 2024;11:57. doi: 10.3390/informatics11030057. [DOI] [Google Scholar]
4.Dias R, Torkamani A. Artificial intelligence in clinical and genomic diagnostics. Genome Med. 2019;11:70. doi: 10.1186/s13073-019-0689-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Poalelungi DG, Musat CL, Fulga A, Neagu M, Neagu A-I, Piraianu AI, et al. Advancing patient care: how artificial intelligence is transforming healthcare. J Personalized Med. 2023;13 doi: 10.3390/jpm13081214. [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Chugh V, Basu A, Kaushik A, Bhansali S, Basu AK. Employing nano-enabled artificial intelligence (AI)-based smart technologies for prediction, screening, and detection of cancer. Nanoscale. 2024;16:5458–86. doi: 10.1039/d3nr05648a. [DOI] [PubMed] [Google Scholar]
7.Pang K, Song Z, Liu Y, Sun H, Zhang R, Fu Y, et al. In vivo dynamic monitoring of circulating melanoma cells and the inhibitory effect of PD-L1 inhibitor based on PAFC equipped with a deep learning framework. APL Photon. 2024;9 doi: 10.1063/5.0226328. [DOI] [Google Scholar]
8.Edelson DP, Churpek MM, Carey KA, Lin Z, Huang C, Siner JM, et al. Early warning scores with and without artificial intelligence. JAMA Netw Open. 2024;7:e2438986. doi: 10.1001/jamanetworkopen.2024.38986. [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Berge GT, Granmo OC, Tveit TO, Munkvold BE, Ruthjersen AL, Sharma J. Machine learning-driven clinical decision support system for concept-based searching: a field trial in a Norwegian hospital. BMC Med Inf Decis Making. 2023;23:5. doi: 10.1186/s12911-023-02101-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Bozyel S, Şimşek E, Koçyiğit D, Güler A, Korkmaz Y, Şeker M, et al. Artificial intelligence-based clinical decision support systems in cardiovascular diseases. Anatol J Cardiol. 2024;28:74. doi: 10.14744/anatoljcardiol.2023.3685. [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Talati D. AI in healthcare domain. J Knowl Learn Sci Technol. 2023 ISSN: 2959-6386 (online) [Google Scholar]
12.Kaul V, Enslin S, Gross SA. History of artificial intelligence in medicine. Gastrointest Endosc. 2020;92:807–12. doi: 10.1016/j.gie.2020.06.040. [DOI] [PubMed] [Google Scholar]
13.Moawad AW, Fuentes DT, ElBanan MG, Shalaby AS, Guccione J, Kamel S, et al. Artificial intelligence in diagnostic radiology: where do we stand, challenges, and opportunities. J Comput Assist Tomogr. 2022;46:78–90. doi: 10.1097/rct.0000000000001247. [DOI] [PubMed] [Google Scholar]
14.Nozari H, Ghahremani-Nahr J, Szmelter-Jarosz A. AI and machine learning for real-world problems. Adv Comput. 2024;134:1–12. doi: 10.1016/bs.adcom.2023.02.001. [DOI] [Google Scholar]
15.Senkamalavalli R, Sankar S, Parivazhagan A, Raja R, Yoganand S, Srinivas P, et al. Enhancing clinical decision-making with cloud-enabled integration of image-driven insights. Indones J Electr Eng Comput Sci. 2024;36:338–46. doi: 10.11591/ijeecs.v36.i1.pp338-346. [DOI] [Google Scholar]
16.Moulaei K, Yadegari A, Baharestani M, Farzanbakhsh S, Sabet B, Afrash MR. Generative artificial intelligence in healthcare: a scoping review on benefits, challenges and applications. Int J Med Inf. 2024:105474. doi: 10.1016/j.ijmedinf.2024.105474. [DOI] [PubMed] [Google Scholar]
17.Chen X. AI in healthcare: revolutionizing diagnosis and treatment through machine learning. MZ J Artif Intell. 2024;1 [Google Scholar]
18.Edge D, Trinh H, Cheng N, Bradley J, Chao A, Mody A, et al. From local to global: a graph RAG approach to query-focused summarization. ArXiv. 2024 abs/2404.16130. [Google Scholar]
19.Swanson K, Wu E, Zhang A, Alizadeh AA, Zou J. From patterns to patients: advances in clinical machine learning for cancer diagnosis, prognosis, and treatment. Cell. 2023;186:1772–91. doi: 10.1016/j.cell.2023.01.035. [DOI] [PubMed] [Google Scholar]
20.Liu Y, Liu G, Zhang Q. Deep learning and medical diagnosis. Lancet. 2019;394:1709–10. doi: 10.1016/s0140-6736(19)32501-2. [DOI] [PubMed] [Google Scholar]
21.Chen X, Wang X, Zhang K, Fung KM, Thai TC, Moore K, et al. Recent advances and clinical applications of deep learning in medical image analysis. Med Image Anal. 2022;79:102444. doi: 10.1016/j.media.2022.102444. [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Wang X, Zhao J, Marostica E, Yuan W, Jin J, Zhang J, et al. A pathology foundation model for cancer diagnosis and prognosis prediction. Nature. 2024;634:970–8. doi: 10.1038/s41586-024-07894-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Tu T, Palepu A, Schaekermann M, Saab K, Freyberg J, Tanno R, et al. Towards conversational diagnostic AI. . 2024 doi: 10.1038/s41586-025-08866-7. arXiv preprint arXiv:240105654. [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Xie X, Yu W, Wang L, Yang J, Tu X, Liu X, et al. SERS-based AI diagnosis of lung and gastric cancer via exhaled breath. Spectrochim Acta Mol Biomol Spectrosc. 2024;314:124181. doi: 10.1016/j.saa.2024.124181. [DOI] [PubMed] [Google Scholar]
25.Wani NA, Kumar R, Bedi J. DeepXplainer: an interpretable deep learning based approach for lung cancer detection using explainable artificial intelligence. Comput Methods Progr Biomed. 2024;243:107879. doi: 10.1016/j.cmpb.2023.107879. [DOI] [PubMed] [Google Scholar]
26.Kale MB, Wankhede NL, Pawar RS, Ballal S, Kumawat R, Goswami M, et al. AI-driven innovations in Alzheimer’s disease: integrating early diagnosis, personalized treatment, and prognostic modelling. Ageing Res Rev. 2024:102497. doi: 10.1016/j.arr.2024.102497. [DOI] [PubMed] [Google Scholar]
27.Mikhael PG, Wohlwend J, Yala A, Karstens L, Xiang J, Takigami AK, et al. Sybil: a validated deep learning model to predict future lung cancer risk from a single low-dose chest computed tomography. J Clin Oncol. 2023;41:2191–200. doi: 10.1200/jco.22.01345. [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Vedula RS, Karp HQ, Koob J, Lim F, Garcia JS, Winer ES, et al. CRISPR-based rapid molecular diagnostic tests for fusion-driven leukemias. Blood. 2024;144:1290–9. doi: 10.1182/blood.2023022908. [DOI] [PubMed] [Google Scholar]
29.Placido D, Yuan B, Hjaltelin JX, Zheng C, Haue AD, Chmura PJ, et al. A deep learning algorithm to predict risk of pancreatic cancer from disease trajectories. Nat Med. 2023;29:1113–22. doi: 10.1038/s41591-023-02332-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Jack CR, Andrews JS, Beach TG, Buracchio T, Dunn B, Graf A, et al. Revised criteria for diagnosis and staging of Alzheimer’s disease: Alzheimer’s Association Workgroup. Alzheimer’s Dement. 2024;20:5143–69. doi: 10.1002/alz.13859. [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Zolnoori M, Zolnour A, Topaz M. ADscreen: a speech processing-based screening system for automatic identification of patients with Alzheimer’s disease and related dementia. Artif Intell Med. 2023;143:102624. doi: 10.1016/j.artmed.2023.102624. [DOI] [PMC free article] [PubMed] [Google Scholar]
32.O’Malley RPD, Mirheidari B, Harkness K, Reuber M, Venneri A, Walker T, et al. Fully automated cognitive screening tool based on assessment of speech and language. J Neurol Neurosurg Psychiatry. 2020;92:12–5. doi: 10.1136/jnnp-2019-322517. [DOI] [PubMed] [Google Scholar]
33.Leming M, Das S, Im H. Adversarial confound regression and uncertainty measurements to classify heterogeneous clinical MRI in Mass General Brigham. PLoS One. 2023;18:e0277572. doi: 10.1371/journal.pone.0277572. [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Amini S, Hao B, Yang J, Karjadi C, Kolachalama VB, Au R, et al. Prediction of Alzheimer’s disease progression within 6 years using speech: a novel approach leveraging language models. Alzheimer’s Dement. 2024;20:5262–70. doi: 10.1002/alz.13886. [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Mohsen F, Al-Absi HRH, Yousri NA, El Hajj N, Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. NPJ Digit Med. 2023;6:197. doi: 10.1038/s41746-023-00933-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Huang J, Yeung AM, Armstrong DG, Battarbee AN, Cuadros J, Espinoza JC, et al. Artificial intelligence for predicting and diagnosing complications of diabetes. J Diabetes Sci Technol. 2023;17:224–38. doi: 10.1177/19322968221124583. [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Gudiño-Ochoa A, García-Rodríguez JA, Ochoa-Ornelas R, Cuevas-Chávez JI, Sánchez-Arias DA. Noninvasive diabetes detection through human breath using TinyML-Powered E-Nose. Sensors. 2024;24:1294. doi: 10.3390/s24041294. [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Eben H, Bolanle M. Deep learning for diabetes diagnosis: convolutional neural networks and recurrent neural networks. MZ J Artif Intell. 2024;1 [Google Scholar]
39.Fu H, Tian Y, Zha G, Xiao X, Zhu H, Zhang Q, et al. Microstrip isoelectric focusing with deep learning for simultaneous screening of diabetes, anemia, and thalassemia. Anal Chim Acta. 2024;1312:342696. doi: 10.1016/j.aca.2024.342696. [DOI] [PubMed] [Google Scholar]
40.Iftikhar M, Saqib M, Qayyum SN, Asmat R, Mumtaz H, Rehan M, et al. Artificial intelligence-driven transformations in diabetes care: a comprehensive literature review. Ann Med Surg (Lond). 2024;86:5334–42. doi: 10.1097/ms9.0000000000002369. [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Elias P, Jain SS, Poterucha T, Randazzo M, Lopez Jimenez F, Khera R, et al. Artificial intelligence for cardiovascular care—part 1: advances: JACC review topic of the week. J Am Coll Cardiol. 2024;83:2472–86. doi: 10.1016/j.jacc.2024.03.400. [DOI] [PubMed] [Google Scholar]
42.Ogawa S, Namino F, Mori T, Sato G, Yamakawa T, Saito S. AI diagnosis of heart sounds differentiated with super StethoScope. J Cardiol. 2024;83:265–71. doi: 10.1016/j.jjcc.2023.09.007. [DOI] [PubMed] [Google Scholar]
43.Omarov B, Saparkhojayev N, Shekerbekova S, Akhmetova O, Sakypbekova M, Kamalova G, et al. Artificial intelligence in medicine: real time electronic stethoscope for heart diseases detection. Comput Mater Continua (CMC) 2022;70 doi: 10.32604/cmc.2022.019246. [DOI] [Google Scholar]
44.Thoenes M, Agarwal A, Grundmann D, Ferrero C, McDonald A, Bramlage P, et al. Narrative review of the role of artificial intelligence to improve aortic valve disease management. J Thorac Dis. 2021;13:396. doi: 10.21037/jtd-20-1837. [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Baikuvekov M, Baikuvekova A, Sultan D. Development of a digital stethoscopy system for detection of heart diseases; 2024 IEEE 4th International Conference on Smart Information Systems and Technologies (SIST). IEEE; 2024. [Google Scholar]
46.Khanna NN, Maindarkar MA, Viswanathan V, Puvvula A, Paul S, Bhagawati M, et al. Cardiovascular/stroke risk stratification in diabetic foot infection patients using deep learning-based artificial intelligence: an investigative study. J Clin Med. 2022;11:6844. doi: 10.3390/jcm11226844. [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Tatsugami F, Nakaura T, Yanagawa M, Fujita S, Kamagata K, Ito R, et al. Recent advances in artificial intelligence for cardiac CT: enhancing diagnosis and prognosis prediction. Diagn Interv Imaging. 2023;104:521–8. doi: 10.1016/j.diii.2023.06.011. [DOI] [PubMed] [Google Scholar]
48.Gutfleisch M, Ester O, Aydin S, Quassowski M, Spital G, Lommatzsch A, et al. Clinically applicable deep learning-based decision aids for treatment of neovascular AMD. Graefes Arch Clin Exp Ophthalmol. 2022;260:2217–30. doi: 10.1007/s00417-022-05565-1. [DOI] [PubMed] [Google Scholar]
49.Zhou Y, Chia MA, Wagner SK, Ayhan MS, Williamson DJ, Struyven RR, et al. A foundation model for generalizable disease detection from retinal images. Nature. 2023;622:156–63. doi: 10.1038/s41586-023-06555-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Qian B, Sheng B, Chen H, Wang X, Li T, Jin Y, et al. A competition for the diagnosis of myopic maculopathy by artificial intelligence algorithms. JAMA Ophthalmol. 2024 doi: 10.1001/jamaophthalmol.2024.3707. [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Dkhan R, Hamdan R. Respiratory diseases detection and classification based on respiratory voice using artificial intelligence methods. SVUPedia. 2024. pp. 36–45.
52.Ghrabli S, Elgendi M, Menon C. Identifying unique spectral fingerprints in cough sounds for diagnosing respiratory ailments. Sci Rep. 2024;14:593. doi: 10.1038/s41598-023-50371-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Isangula KG, Haule RJ. Leveraging AI and machine learning to develop and evaluate a contextualized user-friendly cough audio classifier for detecting respiratory diseases: protocol for a diagnostic study in rural Tanzania. JMIR Res Protoc. 2024;13:e54388. doi: 10.2196/54388. [DOI] [PMC free article] [PubMed] [Google Scholar]
54.Saeed T, Ijaz A, Sadiq I, Qureshi HN, Rizwan A, Imran A. An AI-enabled bias-free respiratory disease diagnosis model using cough audio. Bioengineering. 2024;11:55. doi: 10.3390/bioengineering11010055. [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Cooper M. AI and spotting the sound of illness. ITNOW. 2024;66:24–5. doi: 10.1093/itnow/bwae012. [DOI] [Google Scholar]
56.Malik H, Anees T. Multi-modal deep learning methods for classification of chest diseases using different medical imaging and cough sounds. PLoS One. 2024;19:e0296352. doi: 10.1371/journal.pone.0296352. [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Eswara Rao G, Rajitha B. HQF-CC: hybrid framework for automated respiratory disease detection based on quantum feature extractor and custom classifier model using chest X-rays. Int J Inf Technol. 2024;16:1145–53. doi: 10.1007/s41870-023-01681-1. [DOI] [Google Scholar]
58.Abou Ghayda R, Cannarella R, Calogero AE, Shah R, Rambhatla A, Zohdy W, et al. Artificial intelligence in andrology: from semen analysis to image diagnostics. World J Mens Health. 2024;42:39. doi: 10.5534/wjmh.230050. [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Suresh K, Wu MP, Benboujja F, Christakis B, Newton A, Hartnick CJ, et al. AI model versus clinician otoscopy in the operative setting for otitis media diagnosis. Otolaryngol Head Neck Surg. 2024;170:1598–601. doi: 10.1002/ohn.559. [DOI] [PubMed] [Google Scholar]
60.Dang B, Ma D, Li S, Qi Z, Zhu EY. Deep learning-based snore sound analysis for the detection of night-time breathing disorders. Appl Comput Eng. 2024;76:109–14. [Google Scholar]
61.Peng D, Yue H, Tan W, Lei W, Chen G, Shi W, et al. A bimodal feature fusion convolutional neural network for detecting obstructive sleep apnea/hypopnea from nasal airflow and oximetry signals. Artif Intell Med. 2024;150:102808. doi: 10.1016/j.artmed.2024.102808. [DOI] [PubMed] [Google Scholar]
62.Butler L, Gunturkun F, Chinthala L, Karabayir I, Tootooni MS, Bakir-Batu B, et al. AI-based preeclampsia detection and prediction with electrocardiogram data. Front Cardiovasc Med. 2024;11:1360238. doi: 10.3389/fcvm.2024.1360238. [DOI] [PMC free article] [PubMed] [Google Scholar]
63.Sadasivan VS, Kumar A, Balasubramanian S, Wang W, Feizi S. Can AI-generated text be reliably detected? . 2023 arXiv preprint arXiv:230311156. [Google Scholar]
64.Tsopra R, Fernandez X, Luchinat C, Alberghina L, Lehrach H, Vanoni M, et al. A framework for validating AI in precision medicine: considerations from the European ITFoC consortium. BMC Med Inf Decis Making. 2021;21:274. doi: 10.1186/s12911-021-01634-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
65.Sandhu S, Lin AL, Brajer N, Sperling J, Ratliff W, Bedoya AD, et al. Integrating a machine learning system into clinical workflows: qualitative study. J Med Internet Res. 2020;22:e22421. doi: 10.2196/22421. [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Wang F, Beecy A. Implementing AI models in clinical workflows: a roadmap. BMJ Evid-Based Med. 2024 doi: 10.1136/bmjebm-2023-112727. [DOI] [PMC free article] [PubMed] [Google Scholar]
67.Phillips RA, Jani J, Bradley SK. Exploring the literature on artificial intelligence use in oncology. J Clin Oncol. 2024;42:e13642–e. doi: 10.1200/jco.2024.42.16_suppl.e13642. [DOI] [Google Scholar]
68.Varghese J. Artificial intelligence in medicine: chances and challenges for wide clinical adoption. Visc Med. 2020;36:443–9. doi: 10.1159/000511930. [DOI] [PMC free article] [PubMed] [Google Scholar]
69.Jin Q, Chen F, Zhou Y, Xu Z, Cheung JM, Chen R, et al. Hidden flaws behind expert-level accuracy of multimodal GPT-4 vision in medicine. npj Digit Med. 2024;7:190. doi: 10.1038/s41746-024-01185-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
70.Blumenthal D, Patel B. The regulation of clinical artificial intelligence. NEJM AI. 2024;1:AIpc2400545. doi: 10.1056/aipc2400545. [DOI] [Google Scholar]
71.Romagnoli A, Ferrara F, Langella R, Zovi A. Healthcare systems and artificial intelligence: focus on challenges and the international regulatory framework. Pharm Res. 2024;41:721–30. doi: 10.1007/s11095-024-03685-3. [DOI] [PubMed] [Google Scholar]
72.Grzybowski A, Jin K, Wu H. Challenges of artificial intelligence in medicine and dermatology. Clin Dermatol. 2024;42:210–5. doi: 10.1016/j.clindermatol.2023.12.013. [DOI] [PubMed] [Google Scholar]
73.Freyer N, Groß D, Lipprandt M. The ethical requirement of explainability for AI-DSS in healthcare: a systematic review of reasons. BMC Med Ethics. 2024;25:104. doi: 10.1186/s12910-024-01103-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
74.Shick AA, Webber CM, Kiarashi N, Weinberg JP, Deoras A, Petrick N, et al. Transparency of artificial intelligence/machine learning-enabled medical devices. NPJ Digit Med. 2024;7:21. doi: 10.1038/s41746-023-00992-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
75.Wu J, Wang H, Ni C, Zhang C, Lu W. Case study of Next-generation artificial intelligence in medical image diagnosis based on cloud computing. JTPES. 2024;4:66–73. [Google Scholar]
76.Haggenmüller S, Schmitt M, Krieghoff-Henning E, Hekler A, Maron RC, Wies C, et al. Federated learning for decentralized artificial intelligence in melanoma diagnostics. JAMA Dermatol. 2024;160:303–11. doi: 10.1001/jamadermatol.2023.5550. [DOI] [PMC free article] [PubMed] [Google Scholar]
77.Mikhail D, Farah A, Milad J, Nassrallah W, Mihalache A, Milad D, et al. Performance of DeepSeek-R1 in ophthalmology: an evaluation of clinical decision-making and cost-effectiveness. medRxiv. 2025:2025–02. doi: 10.1136/bjo-2025-327360. [DOI] [PubMed] [Google Scholar]
78.Peng Y, Malin BA, Rousseau JF, Wang Y, Xu Z, Xu X, et al. From GPT to DeepSeek: significant gaps remains in realizing AI in healthcare. United States: Journal of Biomedical Informatics; 2025. p. 104791. [DOI] [PMC free article] [PubMed] [Google Scholar]
79.Mondillo G, Colosimo S, Perrotta A, Frattolillo V, Masino M. Comparative evaluation of advanced AI reasoning models in paediatric clinical decision support: ChatGPT O1 vs. DeepSeek-R1. medRxiv. 2025 doi: 10.1101/2025.01.27.25321169. [DOI] [Google Scholar]
80.Kayaalp ME, Prill R, Sezgin EA, Cong T, Królikowska A, Hirschmann MT. . Hoboken, United States: Wiley Online Library; 2023. DeepSeek versus ChatGPT: multimodal artificial intelligence revolutionizing scientific discovery. From language editing to autonomous content generation—redefining innovation in research and practice. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material

j_mr-2024-0097_suppl_001.csv^{(477.1KB, csv)}

[j_mr-2024-0097_ref_001] 1.Alhur A, Alhur AA, Hamdan Alghamdi SJ, Asiri ZI, Saeed Alzamil SK, Alfateih S, et al. Advancing the frontiers of artificial intelligence in transforming healthcare: a comprehensive literature review. J Popul Ther Clin Pharmacol. 2023;30:2107–12. doi: 10.53555/jptcp.v30i19.4744. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_002] 2.Zeb S, Nizamullah F, Abbasi N, Fahad M. AI in healthcare: revolutionizing diagnosis and therapy. Int J Multidiscip Sci Arts. 2024;3:118–28. [Google Scholar]

[j_mr-2024-0097_ref_003] 3.Nazi ZA, Peng W. Large language models in healthcare and medical domain: a review. Informatics. 2024;11:57. doi: 10.3390/informatics11030057. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_004] 4.Dias R, Torkamani A. Artificial intelligence in clinical and genomic diagnostics. Genome Med. 2019;11:70. doi: 10.1186/s13073-019-0689-8. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_005] 5.Poalelungi DG, Musat CL, Fulga A, Neagu M, Neagu A-I, Piraianu AI, et al. Advancing patient care: how artificial intelligence is transforming healthcare. J Personalized Med. 2023;13 doi: 10.3390/jpm13081214. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_006] 6.Chugh V, Basu A, Kaushik A, Bhansali S, Basu AK. Employing nano-enabled artificial intelligence (AI)-based smart technologies for prediction, screening, and detection of cancer. Nanoscale. 2024;16:5458–86. doi: 10.1039/d3nr05648a. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_007] 7.Pang K, Song Z, Liu Y, Sun H, Zhang R, Fu Y, et al. In vivo dynamic monitoring of circulating melanoma cells and the inhibitory effect of PD-L1 inhibitor based on PAFC equipped with a deep learning framework. APL Photon. 2024;9 doi: 10.1063/5.0226328. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_008] 8.Edelson DP, Churpek MM, Carey KA, Lin Z, Huang C, Siner JM, et al. Early warning scores with and without artificial intelligence. JAMA Netw Open. 2024;7:e2438986. doi: 10.1001/jamanetworkopen.2024.38986. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_009] 9.Berge GT, Granmo OC, Tveit TO, Munkvold BE, Ruthjersen AL, Sharma J. Machine learning-driven clinical decision support system for concept-based searching: a field trial in a Norwegian hospital. BMC Med Inf Decis Making. 2023;23:5. doi: 10.1186/s12911-023-02101-x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_010] 10.Bozyel S, Şimşek E, Koçyiğit D, Güler A, Korkmaz Y, Şeker M, et al. Artificial intelligence-based clinical decision support systems in cardiovascular diseases. Anatol J Cardiol. 2024;28:74. doi: 10.14744/anatoljcardiol.2023.3685. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_011] 11.Talati D. AI in healthcare domain. J Knowl Learn Sci Technol. 2023 ISSN: 2959-6386 (online) [Google Scholar]

[j_mr-2024-0097_ref_012] 12.Kaul V, Enslin S, Gross SA. History of artificial intelligence in medicine. Gastrointest Endosc. 2020;92:807–12. doi: 10.1016/j.gie.2020.06.040. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_013] 13.Moawad AW, Fuentes DT, ElBanan MG, Shalaby AS, Guccione J, Kamel S, et al. Artificial intelligence in diagnostic radiology: where do we stand, challenges, and opportunities. J Comput Assist Tomogr. 2022;46:78–90. doi: 10.1097/rct.0000000000001247. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_014] 14.Nozari H, Ghahremani-Nahr J, Szmelter-Jarosz A. AI and machine learning for real-world problems. Adv Comput. 2024;134:1–12. doi: 10.1016/bs.adcom.2023.02.001. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_015] 15.Senkamalavalli R, Sankar S, Parivazhagan A, Raja R, Yoganand S, Srinivas P, et al. Enhancing clinical decision-making with cloud-enabled integration of image-driven insights. Indones J Electr Eng Comput Sci. 2024;36:338–46. doi: 10.11591/ijeecs.v36.i1.pp338-346. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_016] 16.Moulaei K, Yadegari A, Baharestani M, Farzanbakhsh S, Sabet B, Afrash MR. Generative artificial intelligence in healthcare: a scoping review on benefits, challenges and applications. Int J Med Inf. 2024:105474. doi: 10.1016/j.ijmedinf.2024.105474. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_017] 17.Chen X. AI in healthcare: revolutionizing diagnosis and treatment through machine learning. MZ J Artif Intell. 2024;1 [Google Scholar]

[j_mr-2024-0097_ref_018] 18.Edge D, Trinh H, Cheng N, Bradley J, Chao A, Mody A, et al. From local to global: a graph RAG approach to query-focused summarization. ArXiv. 2024 abs/2404.16130. [Google Scholar]

[j_mr-2024-0097_ref_019] 19.Swanson K, Wu E, Zhang A, Alizadeh AA, Zou J. From patterns to patients: advances in clinical machine learning for cancer diagnosis, prognosis, and treatment. Cell. 2023;186:1772–91. doi: 10.1016/j.cell.2023.01.035. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_020] 20.Liu Y, Liu G, Zhang Q. Deep learning and medical diagnosis. Lancet. 2019;394:1709–10. doi: 10.1016/s0140-6736(19)32501-2. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_021] 21.Chen X, Wang X, Zhang K, Fung KM, Thai TC, Moore K, et al. Recent advances and clinical applications of deep learning in medical image analysis. Med Image Anal. 2022;79:102444. doi: 10.1016/j.media.2022.102444. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_022] 22.Wang X, Zhao J, Marostica E, Yuan W, Jin J, Zhang J, et al. A pathology foundation model for cancer diagnosis and prognosis prediction. Nature. 2024;634:970–8. doi: 10.1038/s41586-024-07894-z. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_023] 23.Tu T, Palepu A, Schaekermann M, Saab K, Freyberg J, Tanno R, et al. Towards conversational diagnostic AI. . 2024 doi: 10.1038/s41586-025-08866-7. arXiv preprint arXiv:240105654. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_024] 24.Xie X, Yu W, Wang L, Yang J, Tu X, Liu X, et al. SERS-based AI diagnosis of lung and gastric cancer via exhaled breath. Spectrochim Acta Mol Biomol Spectrosc. 2024;314:124181. doi: 10.1016/j.saa.2024.124181. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_025] 25.Wani NA, Kumar R, Bedi J. DeepXplainer: an interpretable deep learning based approach for lung cancer detection using explainable artificial intelligence. Comput Methods Progr Biomed. 2024;243:107879. doi: 10.1016/j.cmpb.2023.107879. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_026] 26.Kale MB, Wankhede NL, Pawar RS, Ballal S, Kumawat R, Goswami M, et al. AI-driven innovations in Alzheimer’s disease: integrating early diagnosis, personalized treatment, and prognostic modelling. Ageing Res Rev. 2024:102497. doi: 10.1016/j.arr.2024.102497. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_027] 27.Mikhael PG, Wohlwend J, Yala A, Karstens L, Xiang J, Takigami AK, et al. Sybil: a validated deep learning model to predict future lung cancer risk from a single low-dose chest computed tomography. J Clin Oncol. 2023;41:2191–200. doi: 10.1200/jco.22.01345. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_028] 28.Vedula RS, Karp HQ, Koob J, Lim F, Garcia JS, Winer ES, et al. CRISPR-based rapid molecular diagnostic tests for fusion-driven leukemias. Blood. 2024;144:1290–9. doi: 10.1182/blood.2023022908. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_029] 29.Placido D, Yuan B, Hjaltelin JX, Zheng C, Haue AD, Chmura PJ, et al. A deep learning algorithm to predict risk of pancreatic cancer from disease trajectories. Nat Med. 2023;29:1113–22. doi: 10.1038/s41591-023-02332-5. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_030] 30.Jack CR, Andrews JS, Beach TG, Buracchio T, Dunn B, Graf A, et al. Revised criteria for diagnosis and staging of Alzheimer’s disease: Alzheimer’s Association Workgroup. Alzheimer’s Dement. 2024;20:5143–69. doi: 10.1002/alz.13859. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_031] 31.Zolnoori M, Zolnour A, Topaz M. ADscreen: a speech processing-based screening system for automatic identification of patients with Alzheimer’s disease and related dementia. Artif Intell Med. 2023;143:102624. doi: 10.1016/j.artmed.2023.102624. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_032] 32.O’Malley RPD, Mirheidari B, Harkness K, Reuber M, Venneri A, Walker T, et al. Fully automated cognitive screening tool based on assessment of speech and language. J Neurol Neurosurg Psychiatry. 2020;92:12–5. doi: 10.1136/jnnp-2019-322517. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_033] 33.Leming M, Das S, Im H. Adversarial confound regression and uncertainty measurements to classify heterogeneous clinical MRI in Mass General Brigham. PLoS One. 2023;18:e0277572. doi: 10.1371/journal.pone.0277572. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_034] 34.Amini S, Hao B, Yang J, Karjadi C, Kolachalama VB, Au R, et al. Prediction of Alzheimer’s disease progression within 6 years using speech: a novel approach leveraging language models. Alzheimer’s Dement. 2024;20:5262–70. doi: 10.1002/alz.13886. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_035] 35.Mohsen F, Al-Absi HRH, Yousri NA, El Hajj N, Shah Z. A scoping review of artificial intelligence-based methods for diabetes risk prediction. NPJ Digit Med. 2023;6:197. doi: 10.1038/s41746-023-00933-5. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_036] 36.Huang J, Yeung AM, Armstrong DG, Battarbee AN, Cuadros J, Espinoza JC, et al. Artificial intelligence for predicting and diagnosing complications of diabetes. J Diabetes Sci Technol. 2023;17:224–38. doi: 10.1177/19322968221124583. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_037] 37.Gudiño-Ochoa A, García-Rodríguez JA, Ochoa-Ornelas R, Cuevas-Chávez JI, Sánchez-Arias DA. Noninvasive diabetes detection through human breath using TinyML-Powered E-Nose. Sensors. 2024;24:1294. doi: 10.3390/s24041294. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_038] 38.Eben H, Bolanle M. Deep learning for diabetes diagnosis: convolutional neural networks and recurrent neural networks. MZ J Artif Intell. 2024;1 [Google Scholar]

[j_mr-2024-0097_ref_039] 39.Fu H, Tian Y, Zha G, Xiao X, Zhu H, Zhang Q, et al. Microstrip isoelectric focusing with deep learning for simultaneous screening of diabetes, anemia, and thalassemia. Anal Chim Acta. 2024;1312:342696. doi: 10.1016/j.aca.2024.342696. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_040] 40.Iftikhar M, Saqib M, Qayyum SN, Asmat R, Mumtaz H, Rehan M, et al. Artificial intelligence-driven transformations in diabetes care: a comprehensive literature review. Ann Med Surg (Lond). 2024;86:5334–42. doi: 10.1097/ms9.0000000000002369. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_041] 41.Elias P, Jain SS, Poterucha T, Randazzo M, Lopez Jimenez F, Khera R, et al. Artificial intelligence for cardiovascular care—part 1: advances: JACC review topic of the week. J Am Coll Cardiol. 2024;83:2472–86. doi: 10.1016/j.jacc.2024.03.400. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_042] 42.Ogawa S, Namino F, Mori T, Sato G, Yamakawa T, Saito S. AI diagnosis of heart sounds differentiated with super StethoScope. J Cardiol. 2024;83:265–71. doi: 10.1016/j.jjcc.2023.09.007. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_043] 43.Omarov B, Saparkhojayev N, Shekerbekova S, Akhmetova O, Sakypbekova M, Kamalova G, et al. Artificial intelligence in medicine: real time electronic stethoscope for heart diseases detection. Comput Mater Continua (CMC) 2022;70 doi: 10.32604/cmc.2022.019246. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_044] 44.Thoenes M, Agarwal A, Grundmann D, Ferrero C, McDonald A, Bramlage P, et al. Narrative review of the role of artificial intelligence to improve aortic valve disease management. J Thorac Dis. 2021;13:396. doi: 10.21037/jtd-20-1837. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_045] 45.Baikuvekov M, Baikuvekova A, Sultan D. Development of a digital stethoscopy system for detection of heart diseases; 2024 IEEE 4th International Conference on Smart Information Systems and Technologies (SIST). IEEE; 2024. [Google Scholar]

[j_mr-2024-0097_ref_046] 46.Khanna NN, Maindarkar MA, Viswanathan V, Puvvula A, Paul S, Bhagawati M, et al. Cardiovascular/stroke risk stratification in diabetic foot infection patients using deep learning-based artificial intelligence: an investigative study. J Clin Med. 2022;11:6844. doi: 10.3390/jcm11226844. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_047] 47.Tatsugami F, Nakaura T, Yanagawa M, Fujita S, Kamagata K, Ito R, et al. Recent advances in artificial intelligence for cardiac CT: enhancing diagnosis and prognosis prediction. Diagn Interv Imaging. 2023;104:521–8. doi: 10.1016/j.diii.2023.06.011. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_048] 48.Gutfleisch M, Ester O, Aydin S, Quassowski M, Spital G, Lommatzsch A, et al. Clinically applicable deep learning-based decision aids for treatment of neovascular AMD. Graefes Arch Clin Exp Ophthalmol. 2022;260:2217–30. doi: 10.1007/s00417-022-05565-1. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_049] 49.Zhou Y, Chia MA, Wagner SK, Ayhan MS, Williamson DJ, Struyven RR, et al. A foundation model for generalizable disease detection from retinal images. Nature. 2023;622:156–63. doi: 10.1038/s41586-023-06555-x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_050] 50.Qian B, Sheng B, Chen H, Wang X, Li T, Jin Y, et al. A competition for the diagnosis of myopic maculopathy by artificial intelligence algorithms. JAMA Ophthalmol. 2024 doi: 10.1001/jamaophthalmol.2024.3707. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_051] 51.Dkhan R, Hamdan R. Respiratory diseases detection and classification based on respiratory voice using artificial intelligence methods. SVUPedia. 2024. pp. 36–45.

[j_mr-2024-0097_ref_052] 52.Ghrabli S, Elgendi M, Menon C. Identifying unique spectral fingerprints in cough sounds for diagnosing respiratory ailments. Sci Rep. 2024;14:593. doi: 10.1038/s41598-023-50371-2. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_053] 53.Isangula KG, Haule RJ. Leveraging AI and machine learning to develop and evaluate a contextualized user-friendly cough audio classifier for detecting respiratory diseases: protocol for a diagnostic study in rural Tanzania. JMIR Res Protoc. 2024;13:e54388. doi: 10.2196/54388. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_054] 54.Saeed T, Ijaz A, Sadiq I, Qureshi HN, Rizwan A, Imran A. An AI-enabled bias-free respiratory disease diagnosis model using cough audio. Bioengineering. 2024;11:55. doi: 10.3390/bioengineering11010055. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_055] 55.Cooper M. AI and spotting the sound of illness. ITNOW. 2024;66:24–5. doi: 10.1093/itnow/bwae012. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_056] 56.Malik H, Anees T. Multi-modal deep learning methods for classification of chest diseases using different medical imaging and cough sounds. PLoS One. 2024;19:e0296352. doi: 10.1371/journal.pone.0296352. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_057] 57.Eswara Rao G, Rajitha B. HQF-CC: hybrid framework for automated respiratory disease detection based on quantum feature extractor and custom classifier model using chest X-rays. Int J Inf Technol. 2024;16:1145–53. doi: 10.1007/s41870-023-01681-1. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_058] 58.Abou Ghayda R, Cannarella R, Calogero AE, Shah R, Rambhatla A, Zohdy W, et al. Artificial intelligence in andrology: from semen analysis to image diagnostics. World J Mens Health. 2024;42:39. doi: 10.5534/wjmh.230050. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_059] 59.Suresh K, Wu MP, Benboujja F, Christakis B, Newton A, Hartnick CJ, et al. AI model versus clinician otoscopy in the operative setting for otitis media diagnosis. Otolaryngol Head Neck Surg. 2024;170:1598–601. doi: 10.1002/ohn.559. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_060] 60.Dang B, Ma D, Li S, Qi Z, Zhu EY. Deep learning-based snore sound analysis for the detection of night-time breathing disorders. Appl Comput Eng. 2024;76:109–14. [Google Scholar]

[j_mr-2024-0097_ref_061] 61.Peng D, Yue H, Tan W, Lei W, Chen G, Shi W, et al. A bimodal feature fusion convolutional neural network for detecting obstructive sleep apnea/hypopnea from nasal airflow and oximetry signals. Artif Intell Med. 2024;150:102808. doi: 10.1016/j.artmed.2024.102808. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_062] 62.Butler L, Gunturkun F, Chinthala L, Karabayir I, Tootooni MS, Bakir-Batu B, et al. AI-based preeclampsia detection and prediction with electrocardiogram data. Front Cardiovasc Med. 2024;11:1360238. doi: 10.3389/fcvm.2024.1360238. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_063] 63.Sadasivan VS, Kumar A, Balasubramanian S, Wang W, Feizi S. Can AI-generated text be reliably detected? . 2023 arXiv preprint arXiv:230311156. [Google Scholar]

[j_mr-2024-0097_ref_064] 64.Tsopra R, Fernandez X, Luchinat C, Alberghina L, Lehrach H, Vanoni M, et al. A framework for validating AI in precision medicine: considerations from the European ITFoC consortium. BMC Med Inf Decis Making. 2021;21:274. doi: 10.1186/s12911-021-01634-3. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_065] 65.Sandhu S, Lin AL, Brajer N, Sperling J, Ratliff W, Bedoya AD, et al. Integrating a machine learning system into clinical workflows: qualitative study. J Med Internet Res. 2020;22:e22421. doi: 10.2196/22421. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_066] 66.Wang F, Beecy A. Implementing AI models in clinical workflows: a roadmap. BMJ Evid-Based Med. 2024 doi: 10.1136/bmjebm-2023-112727. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_067] 67.Phillips RA, Jani J, Bradley SK. Exploring the literature on artificial intelligence use in oncology. J Clin Oncol. 2024;42:e13642–e. doi: 10.1200/jco.2024.42.16_suppl.e13642. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_068] 68.Varghese J. Artificial intelligence in medicine: chances and challenges for wide clinical adoption. Visc Med. 2020;36:443–9. doi: 10.1159/000511930. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_069] 69.Jin Q, Chen F, Zhou Y, Xu Z, Cheung JM, Chen R, et al. Hidden flaws behind expert-level accuracy of multimodal GPT-4 vision in medicine. npj Digit Med. 2024;7:190. doi: 10.1038/s41746-024-01185-7. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_070] 70.Blumenthal D, Patel B. The regulation of clinical artificial intelligence. NEJM AI. 2024;1:AIpc2400545. doi: 10.1056/aipc2400545. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_071] 71.Romagnoli A, Ferrara F, Langella R, Zovi A. Healthcare systems and artificial intelligence: focus on challenges and the international regulatory framework. Pharm Res. 2024;41:721–30. doi: 10.1007/s11095-024-03685-3. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_072] 72.Grzybowski A, Jin K, Wu H. Challenges of artificial intelligence in medicine and dermatology. Clin Dermatol. 2024;42:210–5. doi: 10.1016/j.clindermatol.2023.12.013. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_073] 73.Freyer N, Groß D, Lipprandt M. The ethical requirement of explainability for AI-DSS in healthcare: a systematic review of reasons. BMC Med Ethics. 2024;25:104. doi: 10.1186/s12910-024-01103-2. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_074] 74.Shick AA, Webber CM, Kiarashi N, Weinberg JP, Deoras A, Petrick N, et al. Transparency of artificial intelligence/machine learning-enabled medical devices. NPJ Digit Med. 2024;7:21. doi: 10.1038/s41746-023-00992-8. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_075] 75.Wu J, Wang H, Ni C, Zhang C, Lu W. Case study of Next-generation artificial intelligence in medical image diagnosis based on cloud computing. JTPES. 2024;4:66–73. [Google Scholar]

[j_mr-2024-0097_ref_076] 76.Haggenmüller S, Schmitt M, Krieghoff-Henning E, Hekler A, Maron RC, Wies C, et al. Federated learning for decentralized artificial intelligence in melanoma diagnostics. JAMA Dermatol. 2024;160:303–11. doi: 10.1001/jamadermatol.2023.5550. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_077] 77.Mikhail D, Farah A, Milad J, Nassrallah W, Mihalache A, Milad D, et al. Performance of DeepSeek-R1 in ophthalmology: an evaluation of clinical decision-making and cost-effectiveness. medRxiv. 2025:2025–02. doi: 10.1136/bjo-2025-327360. [DOI] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_078] 78.Peng Y, Malin BA, Rousseau JF, Wang Y, Xu Z, Xu X, et al. From GPT to DeepSeek: significant gaps remains in realizing AI in healthcare. United States: Journal of Biomedical Informatics; 2025. p. 104791. [DOI] [PMC free article] [PubMed] [Google Scholar]

[j_mr-2024-0097_ref_079] 79.Mondillo G, Colosimo S, Perrotta A, Frattolillo V, Masino M. Comparative evaluation of advanced AI reasoning models in paediatric clinical decision support: ChatGPT O1 vs. DeepSeek-R1. medRxiv. 2025 doi: 10.1101/2025.01.27.25321169. [DOI] [Google Scholar]

[j_mr-2024-0097_ref_080] 80.Kayaalp ME, Prill R, Sezgin EA, Cong T, Królikowska A, Hirschmann MT. . Hoboken, United States: Wiley Online Library; 2023. DeepSeek versus ChatGPT: multimodal artificial intelligence revolutionizing scientific discovery. From language editing to autonomous content generation—redefining innovation in research and practice. [DOI] [PubMed] [Google Scholar]

PERMALINK

Artificial intelligence-driven transformative applications in disease diagnosis technology

Junyu Zhou

Sunmin Park

Sihan Dong

Xiaoying Tang

Xunbin Wei

Abstract

Introduction

Bibliometric analysis of AI applications in medical diagnostics

Figure 1:

AI-powered non-invasive cancer diagnostics and specialized applications

Figure 2:

A paradigm shift in Alzheimer’s patient-friendly diagnostics

Figure 3:

AI integrates multi-modal data for comprehensive diabetes diagnostics

Figure 4:

Comprehensive approaches in imaging precision and early diagnosis to heart disease

Figure 5:

AI’s transformative role in making ophthalmology more accessible and efficient

Figure 6:

Advanced sound analysis and machine learning enhancing respiratory disease diagnosis

Figure 7:

The impact of AI across medical diagnosis in other disease

Figure 8:

The future is now

Technical and ethical considerations

Current limitations and regulatory framework of diagnostic AI

A new era for diagnostic AI

Figure 9:

Contributions and limitations

Supplementary Material

Acknowledgments

Supplementary Material

Footnotes

References

Associated Data

Supplementary Materials

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases