Frontiers in medical technology最新文献

{"title":"Augmenting medical image specific foundation model with classical radiomic signatures for improved nasopharyngeal carcinoma stage classification.","authors":"Yu Li, Jiacheng Liu, Xin Wen, Xiaomin Niu, Lai Wei, Junjie Yuan, Zhiqiang Wang, Rong Na","doi":"10.3389/fmedt.2026.1863203","DOIUrl":"10.3389/fmedt.2026.1863203","url":null,"abstract":"<p><strong>Background: </strong>Accurate clinical staging is the critical process of prognosis and treatment planning for nasopharyngeal carcinoma (NPC). Conventional radiomics is limited to hand-crafted features that may overlook complex patterns, while generic deep learning models often lack domain specificity. Currently, SAM-Med3D, a vision transformer (ViT)-based medical model excels at capturing global volumetric context. We hypothesized that fusing radiomic and SAM-Med3D feature-extraction would yield synergistic value. This study aims to construct an MRI-based fusion model for NPC stage classification (Stage I, II and III vs. Stage IVa) and benchmark its performance against experienced experts.</p><p><strong>Methods: </strong>We retrospectively analyzed 264 patients with non-metastatic nasopharyngeal carcinoma, partitioned into training (<i>n</i> = 198) and independent validation (<i>n</i> = 66) cohorts. MRI sequences (T1WI, T2WI, CE-T1WI) were utilized to extract high-dimensional features by two distinct paradigms: hand-crafted radiomics and deep features from SAM-Med3D. An early-fusion dataset (SAM-PR) integrating both feature types was also constructed. Following LASSO-based dimensionality reduction, five machine learning classifiers (LR, SVM, RF, XGB, LGBM) were developed for staging. Model performance was evaluated using the area under the curve (AUC) and benchmarked against the independent, blinded staging of two senior otolaryngologists. Optimism-Corrected Bootstrap analysis was applied to evaluate the reliability of AUC, and the learning curve was conducted to avoid overfitting.</p><p><strong>Results: </strong>LASSO regression identified an optimal fusion set of 17 features (9 hand-crafted and 8 deep-learned). Among all 15 constructed DL models, the fusion-based LR yielded the optimal performance (AUC = 0.760, optimism-corrected AUC of 0.836), and the fusion approaches consistently yielded higher AUCs than individual feature approaches. Compared to human experts, our optimal model achieved a more balanced performance (Sensitivity: 73.91% vs. 88%; Specificity: 69.77% vs. 26%-30%) and a higher F1-score (0.615 vs. 0.353-0.400).</p><p><strong>Conclusions: </strong>The integrated model, combining conventional radiomics with SAM-Med3D features, consistently showed higher accuracy than its unimodal counterparts across five distinct algorithms (LR, SVM, RF, XGB, LGBM) for NPC stage I, II, III and stage IVa differentiation. All five fusion models demonstrated comparable discriminative ability compared to experienced experts. These findings suggest that hybrid feature frameworks can serve as powerful objective tools in NPC staging.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1863203"},"PeriodicalIF":4.6,"publicationDate":"2026-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13415764/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148623240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development of a new suction instrument to reduce surgical smoke exposure in the operating room.","authors":"Weiwei Chen, Lumin Lei, Mingzi Xiao, Qiaoling Chen, Yanlan Ma","doi":"10.3389/fmedt.2026.1796828","DOIUrl":"10.3389/fmedt.2026.1796828","url":null,"abstract":"<p><strong>Objective: </strong>To remove surgical smoke and its toxic components, we developed a novel smoke removal tube nested with the intraoperative suction tube, enabling convenient application during surgery and reducing hazards of surgical smoke exposure to medical staff.</p><p><strong>Methods: </strong>Standardized cuts with an electric knife were performed on pig dorsal tissue in an animal laboratory operating room without vertical laminar flow. The generated surgical fumes were analyzed using microcomputer laser dust meter counters for real-time monitoring of particulate matter 2.5 μm (PM2.5). Three new design suction tubes with different orifice diameters (1.5, 2.0, and 2.5 mm), each connected to local smoke evacuation, were compared with a conventional instrument without smoke evacuation.</p><p><strong>Results: </strong>The PM2.5 levels were significantly reduced with all three new tools, with smoke reduction rates of 24.20%, 66.86%, and 43.02%, respectively. Ten seconds after cutting, the 2.0-mm-diameter suction tube demonstrated lower PM2.5 concentrations compared with the other two groups, and the difference was statistically significant (<i>P</i> < 0.05).</p><p><strong>Conclusion: </strong>This study is the first to develop an \"all-in-one\" surgical smoke device that enables simultaneous suction of smoke and blood with one hand. Among the tested tools, the 2.0-mm-diameter instrument is the best one. These results promote the use of local smoke evacuation and may inspire other surgeons to further innovation in this field.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1796828"},"PeriodicalIF":4.6,"publicationDate":"2026-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13396161/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148586413","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An attention-based transfer learning framework for breast cancer classification in mammography under limited data conditions.","authors":"Niketa Banyal, Aman Kumar","doi":"10.3389/fmedt.2026.1842859","DOIUrl":"10.3389/fmedt.2026.1842859","url":null,"abstract":"<p><p>Breast cancer is a leading cause of death for women worldwide. There is a critical need for an early and accurate diagnosis to improve survival rates. This study evaluates the performance of transfer learning-based MobileNetV2 and ResNet50 architectures for binary breast cancer classification using mammography images. To enhance channel-wise feature learning, the Squeeze-and-Excitation (SE) attention mechanism was integrated into the network and partial fine tuning was performed by only freezing the early layers of the model. Furthermore, gradient-weighted class activation mapping was applied to visualize the important regions of mammogram images responsible for the model's predictions. The performance evaluation was carried out on the Contrast Limited Adaptive Histogram Equalization (CLAHE) binary, Digital Database for Screening Mammography (DDSM), INbreast, and Mammographic Image Analysis Society (MIAS) datasets using stratified k-fold cross-validation. The experimental results demonstrate that one of the proposed frameworks, MobileNetV2-SE, achieved mean classification accuracies of 99.14%, 93.53%, 92.35%, and 96.33% on the DDSM, INbreast, MIAS, and CLAHE-binary datasets, respectively. In comparison, the ResNet50-SE model attained mean accuracies of 100%, 88.89%, 71.60%, and 97.98% on the same datasets. The experimental results demonstrate that the frameworks achieve reliable and competitive performance for breast cancer detection under realistic evaluation conditions and improve the diagnostic accuracy of mammogram classification. This work contributes to the field of medical image analysis and computer-aided diagnosis. Future work should focus on applying advanced class balancing methods to improve diagnostic accuracy and model robustness in a variety of medical imaging applications.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1842859"},"PeriodicalIF":4.6,"publicationDate":"2026-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13391876/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148581926","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Generative adversarial networks for balancing and expanding data resources for computer-aided detection in colonoscopy.","authors":"J J H van der Laan, J van Lune, L R B Schomaker, P M A van Ooijen, W B Nagengast","doi":"10.3389/fmedt.2026.1824811","DOIUrl":"10.3389/fmedt.2026.1824811","url":null,"abstract":"<p><strong>Background and aims: </strong>Developing computer-aided detection (CADe) algorithms in colonoscopy requires high-quality databases sourced from real-world data. However, such databases demand costly investments and display imbalanced distributions that underrepresent flat adenomas. Therefore, we evaluated whether StyleGAN2-ADA, a generative adversarial network, can generate synthetic images as alternative training resources for CADe, and whether a modified StyleGAN2-ADA can enhance synthesis control for better balanced datasets.</p><p><strong>Methods: </strong>Two synthetic datasets were generated using the original StyleGAN2-ADA and our modified version with feature-clustered conditioning vectors. Generative adversarial networks were trained on images from our local colonoscopy database, and their outputs were evaluated through Fréchet Inception Distance where lower scores indicate more realistic and diverse data. Three CADe models were trained using synthetic or real-world data and tested in two external databases comprising only polyp images. CADe performance included mean average precision (mAP) score as critical metric for polyp identification.</p><p><strong>Results: </strong>The modified StyleGAN2-ADA achieved a lower Fréchet Inception Distance-score (7.54 vs. 13.68) than the original StyleGAN2-ADA. During external testing, CADe trained on synthetic data from the modified version achieved performance comparable to the model trained on real-world data across all metrics (<i>p</i> > 0.05). It also outperformed the model trained on synthetic data from the original StyleGAN2-ADA, achieving mAP-scores of 0.77 ± 0.03 vs. 0.64 ± 0.02 in the first test set (<i>p</i> < 0.001) and 0.91 ± 0.02 vs. 0.87 ± 0.02 in the second (<i>p</i> = 0.012).</p><p><strong>Conclusions: </strong>StyleGAN2-ADA incorporating feature-clustered conditioning vectors can synthesize better balanced colonoscopy databases that offer alternative training resources to real-world data for CADe training and validation.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1824811"},"PeriodicalIF":4.6,"publicationDate":"2026-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13391578/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148581866","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Continuous BP monitoring of ICU Taiwanese patients: training-phase evaluation of oCare™ BP100 according to ISO 81060-3: 2022.","authors":"Chao-Lun Lai, Yeh-Wen Lee, Chia-Fang Chang, Chang-Sheng Chu","doi":"10.3389/fmedt.2026.1681323","DOIUrl":"10.3389/fmedt.2026.1681323","url":null,"abstract":"<p><strong>Objective: </strong>This study aimed to conduct a training-phase evaluation of a novel continuous automated non-invasive sphygmomanometer, oCare™ BP100, in intensive care unit (ICU) Taiwanese patients according to ISO 81060-3:2022 in Type T mode. The evaluation included stability testing and blood pressure change tracking.</p><p><strong>Methods: </strong>A total of forty ICU patients were recruited from the National Taiwan University Hospital Hsin-Chu Branch. All enrolled patients had undergone unilateral radial artery catheterization in the forearm as part of routine clinical care. The catheter was connected to Philips patient monitors (Models: IntelliVue MX800 and X2) to obtain invasive blood pressure measurements for calibration purposes. Each subject underwent simultaneous monitoring with oCare™ BP100 sensors placed on the fingers of both hands. Paired measurements of the same side were further processed and followed ISO 81060-3:2022, for the defined Type T blood pressure monitor, the well-established stability method and the blood pressure change method were implemented and analyzed.</p><p><strong>Results: </strong>The present study belongs to the training phase and therefore the study was not carried out completely according to the ISO 81060-3:2022 protocol, e.g., the distribution of gender and the eligible subject in BP change test not fulfilled. For all 6 analysis periods, for the corrected standard deviations of SBP, DBP, and MAP parameter defined ISO target thresholds (A) s_corr ≤ 6 mmHg for each analysis period. In the Blood Pressure Change Test, at most 22 subjects the minimum requirement of 50 change points per subject, the for the change requirements were (B1) <math> <msub> <mrow><mover><mi>E</mi> <mo>¯</mo></mover> </mrow> <mrow><mn>50</mn> <mi>t</mi> <mi>h</mi></mrow> </msub> <mo>≤</mo> <mn>25</mn> <mtext>%</mtext></math> and (B2) <math> <msub> <mrow><mover><mi>E</mi> <mo>¯</mo></mover> </mrow> <mrow><mn>85</mn> <mi>t</mi> <mi>h</mi></mrow> </msub> <mo>≤</mo> <mn>50</mn> <mtext>%</mtext></math> error for blood pressure.</p><p><strong>Conclusion: </strong>The present study demonstrated that a training-phase evaluation of Type T mode continuous automated non-invasive sphygmomanometer repeated SBP, DBP, and MAP measurement of oCare™ BP100 in Type T mode are stable across various analysis periods and patients. oCare™ BP100 in Type T Mode can track blood pressure changes in ICU patients during different activities. Even with the absence of females in the sex ratio and an insufficient number of participants in the blood pressure change test, the exploratory study still showed training-phase results that were within ISO target thresholds, full ISO validation remains to be completed.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1681323"},"PeriodicalIF":4.6,"publicationDate":"2026-07-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13384929/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148551886","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Member selection in traditional Chinese medicine innovation consortia: an evolutionary game theory approach.","authors":"Dong Hua, Yanhui Wang","doi":"10.3389/fmedt.2026.1786671","DOIUrl":"https://doi.org/10.3389/fmedt.2026.1786671","url":null,"abstract":"<p><strong>Introduction: </strong>In recent years, long and uncertain research and development (R&D) cycles, difficulties in patent protection, and fragmented innovation activities have jointly constrained the traditional Chinese medicine (TCM) industry, making many enterprises hesitant to pursue independent innovation. Despite a relatively complete industrial chain, traditional Chinese medicine (TCM) industry faces persistent challenges in technological innovation and commercialization. Collaborative innovation through TCM innovation consortia has emerged as a strategic mechanism to integrate heterogeneous resources and enhance innovation capacity. Understanding the determinants of member selection and consortium formation dynamics is essential for promoting high-quality development.</p><p><strong>Methods: </strong>This study develops an evolutionary game model under bounded rationality to simulate potential consortium members' decision-making. Key factors-resource complementarity, transaction costs, participant strengths, and third-party involvement-are incorporated to evaluate their impact on stable consortium formation.</p><p><strong>Results and discussion: </strong>Analysis indicates that higher resource complementarity, balanced participant strengths, greater expected returns, and third-party support significantly increase the likelihood of stable consortium formation. Similar mechanisms are observed in international innovation ecosystems, where intermediaries, policy guidance, and cross-boundary partnerships reduce information asymmetries and transaction costs. These findings offer theoretical insights into dynamic member selection and practical guidance for enhancing governance, operational stability, and coordination efficiency in TCM innovation consortia. By learning from global collaboration practices, more effective resource integration and cooperation can strengthen innovation capacity and support the high-quality development of China's TCM industry.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1786671"},"PeriodicalIF":4.6,"publicationDate":"2026-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13372755/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148474854","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Reproducible candidate kinematic-electromyographic waveform markers of post-stroke gait from public multimodal waveform exports.","authors":"Rocco Salvatore Calabrò, Andrea Calderone, Fabrizio Sottile, Carmela Casella, Antonino Naro, Gokhan Ozkocak, Angelo Quartarone","doi":"10.3389/fmedt.2026.1863908","DOIUrl":"10.3389/fmedt.2026.1863908","url":null,"abstract":"<p><strong>Background: </strong>Instrumented gait analysis after stroke is informative but difficult to scale for routine rehabilitation use. This secondary analysis examined whether a spreadsheet-restricted subset of public sagittal kinematic and surface electromyography (sEMG) waveforms could yield clinically legible candidate waveform-derived markers and reproducible subject-level summaries.</p><p><strong>Methods: </strong>We analyzed a public multimodal gait dataset with 138 able-bodied adults and 50 adults with stroke. The workflow used only public spreadsheet exports and 11 domains shared across cohorts: four sagittal kinematic waveforms and seven normalized sEMG waveforms. Subject-level normative deviation and within-stroke asymmetry summaries were derived. Kinematics-only, sEMG-only, and combined kinematic-sEMG panels were compared using internal information-retention metrics within a nested benchmarking framework, not against an external clinical endpoint.</p><p><strong>Results: </strong>All kinematic domains were complete in both cohorts. Seven-channel sEMG completeness was lower, yielding 102 able-bodied controls, 43 paretic stroke sides, 44 non-paretic stroke sides, and 43 paired stroke participants for sEMG-containing complete-case benchmarking. Combined-panel deviation burden remained non-trivial bilaterally, with the largest median domain-level abnormalities in gastrocnemius activity and knee-angle waveforms. Relative to the combined internal reference panel, kinematics-only and sEMG-only reductions preserved substantial ranking information in 43 paired complete cases, with Pearson correlations of 0.865 and 0.875 and Spearman correlations of 0.839 and 0.884, respectively. Bootstrap intervals and sensitivity analyses indicated overlap between reduced panels and strongest robustness for kinematics-only summaries.</p><p><strong>Conclusion: </strong>Public spreadsheet waveform exports can support reproducible candidate gait markers, but reduced panels should be interpreted as internally benchmarked summaries rather than validated clinical biomarkers or prospective rehabilitation decision endpoints.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1863908"},"PeriodicalIF":4.6,"publicationDate":"2026-07-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13373056/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148474877","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Transcatheter dual-device closure for symptomatic iatrogenic atrial septal defect following bioprosthetic mitral valve replacement: a case report.","authors":"Ping-An Lian, Shan-Fu Liang, Fei Xie, Shuai Cheng, Zhan-Zhan Zhu, Yun-Fei Zhao, Lei Yin, Shen-Wei Zhang","doi":"10.3389/fmedt.2026.1847831","DOIUrl":"10.3389/fmedt.2026.1847831","url":null,"abstract":"<p><p>Iatrogenic atrial septal defect (iASD) is a relatively common complication after cardiac surgical or catheter-based procedures involving transseptal access. While most defects close spontaneously, a subset may persist and become hemodynamically significant, leading to clinical deterioration and necessitating intervention. We report a case of a large, persistent iASD that led to progressive right-sided heart failure following bioprosthetic mitral valve replacement combined with tricuspid annuloplasty. A 69-year-old male underwent bioprosthetic mitral valve replacement and tricuspid annuloplasty. Postoperatively, the patient developed a large, persistent iASD, which contributed to severe right-sided heart failure. The patient also had end-stage liver disease, portal hypertension, and severe coagulopathy, which made reoperation extremely high-risk. After heart team evaluation, a transcatheter closure approach was chosen. Initially, a 24-mm occluder was deployed, but significant residual shunting remained. A planned two-device strategy was then performed, resulting in complete closure of the defect. Post-procedure imaging confirmed the absence of residual shunt, and tricuspid regurgitation significantly improved. The patient's symptoms improved to New York Heart Association NYHA I-II, and at six-month follow-up, the device remained stable, and cardiac function had recovered. This case illustrates that in patients with large, complex, or poorly rimmed iASDs, particularly those who are not suitable candidates for reoperation, the dual-device strategy represents a safe and effective percutaneous alternative.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1847831"},"PeriodicalIF":4.6,"publicationDate":"2026-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13364760/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148451052","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep learning and body composition model for predicting postoperative complications in colorectal cancer.","authors":"Ding Ding, Ran Xuan, Rui Li","doi":"10.3389/fmedt.2026.1817439","DOIUrl":"10.3389/fmedt.2026.1817439","url":null,"abstract":"<p><strong>Background: </strong>Colorectal cancer (CRC) is the third most commonly diagnosed malignancy worldwide and remains a leading cause of cancer-related mortality. Surgical resection remains the cornerstone of curative treatment for CRC; however, postoperative complications, including anastomotic leakage, infections, and thromboembolic events, continue to substantially affect patient prognosis. These complications are associated with prolonged hospitalization, increased healthcare expenditures, delayed postoperative recovery, and elevated mortality risk. Therefore, the accurate identification of patients at high risk for postoperative complications is of considerable clinical importance for optimizing perioperative management and improving surgical outcomes.</p><p><strong>Methods: </strong>A total of 154 patients were retrospectively enrolled, including 99 patients in the training cohort and 55 patients in the external validation cohort. Abdominal CT images at the L1-L5 levels were automatically segmented using a pretrained DeepLabv3-ResNet101 model implemented on the Onekey platform, followed by manual correction to ensure segmentation accuracy. Visceral adipose tissue, subcutaneous adipose tissue, skeletal muscle, and intramuscular adipose tissue were segmented, and body composition indices (VAT_h, SAT_h, SMA_h, and IMAT_h) were subsequently calculated after normalization by patient height. Deep learning features were extracted from CT images using a pretrained 3D ResNet-18 model. Feature selection and model development were conducted independently. Random forest, logistic regression, and extremely randomized trees combined with recursive feature elimination were applied for feature selection. Subsequently, logistic regression, random forest, support vector machine (SVM), k-nearest neighbors (KNN), XGBoost, and LightGBM algorithms were constructed and compared for predictive performance.</p><p><strong>Results: </strong>Univariate and multivariate analyses identified BMI, SMA_h, SAT_h, and VAT_h as independent risk factors for postoperative complications. Among all predictive models, the combined model incorporating the deep learning score (DL-score) achieved the best performance, with AUC values of 0.944 and 0.855 in the training and validation cohorts, respectively. DeLong test results further demonstrated that both the deep learning model and the combined model significantly outperformed the clinical and body composition models, whereas the combined model provided superior discrimination in identifying high-risk patients.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1817439"},"PeriodicalIF":4.6,"publicationDate":"2026-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13357902/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148440636","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Economic evaluation of AI-based oral disease screening: a systematic review.","authors":"Rakesh Kumar Sahoo, Abhinav Sinha, Sayani Satpathi, Gunjan Kumar, Krushna Chandra Sahoo, Debdutta Bhattacharya, Bhuputra Panda, Sanghamitra Pati","doi":"10.3389/fmedt.2026.1792401","DOIUrl":"10.3389/fmedt.2026.1792401","url":null,"abstract":"<p><strong>Objectives: </strong>Artificial intelligence (AI) has made a significant contribution to dentistry, particularly in improving screening and diagnostic accuracy. This systematic review assessed the cost-effectiveness and broader economic implications of incorporating AI into oral disease screening.</p><p><strong>Methods: </strong>We included studies that conducted economic analyses and/or cost-effectiveness evaluations of AI-driven tools compared with conventional techniques for detecting oral diseases. We searched Cochrane Library, CINAHL, Embase, PubMed, Scopus, and Web of Science. The search strategy combined terms related to economic evaluation, artificial intelligence, and oral diseases. The review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and was registered in PROSPERO (CRD42024548039). Articles published up to January 12, 2025, were included.</p><p><strong>Results: </strong>Of 334 records identified, four studies met the inclusion criteria - all these studies dealt with dental economic evaluations, and were conducted in the German dental care settings. All four evaluated AI-assisted detection of dental or proximal caries on bitewing radiographs. Across studies, AI generally improved diagnostic performance, particularly sensitivity. In one modelling study, a U-Net-based system achieved higher accuracy than dentists (0.80 vs. 0.71). Three model-based economic evaluations reported lower costs and greater tooth-retention outcomes with AI than with conventional assessment. Costs ranged from €298 to €378 for AI and from €322 to €419 for conventional care. Tooth retention ranged from 62.4 to 64.0 years with AI and from 60.4 to 62.0 years with conventional care. In contrast, the randomized trial found that AI-supported assessment increased sensitivity but was associated with more invasive treatment decisions. This resulted in similar costs (€330 vs. €330) and identical mean tooth retention (49 years) compared with standard assessment. Overall, the evidence suggests a potential economic benefit of AI in caries detection, but uncertainty remains under real-world treatment conditions.</p><p><strong>Conclusions: </strong>AI-assisted screening may offer economic value in dental caries detection, particularly when improved lesion detection is linked to appropriate downstream management. These findings are relevant for clinicians, payers, and health-system decision-makers considering the integration of AI into diagnostic pathways. Future multicentre, real-world economic evaluations are needed to guide reimbursement, implementation, and equitable adoption across diverse settings.</p><p><strong>Systematic review registration: </strong>https://www.crd.york.ac.uk/PROSPERO/view/CRD42024548039, PROSPERO: CRD42024548039.</p>","PeriodicalId":94015,"journal":{"name":"Frontiers in medical technology","volume":"8 ","pages":"1792401"},"PeriodicalIF":4.6,"publicationDate":"2026-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13346071/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148427680","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}