-AI-assisted diagnostic potential of CT in bone oncology and its impact on clinical decision-making for intensive care

Highlights

• •AI-assisted diagnostic potential of CT for bone cancer and its impact on patient care.
• •Conducted retrospective analysis of 50 patients with bone cancer by SPECT and histopathology.
• •Timely CT examination is crucial in achieving accurate staging of bone tumors.
• •Specific consolidation patterns and extent of lesion spread were predictive of risks necessitating ICU intervention.
• •CT severity scores proved invaluable in forecasting the need for therapeutic interventions.

Abstract

Objective

This study evaluates the AI-assisted diagnostic potential of computed tomography (CT) for bone cancer and its influence on patient care during the pre- and post-treatment phases. It compares patient management approaches based on CT severity levels and identifies distinct CT phenotypes linked to disease severity.

Methodology

We retrospectively examined 50 patients diagnosed with bone cancer between December 2022 and June 2023. The CT scans were analyzed according to the Radiological Society of North America (RSNA) guidelines. This study was performed using the deep convolutional neutral network (DCNN) model to assist doctors in diagnosing bone tumors through CT scanning. Patients’ management approaches were compared based on the severity levels indicated by CT scans.

Results

Fifty patients participated in this study, with a median age of 67.2 years, ranging from 32 to 89 years. Of them, 38 % were female and 62 % were male. In 2022, 19 individuals (13 males and 6 females, ages 32 to 84) were assessed, with a mean age of 59.9 years. In 2023, 31 individuals, aged 54 to 89 with a mean age of 71.6 years, were assessed; among them were 18 men and 13 women. SPECT scans revealed the following key diagnostic features: 85.9 % of patients exhibited bone lesions with ground-glass opacities, 88 % had multipolar involvement, 92.8 % had bilateral involvement, and 92.8 % showed peripheral involvement. The severity scores based on CT scans were significantly higher in patients requiring intensive care, with scores above 14 being more common in this group.

Conclusion

Distinct CT findings during the AI-assisted diagnosis and treatment of bone cancer provided prompt and sensitive examination capabilities. Notably, two CT phenotypes emerged, associated with large consolidation patterns and high severity scores, offering crucial insights into disease severity and aiding in clinical decision-making for intensive care requirements. The study underscores the importance of CT in the effective monitoring and management of bone cancer pre- and post-treatment.

1. Introduction

The effective management and treatment of bone cancer necessitate sophisticated imaging techniques. Bone cancer includes various malignant conditions such as osteosarcoma, chondrosarcoma, and Ewing sarcoma, which often present with complex clinical manifestations. Imaging studies play a crucial role in the diagnosis of bone cancer. While X-rays are often the first imaging modality used, they may not always provide enough detail for a definitive diagnosis. More advanced imaging techniques such as computed tomography (CT) and magnetic resonance imaging (MRI) offer higher resolution images, allowing for a more accurate assessment of the tumor's location, size, and involvement with surrounding tissues. CT imaging is a key instrument essential in diagnosing bone cancer, enabling precise depiction of bone lesions, tumor size, location, and the extent of disease spread. CT scans facilitate biopsies, assist in the staging process, and evaluate treatment responses, thereby providing crucial insights for clinical decision-making [1], [2], [3]. The unique capabilities of CT imaging in detecting parenchymal changes and complications such as pathological fractures underscore its importance in bone cancer management. This study explores the AI-assisted diagnostic potential of CT in bone cancer and its impact on patient care, aiming to enhance the understanding of CT imaging in both pre- and post-treatment phases [4], [5].

The American Cancer Society (ACS) and the World Health Organization (WHO) have consistently underscored the paramount importance of early detection and effective treatment of bone cancers. In recent years, remarkable advancements in imaging technology have played a pivotal role in enhancing patient outcomes. Bone cancers, comprising malignancies such as osteosarcoma, chondrosarcoma, and Ewing sarcoma, present substantial challenges owing to their aggressive nature and intricate treatment requirements [6]. Notably, as of June 2020, the ACS reported an estimated 3,500 annual new cases of bone and joint cancer solely in the United States, significantly contributing to morbidity and mortality rates [6], [2]. The ACS has highlighted the role of advanced diagnostic techniques, including CT imaging, in providing detailed insights into tumor characteristics, which are crucial for staging and treatment planning. These efforts underscore the necessity for continued advancements in diagnostic imaging to enhance the early detection and management of bone cancers, ultimately improving patient prognosis and quality of life.

Advanced imaging techniques such as computed tomography (CT) scans are widely considered critical in the diagnosis and monitoring of bone cancer due to their high specificity and ability to provide detailed anatomical information. Nevertheless, CT scans are not deviod of limitations, including varied sensitivity and specificity rates, as well as high costs [7]. Studies show that CT scans, along with other imaging modalities like magnetic resonance imaging (MRI) and bone scintigraphy, are essential in identifying bone tumors, assessing their size and location, and evaluating the extent of metastasis [7]. Additionally, the potential variation in CT scan sensitivity and specificity for bone cancer underscores the imperative for embracing comprehensive imaging approaches to improve diagnostic accuracy [8], [9].

In the realm of bone oncology, CT scans play a pivotal role in classifying various bone cancer types, assessing disease severity, tracking progression, and evaluating treatment outcomes [10], [11], [12]. These imaging techniques help in monitoring the effectiveness of treatments such as chemotherapy, radiotherapy, and surgical interventions, thereby aiding in treatment planning and follow-up [13], [14]. The aim of this work is to evaluate the diagnostic validity of CT scans in detecting bone cancer by contrasting their accuracy with other imaging methods and following established guidelines for CT interpretation in oncology. Research endeavors focusing on hemodynamic parameters and their influence on flow resistance, myocardial ischemia, and plaque vulnerability has provided valuable insights that can be paralleled in the study of tumor vascularization and perfusion in oncology [9]. Specifically, examining the effects of fluid and structural characteristics on tumor growth and response to treatment can offer critical information for clinical studies and treatment planning. Advances in imaging techniques and their application in bone cancer diagnosis and monitoring are crucial for improving patient outcomes and guiding future research in the field.

The current study endeavors to examine the correlation between CT severity ratings and the probability of requiring intensive treatment interventions. Furthermore, the research aims to identify imaging characteristics such as unique phenotypes and severity indicators that are associated with advanced stages of bone cancer. Through these investigations, the study hopes to advance knowledge on the role of CT imaging in bone cancer diagnosis and treatment. This could help guide clinical decision-making and improve patient outcomes by providing a more precise assessment of tumor severity and progression.

Fig. 1 illustrates the comprehensive role of Computed Tomography (CT) in the diagnosis and management of bone cancer. On the diagnosis side, initial CT scans are utilized to detect the presence of tumors, providing detailed imaging that guides subsequent biopsies for confirmation. Following diagnosis, CT scans play a crucial role in monitoring the effectiveness of treatment. Pre-treatment CT imaging assesses the extent of the disease, helping to plan surgical interventions. Throughout the treatment phases, including surgery, chemotherapy, and radiation therapy, CT scans are employed to track progress, ensuring the treatment is effective and on the right course. Post-treatment, CT scans continue to be instrumental in adjusting therapeutic strategies based on the patient’s response, ensuring optimized and personalized patient care.

Fig. 1.

Computed Tomography (CT) for Bone Cancer Diagnosis and Monitoring.

2. Methodology

This retrospective investigation was conducted at Northern Jiangsu People's Hospital, a single tertiary facility, focusing on 60 patients diagnosed with bone cancer between December 2022 and June 2023. The study aimed to assess the diagnostic accuracy and effectiveness of CT imaging in evaluating bone cancer, monitoring disease progression, and guiding treatment strategies. Each patient underwent a high-resolution CT scan of the affected bones using our advanced CT machine, which operated on low kVp and low mAs settings to minimize radiation exposure. The specific radiation dose varies depending on the scanning site. This protocol ensured detailed imaging while prioritizing patient safety by avoiding the need for intravenous contrast agents. All CT scans were performed using the same machine and protocol. The CT scans were meticulously analyzed for various diagnostic indicators, including tumor size, location, bone involvement, and the presence of metastases. These parameters were then correlated with clinical outcomes to evaluate the prognostic value of CT imaging in bone cancer management. The study also sought to identify specific CT phenotypes and severity markers that could predict treatment response and guide clinical decision-making.

The institutional ethical committee approved the study protocol, ensuring that all research activities adhered to ethical standards. As the study involved the use of existing medical imaging data and posed minimal risk to participants, the requirement for written informed consent was waived. This ethical consideration facilitated a comprehensive evaluation of CT imaging's role in bone oncology, ultimately aiming to enhance diagnostic accuracy, improve patient outcomes, and contribute to the body of knowledge in this field.

Fig. 2 illustrates the CT imaging outcomes for a cohort of 60 patients suspected of having primary bone cancer within a six-month period from December 2022 to June 2023. Out of these, 50 individuals underwent CT imaging to determine the likelihood of having primary bone cancer. The distribution of the patients by gender showed that 62 % were male and 38 % were female, as confirmed by histopathological examination. For the male patients, CT imaging revealed that 8 were CT positive, 5 were classified with a low probability of having cancer, 7 were classified with an intermediate probability, and 11 were classified with a high probability of having cancer. In contrast, among the female patients, 6 were CT positive, 3 were categorized with a low probability, 4 with an intermediate probability, and 6 with a high probability of having bone cancer. Additionally, there were 10 patients who tested negative on the CT imaging. This data underscores the role of CT imaging in the diagnosis and probability assessment of primary bone cancer across different patient demographics.

Fig. 2.

CT imaging results for 60 patients with suspected primary bone cancer, detailing gender distribution and probability assessments.

2.1. Requirements for Inclusion

Each patient presenting to the ER or outpatient clinic with symptoms suggestive of bone cancer—such as persistent bone pain, swelling, fractures, or unexplained weight loss—was subjected to a standardized diagnostic procedure. This process involved comprehensive laboratory examinations, including biopsy and blood tests for tumor markers, in addition to high-resolution non-contrast CT scans of the affected bones. These diagnostic procedures were crucial in promptly and accurately identifying potential cases of bone cancer, facilitating timely intervention and appropriate care for the affected individuals. This approach ensured that each patient received a thorough evaluation, enabling the early detection of malignancies and the initiation of tailored treatment plans.

2.2. Requirements for Exclusion

Patients who were unable to undergo a biopsy or who lacked adequate clinical information to form a comprehensive management plan were excluded from the study. Additionally, individuals admitted to the hospital for reasons unrelated to bone cancer, as well as those whose imaging results were compromised by motion artifacts or other technical issues, were not considered. This ensured that the study remained focused on patients with definitive and evaluable bone cancer diagnoses, providing accurate and relevant data for analysis.

2.3. CT approaches for bone cancer diagnosis and monitoring

The procedure was performed using an 80-slice CT scanner from Prime Aquilion, Toshiba, USA. The hospital's radiology department adhered to its standard protocols for patient safety and image quality. Patients did not need any special preparations for the scanning procedure. They were positioned supine with their arms raised over their heads to minimize artifacts. The imaging was performed with a slice thickness of 1.25 mm and an interval of 0.625 mm using a 512 × 512 matrix. The rotation time was set to 0.5 s, and the tube speed was calibrated to 35 mm/revolution. The CT scan operator adjusted the kVp and mAs settings to the lowest possible values to reduce radiation exposure. Images were then transferred to the workstation for multi-planar reformation and examination of the axial slices. Two experienced radiologists with at least three years of post-fellowship experience in musculoskeletal imaging analyzed the images. Disagreements among the observers' interpretations were resolved by consensus to minimize inter-observer variability; however, this was outside the scope of the current study. The radiologists were blinded to the study's objectives and the clinical details of the patients. The imaging results recorded included. We collected images annotated by radiologists and pre-trained the training model using the most commonly used ImageNet database in computer vision. By training the regression and classification tasks in the ImageNet database, we obtained the parameters of the deep convolutional neutral network (DCNN) model that performed well on the task. Then, we used the pre-trained model to train, fine tune, and optimize the parameters in the image training set and validation set. Subsequently, we used the DCNN model to assist doctors in diagnosing for all patients.

In Fig. 3, the leftmost part of the image represents the input image, and the area in the middle represents the reference network ResNet18, the rightmost side represents the prediction module, which consists of three branches, including the heatmap of the center point branch, center offset branch, target size branch.

Fig. 3.

The process of DCNN model to assist doctors in diagnosing.

2.3.1. Identification and characterization of bone lesions

The use of CT imaging is essential for evaluating bone cancer. Lesions appear as areas with abnormal attenuation, indicating the presence of a tumor. The presence or absence of bone lesions, their location (e.g., femur, pelvis, spine), size, shape (e.g., lytic, sclerotic, mixed), and degree of involvement were accurately recorded. Clinicians also considered the presence or absence of secondary signs that could assist in differential diagnosis and assessing disease severity. These included periosteal reaction, cortical destruction, soft tissue mass, and any signs of pathological fracture [8]. By incorporating these features into the CT scan evaluation, the extent of bone involvement was fully depicted, and bone cancer could be differentiated from other musculoskeletal diseases. Moreover, it was critical to identify the primary CT pattern of the lesions whether predominantly lytic or sclerotic as this distinction is important from a therapeutic standpoint [15], [16]. Lytic lesions, also known as osteolytic lesions, refer to areas of bone destruction where the normal bone matrix is replaced by fluid, fibrous tissue, or tumor. Sclerosis patterns, on the other hand, refer to areas of bone thickening or increased bone density. It provides information about the progression, severity, and potential therapeutic approaches for the condition. Systematically recording these radiological findings ensured a more accurate diagnosis, prognosis, and management of bone cancer, ultimately improving patient outcomes and guiding treatment strategies.

Fig. 4 presents detailed CT scans of the human skeletal system, highlighting crucial aspects pertinent to the diagnosis and study of bone cancer within the field of bone oncology. The central images display full-body scans in - frontal - views, providing a clear representation of the skeletal structure from head to toe. Surrounding these are various views of the skeletal system relevant to bone cancer diagnosis and bone oncology. These scans are instrumental in identifying bone lesions, tumors, and other abnormalities associated with primary bone cancer. The meticulous depiction of the bone structures facilitates a comprehensive analysis, enabling oncologists to assess the extent of cancer spread, plan biopsies, and strategize appropriate treatment plans. This holistic view underscores the significance of CT imaging in enhancing diagnostic accuracy and optimizing therapeutic outcomes in bone oncology.

Fig. 4.

Comprehensive CT scans showcasing various views of the skeletal system relevant to bone cancer diagnosis and bone oncology.

2.3.2. CT likelihood evaluation

Using recommendations from the American College of Radiology (ACR), the likelihood of bone cancer on CT scans was classified as high, moderate, low, or negative. These classifications corresponded to categories such as highly suggestive of malignancy, indeterminate, suggestive of benignity, and normal.

2.3.3. Evaluation of CT scan severity

The severity of bone involvement was assessed by examining the extent and distribution of lesions. Each affected bone was divided into zones, and the level of involvement in each zone was scored: 1 represented less than 25 % involvement, 2 represented involvement between 25 % and 50 %, 3 represented involvements between 50 % and 75 %, and 4 represented more than 75 % involvement. A maximum score was calculated based on the number of involved zones. This severity score was then compared with the patient's clinical management and treatment decisions. By employing these CT techniques, the study aimed to enhance the understanding of the role of CT imaging in the diagnosis and monitoring of bone cancer, thereby aiding in clinical decision-making and improving patient outcomes.

2.3.4. Examination and analysis of diagnostic techniques

Each participant in our study underwent multiple imaging and biopsy tests over the course of their diagnosis and treatment. If initial biopsy results were deemed inadequate or inconclusive, up to three additional biopsies were performed to ensure diagnostic accuracy. Additionally, we meticulously recorded clinical decisions made at the time of presentation, such as whether surgery, chemotherapy, radiation therapy, or palliative care was required. These decisions were made by the oncology team in compliance with the hospital's established bone cancer management protocols. A range of clinical characteristics, including symptomatology, biopsy results, imaging findings, and the presence of risk factors that predispose individuals to the progression of bone cancer, were considered when making these determinations. Factors such as pain severity, functional impairment, lesion size, location, and metastasis were crucial in guiding treatment plans. The triage team also took into account the patient's overall health status, potential for recovery, and personal treatment preferences [17]. By integrating these comprehensive diagnostic and clinical evaluation methods, the study aimed to improve the accuracy of bone cancer diagnosis and optimize treatment strategies, ultimately enhancing patient outcomes and quality of life [18], [19], [20].

2.4. Determining the sample size for bone cancer diagnosis

In this study, we aim to determine the sensitivity of advanced imaging techniques in the detection of bone cancer. Assuming that 50 % of patients will have positive findings for bone cancer using these imaging techniques, we utilized the PSAA-11 program and the methods described by Simpson et al. [8] to calculate the necessary sample size. The calculations indicate that a minimum sample size of 100 patients is required to detect a sensitivity of 70 % for the imaging-based detection of bone cancer. This sample size ensures a statistical power of 80 %, meaning there is an 80 % probability of correctly rejecting the null hypothesis when it is false. Additionally, the significance level (α-error) is set at 0.05, indicating a 5 % risk of concluding that the imaging technique has a sensitivity different from 70 % when it actually does not. These parameters ensure that the study is adequately powered to detect clinically meaningful differences in the sensitivity of advanced imaging techniques for bone cancer detection, providing robust and reliable results shown in Fig. 5.

Fig. 5.

Detailed analysis of the frequency and range of severity scores related to bone involvement, highlighting key trends and patterns within the data set.

2.5. Statistical techniques for bone cancer diagnosis

Version 22.0 of IBM SPSS Statistics, created by IBM Corp. in 2013, was used for data analysis in our bone cancer diagnosis study. Descriptive statistical analysis was performed on quantitative data, including measurements like minimum, maximum, mean, and standard deviation (SD) for normally distributed data. The Shapiro-Wilk test assessed data normality. For comparing two independent groups, the independent t-test was used, while the ANOVA test analyzed multiple groups. Post hoc Bonferroni tests followed ANOVA to control type I error rates. For qualitative data, Fisher's exact test and the chi-square test were employed to evaluate proportionate differences, with the Bonferroni test applied for significant results. ROC curve analysis identified distinct diagnostic groups and evaluated test performance. A statistical significance threshold of P<0.050 was set. This rigorous statistical approach ensured robust and reliable findings, enhancing the validity and credibility of the bone cancer diagnosis study.

3. Results

For the first time, fifty patients suspected of having bone cancer were included in a comparative analysis. After excluding 134 individuals, the mean age of the remaining patients was between 30 and 92 years old, with a standard deviation of 71 years. Among these patients, only 11 had highly positive diagnostic results for bone cancer, and 31 were younger than 58 years old. Males comprised approximately 63.3 % of the patient group overall, with a male-to-female ratio of 1.7 among positive patients (Table 1). The interval between initial clinical symptoms and the diagnostic imaging ranged from two to seven days.

Table 1.

Patients range with yearly effect.

Year	Total Patients	Male	Female	Age	Mean Age
2022	19	13	6	32–84	59.9
2023	31	18	13	54–89	71.6
Total/Avg	50	31	19	32–89	67.2

This clinical trial involved single photon emission computed tomography (SPECT) imaging on a group of fifty patients diagnosed with bone cancer, both before and after treatment. The group consisted of 19 female and 31 male patients, with ages ranging from 30 to 92 years and an average age of -67.2 years. The participants received therapy and later SPECT imaging evaluations between December 2022 and June 2023. Eight patients, representing a subgroup of the patient population, had SPECT scans in June, both prior to and following treatment. These tests aimed to assess the effectiveness of the treatment through the identification of changes in bone lesions. Interestingly, the SPECT images provided in Fig. 6 and Fig. 7 had red circles drawn around the bone lesions. This thorough SPECT image analysis offers insightful information about the efficacy of the prescribed treatment plan, providing researchers and doctors with critical data for further assessment and improvement of therapeutic strategies for bone cancer management. Additional findings are covered in Fig. 6, Fig. 7 below.

Fig. 6.

Using SPECT scans from set 1–3 of medical records prior to and following the period. Red circles indicated bone lesions. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 7.

Using SPECT scans from groups 4–5 of medical records prior to and following in the period. Red circles indicated bone lesions. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 2 shows that the high probability group had the highest positive predictive value (PPV) and specificity, whereas the CT specificity decreased from the high probability group to the low probability group. As the positive events shown in Table 1 were interpreted, SPECT abnormalities were found and noted. Lytic lesions, present in 85.9 % of the positive patients, were the most commonly seen feature. As seen in Fig. 6, the majority of lytic lesions were peripheral, multifocal, and bilateral, affecting 88 %, 92.8 %, and 92.8 % of patients, respectively. Round forms accounted for 55.5 % of the patterns seen in lytic lesions (see Fig. 6). The “moth-eaten” pattern, which was observed in 45.3 % of positive patients, was the next most frequent SPECT feature (see Fig. 7). Of the positive cases, 74.5 % showed a subsegmental/segmental sclerosis pattern, with consolidation recorded in 28.8 % of the cases. Furthermore, 15.3 % of the patients exhibited a distinct accumulation pattern that was higher than the lytic lesions. Thus, our work identifies two SPECT phenotypes: one with dominant lytic lesions and the other with dominant sclerosis, based on prominent SPECT features shown in Fig. 6, Fig. 7. It was shown that the two management decision groups differed significantly in terms of pleural effusion and consolidation. In 44 of the cases requiring admission to the intensive care unit, chest CT scans showed a consolidation pattern. In 66.7 % of ICU patients, consolidation was the most prevalent pattern in contrast to lytic lesions. Moreover, Table 2 shows that 23 of the 38 patients with pleural effusions found on CT scans needed to be admitted to the intensive care unit (ICU).

Table 2.

Comparison of Single Photon Emission Computed Tomography (SPECT) Results for Bone Cancer and Bone Oncology Cases.

Variables		Total (N=50)	Positive (N=37)	Negative (N=13)	P value
Mean age (years)		46.0 ± 14.8	38.2 ± 21.2	44.0 ± 17.1	< 0.001 *
Sex	Male	31 (62 %)	24 (64.9 %)	7 (53.8 %)	0.484
	Female	19 (38 %)	13 (35.1 %)	6 (46.2 %)
CT likelihood	High	25 (50 %)	22 (59.5 %)	3 (23.1 %)	< 0.001*
	Intermediate	13 (36 %)	9 (24.3 %)	4 (30.8 %)
	Low	12 (24 %)	6 (16.2 %)	6 (46.2 %)

Three examples (27.3 %) exhibited a consolidation pattern, indicating significant bone sclerosis and density increase. Two patients showed mostly central giant cell tumor of bone (GCTB) lesions, characterized by large, lytic areas typically found in the metaphysis of long bones. One patient had peripheral GCTB (9.1 %), where the tumor was located towards the outer regions of the bone, leading to thinning and possible cortical breaches. Among the remaining patients, five (45.4 %) showed negative SPECT results for bone cancer, with no visible lesions or abnormalities. These negative findings underscore the variability in radiologic presentation of bone cancer, especially between adolescent and adult patients. In adolescents, bone tumors can sometimes be more challenging to detect due to the ongoing bone growth and remodeling processes, which can mask or mimic tumor characteristics.

This analysis highlights the complexity of diagnosing bone cancer using CT imaging, emphasizing the need for careful interpretation of different CT phenotypes, such as consolidation patterns and GCTB. Additionally, it illustrates the importance of considering patient age when evaluating radiologic images, as the presentation of bone cancer can differ significantly between younger and older patients. This comprehensive approach provides valuable insights for improving diagnostic accuracy and tailoring treatment strategies in bone oncology.

The study comprised 50 patients, with a mean age of 67.2 years. In 2022, the group consisted of 19 patients (13 males and 6 females) with an age range of 32 to 84 years and a mean age of 59.9 years. In 2023, there were 31 patients (18 males and 13 females) with an age range of 54 to 89 years and a mean age of 71.6 years. Significant differences were observed between the positive and negative cases (P<0.001*). The distribution of genders showed that 62 % of the cases were male, with similar percentages in both positive and negative instances (P=0.484). Remarkably, CT likelihood showed a considerable correlation with positive cases (P<0.001*), with a high prevalence of giant cell tumor of bone (GCTB) among them. There were no appreciable variations (P>0.05) in the GCTB attributes of rounding, linearity, and patchiness between the positive and negative situations. Peripheral GCTB placement, however, was significantly more closely linked to positive cases than negative cases (P<0.001*). These results point to a unique radiological profile linked to bone cancer, typified by the presence of GCTB and particular patterns of distribution.

Subsequent examination between positive and negative cases demonstrated significant variations in the type of consolidation and related radiological characteristics. Segmental/subsegmental consolidation was more common in positive instances, while lobar consolidation was more common in negative cases compared to positive cases (P=0.001*). Curving-pattern opacities were seen substantially more in positive cases than in negative cases (P<0.001*). Likewise, there was a substantial increase in the frequency of nodules and cavitations in positive cases (P<0.005* and P=0.006*, respectively) shown in Fig. 8. These distinct radiological features provide valuable information on the diagnostic utility of CT imaging in identifying and diagnosing bone cancer. The high prevalence of GCTB, particularly in peripheral locations, and the presence of segmental/subsegmental consolidation in positive cases, highlight the importance of detailed CT imaging in the diagnosis and management of bone cancer. This aids in patient management and treatment decisions, offering a clear pathway for distinguishing between different radiological profiles associated with bone cancer.

Fig. 8.

Comparison of mean age, male percentage (A), and CT likelihood (B) between 2022 and 2023 in bone oncology.

4. Discussion

The lowest crude incidence rates of severe bone lesions observed on radiographs and CT scans were seen among patients treated within 90 days: 18 % (277 out of 1527) and 31 % (66 out of 216), respectively. However, for those treated 91 to 120 days earlier, these rates increased to 22 % (172 out of 783) for radiographs and 40 % (65 out of 161) for CT scans. For those treated 121 to 180 days earlier, the rates were 27 % (274 out of 1032) for radiographs and 34 % (73 out of 213) for CT scans. For those treated 181 to 240 days earlier, the incidence rates were 32 % (159 out of 496) for radiographs and 40 % (51 out of 126) for CT scans. For those treated more than 240 days before diagnosis, the rates were 31 % (110 out of 358) for radiographs and 38 % (61 out of 162) for CT scans shown in Table 1, Table 2. Nonetheless, there was a decline in the crude incidence rate of typical bone lesions: 44 % (69 out of 158) of those who received their treatment within 90 days, 37 % (42 out of 114) of those treated between 91 and 120 days, 32 % (51 out of 159) of those treated between 121 and 180 days, 31 % (31 out of 101) of those treated between 181 and 240 days, and 23 % (27 out of 116) of those treated more than 240 days before diagnosis. Fig. 9 displays the smoothing function of odds ratios (ORs) for (a) severe clinical outcomes (such as invasive procedures or mortality), and (b) based on the period (in days) between the last treatment and the bone cancer diagnosis. The statistics indicate that the likelihood of serious clinical outcomes increases gradually following the last dose of treatment. In the adjusted multivariable model, patients who had received their last treatment more than 240 days prior to diagnosis had a significantly higher probability of experiencing severe clinical consequences than patients who had received their treatment less than 90 days prior (OR=1.94 [95 % CI: 1.16, 3.24]; P=0.01). The odds ratios (ORs) for severe bone lesions on radiographs, both unadjusted and adjusted, showed a steady rise over time when comparing the groups treated beyond 90 days to those treated within 90 days shown in Fig. 9.

Fig. 9.

Incidence Rates and Odds Ratios for Bone Lesions and Clinical Outcomes Over Time.

In comparison to the group that received treatment within 90 days of diagnosis, Fig. 9 demonstrates that the group that received treatment more than 240 days prior to diagnosis had higher adjusted odds ratios (ORs) for severe bone lesions on radiographs (OR=1.65 [95 % CI: 1.13, 2.40]; P=0.009). However, there was no significant difference in the adjusted ORs for severe bone lesions on radiographs between the groups treated fewer than 240 days before and those treated within 90 days of diagnosis (P=0.15 to P=0.86). The adjusted odds ratios (ORs) for severe bone lesions detected on CT scans or presenting typical CT findings did not differ significantly between the groups that received treatment more than 90 days prior to diagnosis and the group that received treatment within 90 days (p-values ranging from 0.17 to 0.88 and 0.23 to 0.72, respectively) shown in and Fig. 10.

Fig. 10.

Relationships between incidence rates and adjusted odds ratios (ORs) for severe bone lesions on radiographs and CT scans over varying time periods.

Furthermore, our study reveals significant associations between clinical characteristics and the presence of severe bone lesions as revealed on SPECT scans or radiographs. These findings underscore the correlation between various clinical aspects and the common CT symptoms of bone cancer. In the context of bone cancer diagnosis, no single imaging modality is superior to the others. Each technique has its unique strengths and limitations, and a combination of these modalities is often necessary for a comprehensive assessment. In X-rays may be the first imaging test performed due to their low cost and wide availability. However, they may miss early signs of bone cancer. CT is used to assess bone destruction, cortical breaks, and the relationship of tumors to adjacent tissues. MRI is used to assess soft tissue involvement, bone marrow infiltration, and functional information. PET-CT is used for whole-body imaging, staging, and monitoring therapy response. Notably, our analysis highlights how timing of treatment relative to diagnosis can influence the severity and detection of bone lesions, providing valuable insights for optimizing treatment strategies in bone oncology.

In recent years, neural networks, particularly those leveraging deep learning techniques, have demonstrated remarkable capabilities in analyzing intricate medical imaging data and extracting essential information for clinical decision-making processes [21], [22], [23]. By automatically detecting and quantifying imaging features, neural networks and AI-based CT analysis can improve diagnostic accuracy, facilitate dynamic disease tracking, and predict treatment response and patient outcomes.

By integrating neural networks [22] and AI-based CT analysis within CT-based bone cancer diagnosis and monitoring frameworks, diagnostic accuracy can be significantly enhanced. This integration facilitates automated feature extraction and enables dynamic tracking of disease progression. Through precise identification of alterations in tumor size, shape, and location, the emergence of new lesions or metastases can be promptly detected, thereby enabling timely interventions and the customization of treatment plans. Moreover, neural networks [22], [23] offer predictive insights into disease severity and outcomes, aiding in the identification of patients who may require intensive care or are at an elevated risk of complications. By anticipating these needs, earlier and more targeted interventions can be implemented, concurrently suggesting personalized treatment strategies tailored to each patient's unique profile. Furthermore, the continuous training and learning capabilities of neural networks, as well as advanced modeling techniques [24], [25] that are fueled by new data inputs, underscore their potential in fostering continuous improvement in diagnostic accuracy and personalized care delivery. Medical image analysis of the bone [26], [27], [28] that is supported by the deep-learning based framework for medical applications [29], [30] has strong clinical potential for orthopedics diagnostics. Ultimately, these advancements hold promise for optimizing patient outcomes and enhancing the efficient utilization of medical resources.

5. Conclusion

In the realm of bone cancer diagnosis and management, where timely histopathological analysis may be challenging, our study thoroughly evaluated the AI-assisted diagnostic efficacy of computed tomography (CT) and its impact on patient care. Conducted under established oncology imaging guidelines, our retrospective analysis of 60 symptomatic patients aligned closely with histopathological findings, affirming CT's pivotal role in early detection and management. CT has emerged as indispensable for early disease detection and management, adhering rigorously to oncological imaging standards to minimize the risk of false-negative results. Our findings underscore the critical importance of prompt CT examinations in achieving precise staging of bone tumors, encompassing both primary malignancies and metastatic lesions. This timely imaging approach is pivotal in optimizing diagnostic accuracy and guiding tailored treatment strategies. Furthermore, our study identified distinct CT phenotypes such as Giant Cell Tumors of Bone (GCTB) and various consolidation patterns indicative of diverse tumor types and stages. Notably, specific consolidation patterns and the extent of lesion spread were predictive of heightened risks necessitating intensive care unit (ICU) intervention. CT severity scores proved invaluable in forecasting the need for aggressive therapeutic interventions and assessing skeletal involvement comprehensively. Overall, our research contributes significant insights into the clinical nuances of bone cancer diagnosis and underscores the pivotal role of CT imaging in enhancing patient outcomes. Our findings emphasize the imperative of early and precise CT-based evaluations to delineate disease extent, differentiate between benign and malignant lesions, and inform effective treatment planning. These insights provide actionable recommendations for clinicians to optimize bone cancer management and resource allocation in clinical practice.

Ethics approval and consent to participate.

This study received approval from the institutional board of ethics.

Consent for publication

Yes.

Funding

This research was funded by Fund of the Natural Science Foundation of Yangzhou, grant number YZ2023161, Young Scientists Fund of the National Natural Science Foundation of China, grant number 82102831 and China Health Promotion Foundation, grant number XH-B047.

CRediT authorship contribution statement

Wei Hua: Writing – original draft, Software, Investigation. Bing Xu: Validation, Methodology. Xianwen Zhang: Supervision, Software, Resources. Tingting Chen: Writing – review & editing, Project administration, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.