The release of ChatGPT has spurred the widespread adoption of generative large language models (LLMs) in healthcare. This scoping review systematically examines their use in healthcare.
Methods
A systematic search was conducted using PubMed, a comprehensive and representative database on biomedical and health science, to identify studies published between January 1, 2023, and July 30, 2024. Studies were included if they assessed the performance of generative LLMs in healthcare applications; review or perspective articles were excluded.
Results
A total of 415 studies were included, with a significant increase in publications observed after April 2023. Generative LLMs were applied across various medical specialties, primarily supporting clinical decision-making (26.7%) and providing patient information (23.9%). Smaller proportions were focused on professional education and training (18.1%), research (16.1%), and workflow support (12.5%). These applications were mainly supported by three key NLP tasks: question answering (36.1%), text classification (27.5%), and text generation (26.3%). Public datasets appeared in 20% of studies, and 15% used clinical patient data. Of the 98 LLMs used, GPT-4 (51.3%), GPT-3.5 (36.6%), and ChatGPT (22.4%) were the most common. Direct prompting was the most common adaptation method (92.5%), with reinforcement learning rarely utilized (1.4%). Accuracy was the most frequently assessed metric, while errors and safety (9.4%) and time efficiency (7.0%) were less commonly evaluated.
Conclusion
LLMs hold promise across healthcare applications. Expanding their use in workflow optimization, trainee education, and research tools could enhance healthcare delivery and innovation. Comprehensive evaluation using standardized criteria is essential for LLMs integration into healthcare.
{"title":"Advancing healthcare with large language models: A scoping review of applications and future directions","authors":"Zhihong Zhang , Mohamad Javad Momeni Nezhad , Seyed Mohammad Bagher Hosseini , Ali Zolnour , Zahra Zonour , Seyedeh Mahdis Hosseini , Maxim Topaz , Maryam Zolnoori","doi":"10.1016/j.ijmedinf.2025.106231","DOIUrl":"10.1016/j.ijmedinf.2025.106231","url":null,"abstract":"<div><h3>Background</h3><div>The release of ChatGPT has spurred the widespread adoption of generative large language models (LLMs) in healthcare. This scoping review systematically examines their use in healthcare.</div></div><div><h3>Methods</h3><div>A systematic search was conducted using PubMed, a comprehensive and representative database on biomedical and health science, to identify studies published between January 1, 2023, and July 30, 2024. Studies were included if they assessed the performance of generative LLMs in healthcare applications; review or perspective articles were excluded.</div></div><div><h3>Results</h3><div>A total of 415 studies were included, with a significant increase in publications observed after April 2023. Generative LLMs were applied across various medical specialties, primarily supporting clinical decision-making (26.7%) and providing patient information (23.9%). Smaller proportions were focused on professional education and training (18.1%), research (16.1%), and workflow support (12.5%). These applications were mainly supported by three key NLP tasks: question answering (36.1%), text classification (27.5%), and text generation (26.3%). Public datasets appeared in 20% of studies, and 15% used clinical patient data. Of the 98 LLMs used, GPT-4 (51.3%), GPT-3.5 (36.6%), and ChatGPT (22.4%) were the most common. Direct prompting was the most common adaptation method (92.5%), with reinforcement learning rarely utilized (1.4%). Accuracy was the most frequently assessed metric, while errors and safety (9.4%) and time efficiency (7.0%) were less commonly evaluated.</div></div><div><h3>Conclusion</h3><div>LLMs hold promise across healthcare applications. Expanding their use in workflow optimization, trainee education, and research tools could enhance healthcare delivery and innovation. Comprehensive evaluation using standardized criteria is essential for LLMs integration into healthcare.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"208 ","pages":"Article 106231"},"PeriodicalIF":4.1,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145829162","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-15Epub Date: 2025-12-28DOI: 10.1016/j.ijmedinf.2025.106249
Wen-Jiang Yang
{"title":"Commentary on “Towards practical federated learning and evaluation for medical prediction models”","authors":"Wen-Jiang Yang","doi":"10.1016/j.ijmedinf.2025.106249","DOIUrl":"10.1016/j.ijmedinf.2025.106249","url":null,"abstract":"","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"208 ","pages":"Article 106249"},"PeriodicalIF":4.1,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145879347","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-15Epub Date: 2025-12-24DOI: 10.1016/j.ijmedinf.2025.106238
Yuto Arai , Yoshiyasu Takefuji
{"title":"Limitations of SHAP-based interpretability in sepsis progression models and paths to more robust feature validation","authors":"Yuto Arai , Yoshiyasu Takefuji","doi":"10.1016/j.ijmedinf.2025.106238","DOIUrl":"10.1016/j.ijmedinf.2025.106238","url":null,"abstract":"","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"208 ","pages":"Article 106238"},"PeriodicalIF":4.1,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145841723","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-15Epub Date: 2025-12-19DOI: 10.1016/j.ijmedinf.2025.106228
Siyi Liu , Zekai Yu
{"title":"Comment on “Medication-based mortality prediction in COPD using machine learning and conventional statistical methods”","authors":"Siyi Liu , Zekai Yu","doi":"10.1016/j.ijmedinf.2025.106228","DOIUrl":"10.1016/j.ijmedinf.2025.106228","url":null,"abstract":"","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"208 ","pages":"Article 106228"},"PeriodicalIF":4.1,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145799805","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-15Epub Date: 2025-12-18DOI: 10.1016/j.ijmedinf.2025.106230
Emiko Shinohara, Yoshimasa Kawazoe
Background
Named entity recognition (NER) is critical in natural language processing (NLP), particularly in the medical field, where accurate identification of entities, such as patient information and clinical events, is essential. Traditional NER approaches rely heavily on large, annotated corpora, which are resource intensive. Large language models (LLMs) offer new NER approaches, particularly through in-context and few-shot learning.
Objective
This study investigates the effects of incorporating annotation guidelines into prompts for NER via LLMs, with a specific focus on their impact on few-shot learning performance across various medical corpora.
Methods
We designed eight different prompt patterns, combining few-shot examples with annotation guidelines of varying complexity, and evaluated their performance via three prominent LLMs: GPT-4o, Claude 3.5 Sonnet, and gpt-oss-120b. Additionally, we employed three diverse medical corpora: i2b2-2014, i2b2-2012, and MedTxt-CR. Accuracy was assessed via precision, recall, and the F1 score, with evaluation methods aligned with those used in relevant shared tasks to ensure the comparability of the results.
Results
Our findings indicate that adding detailed annotation guidelines to few-shot prompts improves the recall and F1 score in most cases.
Conclusion
Including annotation guidelines in prompts enhances the performance of LLMs in NER tasks, making this a practical approach for developing accurate NLP systems in resource-constrained environments. Although annotation guidelines are essential for evaluation and example creation, their integration into LLM prompts can further optimize few-shot learning, especially within specialized domains such as medical NLP.
{"title":"Efficient medical NER with limited data: Enhancing LLM performance through annotation guidelines","authors":"Emiko Shinohara, Yoshimasa Kawazoe","doi":"10.1016/j.ijmedinf.2025.106230","DOIUrl":"10.1016/j.ijmedinf.2025.106230","url":null,"abstract":"<div><h3>Background</h3><div>Named entity recognition (NER) is critical in natural language processing (NLP), particularly in the medical field, where accurate identification of entities, such as patient information and clinical events, is essential. Traditional NER approaches rely heavily on large, annotated corpora, which are resource intensive. Large language models (LLMs) offer new NER approaches, particularly through in-context and few-shot learning.</div></div><div><h3>Objective</h3><div>This study investigates the effects of incorporating annotation guidelines into prompts for NER via LLMs, with a specific focus on their impact on few-shot learning performance across various medical corpora.</div></div><div><h3>Methods</h3><div>We designed eight different prompt patterns, combining few-shot examples with annotation guidelines of varying complexity, and evaluated their performance via three prominent LLMs: GPT-4o, Claude 3.5 Sonnet, and gpt-oss-120b. Additionally, we employed three diverse medical corpora: i2b2-2014, i2b2-2012, and MedTxt-CR. Accuracy was assessed via precision, recall, and the F1 score, with evaluation methods aligned with those used in relevant shared tasks to ensure the comparability of the results.</div></div><div><h3>Results</h3><div>Our findings indicate that adding detailed annotation guidelines to few-shot prompts improves the recall and F1 score in most cases.</div></div><div><h3>Conclusion</h3><div>Including annotation guidelines in prompts enhances the performance of LLMs in NER tasks, making this a practical approach for developing accurate NLP systems in resource-constrained environments. Although annotation guidelines are essential for evaluation and example creation, their integration into LLM prompts can further optimize few-shot learning, especially within specialized domains such as medical NLP.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"208 ","pages":"Article 106230"},"PeriodicalIF":4.1,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145821980","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Radiology faces an unprecedented workload crisis, creating demand for AI solutions to enhance efficiency and quality. Vision-language models (VLMs) represent a paradigm shift from narrow AI tools to integrated systems for image interpretation and report generation. However, their rapid technical progress has outpaced rigorous clinical validation, creating a critical gap between their theoretical potential and safe, practical deployment.
Objective
To critically review the state of VLMs in diagnostic imaging by evaluating their clinical validation, identifying deployment challenges, and assessing their impact on the radiological workflow. This review provides a roadmap for responsible clinical integration by analyzing the gap between model performance and real-world utility.
Method
A narrative review of literature was conducted from January 2017 to May 2025. The search focused on VLM applications in radiology, including automated report generation and visual question answering. We synthesized findings from technical and clinical validation studies, thematically organized around architectural evolution, applications, validation, and implementation barriers.
Results
A clear progression from encoder-decoder models to sophisticated LLM-integrated foundation models was identified. While these models achieve high performance on NLP metrics, their clinical utility is limited. Key findings include: (1) Pervasive model hallucination, with factual errors in ∼ 22 % of AI-generated reports; (2) A lack of external validation on diverse, multi-institutional datasets; (3) Significant implementation barriers, including high computational costs, poor workflow integration, and unresolved liability. Human expert evaluations show that while AI-generated reports for routine cases are often acceptable (77.7 % in one study), accuracy declines significantly in complex cases.
Conclusion
VLMs hold transformative potential but are not ready for autonomous clinical use. Their primary value lies in augmenting radiologists’ workflow. For successful adoption, the field must shift focus from algorithmic metrics to proving clinical safety and efficacy through rigorous validation, developing robust hallucination mitigation strategies, and designing seamless workflow integrations.
{"title":"Vision-language models in diagnostic imaging: review of technical advances, clinical validation, and practical deployment","authors":"Niharika Dutta , Kartik Bose , Emir Syailendra , Linda Chu , Pankaj Gupta","doi":"10.1016/j.ijmedinf.2025.106227","DOIUrl":"10.1016/j.ijmedinf.2025.106227","url":null,"abstract":"<div><h3>Background</h3><div>Radiology faces an unprecedented workload crisis, creating demand for AI solutions to enhance efficiency and quality. Vision-language models (VLMs) represent a paradigm shift from narrow AI tools to integrated systems for image interpretation and report generation. However, their rapid technical progress has outpaced rigorous clinical validation, creating a critical gap between their theoretical potential and safe, practical deployment.</div></div><div><h3>Objective</h3><div>To critically review the state of VLMs in diagnostic imaging by evaluating their clinical validation, identifying deployment challenges, and assessing their impact on the radiological workflow. This review provides a roadmap for responsible clinical integration by analyzing the gap between model performance and real-world utility.</div></div><div><h3>Method</h3><div>A narrative review of literature was conducted from January 2017 to May 2025. The search focused on VLM applications in radiology, including automated report generation and visual question answering. We synthesized findings from technical and clinical validation studies, thematically organized around architectural evolution, applications, validation, and implementation barriers.</div></div><div><h3>Results</h3><div>A clear progression from encoder-decoder models to sophisticated LLM-integrated foundation models was identified. While these models achieve high performance on NLP metrics, their clinical utility is limited. Key findings include: (1) Pervasive model hallucination, with factual errors in ∼ 22 % of AI-generated reports; (2) A lack of external validation on diverse, multi-institutional datasets; (3) Significant implementation barriers, including high computational costs, poor workflow integration, and unresolved liability. Human expert evaluations show that while AI-generated reports for routine cases are often acceptable (77.7 % in one study), accuracy declines significantly in complex cases.</div></div><div><h3>Conclusion</h3><div>VLMs hold transformative potential but are not ready for autonomous clinical use. Their primary value lies in augmenting radiologists’ workflow. For successful adoption, the field must shift focus from algorithmic metrics to proving clinical safety and efficacy through rigorous validation, developing robust hallucination mitigation strategies, and designing seamless workflow integrations.</div></div>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"208 ","pages":"Article 106227"},"PeriodicalIF":4.1,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145885679","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-15Epub Date: 2025-12-25DOI: 10.1016/j.ijmedinf.2025.106242
Janne Kommusaar , Silja Elunurm , Taridzo Chomutare , Mari Kangasniemi , Sanna Salanterä , Laura-Maria Peltonen
<div><h3>Background</h3><div>The rise of digital health data has expanded opportunities for data-driven innovation, yet privacy, legal and ethical barriers frame data sharing and collaborative artificial intelligence development. Federated Learning (FL) offers a privacy-preserving alternative, but current research considers mainly technical aspects. There is no end-to-end roadmap that integrates ethical, legal, technical and administrative principles tailored to FL projects in healthcare. This study addresses that gap by developing a roadmap to guide responsible and scalable FL research in the European context.</div></div><div><h3>Methods</h3><div>A multi-method participatory approach was used to develop a roadmap for scientific projects using FL on health data. The iterative process involved three phases. First, key questions were defined and existing evidence was explored through (i) a survey of domain experts (researchers, data governance specialists and infrastructure providers), (ii) a targeted literature review of FL applications in health research and (iii) systematic mapping of relevant EU-level legislation and policy frameworks. Evidence from these sources was synthesized to identify technical, organizational, legal and sustainability-related requirements for FL-based research. Second, preliminary roadmap components were refined through stakeholder engagement in an online workshop, where feasibility, scalability and sustainability considerations were explicitly discussed. Third, the roadmap was validated and iteratively refined by an expert panel through a structured group discussion, focusing on long-term sustainability, governance and transferability across research contexts. The process was carried out within a Baltic-Nordic collaboration in 2023–2025.</div></div><div><h3>Results</h3><div>The developed roadmap integrates ethical, legal, technical, administrative and sustainability-related considerations essential for applying FL to health data. It emphasizes the importance of multidisciplinary collaboration throughout the FL project lifecycle, with particular attention to long-term governance, scalability and reuse of infrastructures and practices. The process is structured into six phases: (1) Planning, (2) Execution refinement, (3) Data, (4) FL platform, (5) FL experiment and (6) Dissemination. Across these phases, sustainability is addressed through mechanisms such as regulatory alignment, shared governance models, capacity building and integration with existing research and health data infrastructures. By merging ethical, legal, technical and administrative aspects into a unified, end-to-end framework, the roadmap provides actionable, novel guidance beyond existing recommendations.</div></div><div><h3>Conclusions</h3><div>This work consolidates early lessons from FL in healthcare into a practical, step-by-step roadmap that integrates ethical, legal, technical and administrative aspects in the European context. By offering a shared
{"title":"A roadmap for federated learning projects using health data to guide sustainable artificial intelligence development in the European Union","authors":"Janne Kommusaar , Silja Elunurm , Taridzo Chomutare , Mari Kangasniemi , Sanna Salanterä , Laura-Maria Peltonen","doi":"10.1016/j.ijmedinf.2025.106242","DOIUrl":"10.1016/j.ijmedinf.2025.106242","url":null,"abstract":"<div><h3>Background</h3><div>The rise of digital health data has expanded opportunities for data-driven innovation, yet privacy, legal and ethical barriers frame data sharing and collaborative artificial intelligence development. Federated Learning (FL) offers a privacy-preserving alternative, but current research considers mainly technical aspects. There is no end-to-end roadmap that integrates ethical, legal, technical and administrative principles tailored to FL projects in healthcare. This study addresses that gap by developing a roadmap to guide responsible and scalable FL research in the European context.</div></div><div><h3>Methods</h3><div>A multi-method participatory approach was used to develop a roadmap for scientific projects using FL on health data. The iterative process involved three phases. First, key questions were defined and existing evidence was explored through (i) a survey of domain experts (researchers, data governance specialists and infrastructure providers), (ii) a targeted literature review of FL applications in health research and (iii) systematic mapping of relevant EU-level legislation and policy frameworks. Evidence from these sources was synthesized to identify technical, organizational, legal and sustainability-related requirements for FL-based research. Second, preliminary roadmap components were refined through stakeholder engagement in an online workshop, where feasibility, scalability and sustainability considerations were explicitly discussed. Third, the roadmap was validated and iteratively refined by an expert panel through a structured group discussion, focusing on long-term sustainability, governance and transferability across research contexts. The process was carried out within a Baltic-Nordic collaboration in 2023–2025.</div></div><div><h3>Results</h3><div>The developed roadmap integrates ethical, legal, technical, administrative and sustainability-related considerations essential for applying FL to health data. It emphasizes the importance of multidisciplinary collaboration throughout the FL project lifecycle, with particular attention to long-term governance, scalability and reuse of infrastructures and practices. The process is structured into six phases: (1) Planning, (2) Execution refinement, (3) Data, (4) FL platform, (5) FL experiment and (6) Dissemination. Across these phases, sustainability is addressed through mechanisms such as regulatory alignment, shared governance models, capacity building and integration with existing research and health data infrastructures. By merging ethical, legal, technical and administrative aspects into a unified, end-to-end framework, the roadmap provides actionable, novel guidance beyond existing recommendations.</div></div><div><h3>Conclusions</h3><div>This work consolidates early lessons from FL in healthcare into a practical, step-by-step roadmap that integrates ethical, legal, technical and administrative aspects in the European context. By offering a shared","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"208 ","pages":"Article 106242"},"PeriodicalIF":4.1,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145879368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-14DOI: 10.1016/j.ijmedinf.2026.106390
Peter B Hjort, Jacob E Jensen, Jørgen B Jensen, Andreas Ernst
Background: Diagnosis and surveillance of bladder cancer rely on white-light cystoscopy (WLC). However, this modality is operator-dependent and associated with a risk of missed lesions, contributing to high recurrence rates, especially in non-muscle invasive bladder cancer. Recent advances in artificial intelligence (AI) enable software-based decision support for bladder lesion detection, with potential for vendor-independent deployment and broad integration into routine clinical workflows.
Objective: To develop and externally validate an AI-based clinical decision support system for real-time bladder lesion detection during cystoscopy.
Methods: CystoAID, a convolutional neural network-based object detection system, was trained on prospectively collected video recordings from flexible cystoscopies and transurethral resections of bladder tumors. Diagnostic accuracy was evaluated using a retrospective external validation dataset representative of routine clinical practice, in accordance with STARD-AI recommendations.
Results: In the external validation cohort, CystoAID achieved a sensitivity of 1.00 (95% CI 0.95-1.00). Precision was 88.1% (95% CI 81.3-92.7), exceeding published estimates for WLC. Precision-recall analysis showed consistently high precision (>0.8) across clinically relevant recall levels, with declining precision at higher recall, reflecting the expected trade-off between sensitivity and false-positive detections. The system operated with low processing latency, supporting feasibility for real-time clinical use. Sensitivity was prioritized to mitigate the clinical risk associated with false-negative findings.
Conclusions: CystoAID is a real-time, AI-based decision support tool for cystoscopy that demonstrated high sensitivity and favorable precision in external validation. These findings support its potential role as an assistive technology in routine urologic practice. Prospective studies are warranted to evaluate clinical impact, workflow integration, and performance in detecting challenging lesion subtypes, including flat lesions and carcinoma in situ.
背景：膀胱癌的诊断和监测依赖于白光膀胱镜检查。然而，这种手术方式依赖于手术者，并且与漏诊的风险相关，导致高复发率，特别是在非肌肉浸润性膀胱癌中。人工智能（AI）的最新进展使基于软件的膀胱病变检测决策支持成为可能，具有独立于供应商的部署和广泛集成到常规临床工作流程中的潜力。目的：开发并外部验证基于人工智能的膀胱镜下病变实时检测临床决策支持系统。方法：采用基于卷积神经网络的目标检测系统CystoAID，对前瞻性收集的柔性膀胱镜和经尿道膀胱肿瘤切除术的视频记录进行训练。根据star - ai建议，使用代表常规临床实践的回顾性外部验证数据集评估诊断准确性。结果：在外部验证队列中，CystoAID的敏感性为1.00 （95% CI 0.95-1.00）。精密度为88.1% (95% CI 81.3-92.7)，超过了已发表的WLC估计值。查准率-查全率分析显示，在临床相关的查全率水平上，准确率始终较高（>.8），查全率越高，准确率越低，反映了灵敏度和假阳性检测之间的预期权衡。该系统运行时处理延迟低，支持实时临床应用的可行性。敏感性被优先考虑，以减轻与假阴性结果相关的临床风险。结论：CystoAID是一种基于人工智能的实时膀胱镜决策支持工具，在外部验证中具有较高的灵敏度和良好的精度。这些发现支持其在泌尿外科常规实践中作为辅助技术的潜在作用。有必要进行前瞻性研究，以评估临床影响、工作流程整合以及检测具有挑战性的病变亚型（包括扁平病变和原位癌）的性能。
{"title":"Study of bladder cancer detection in standard white light versus AI-supported endoscopy-01 (RAISE-01) - Development and validation of an AI-based support tool.","authors":"Peter B Hjort, Jacob E Jensen, Jørgen B Jensen, Andreas Ernst","doi":"10.1016/j.ijmedinf.2026.106390","DOIUrl":"https://doi.org/10.1016/j.ijmedinf.2026.106390","url":null,"abstract":"<p><strong>Background: </strong>Diagnosis and surveillance of bladder cancer rely on white-light cystoscopy (WLC). However, this modality is operator-dependent and associated with a risk of missed lesions, contributing to high recurrence rates, especially in non-muscle invasive bladder cancer. Recent advances in artificial intelligence (AI) enable software-based decision support for bladder lesion detection, with potential for vendor-independent deployment and broad integration into routine clinical workflows.</p><p><strong>Objective: </strong>To develop and externally validate an AI-based clinical decision support system for real-time bladder lesion detection during cystoscopy.</p><p><strong>Methods: </strong>CystoAID, a convolutional neural network-based object detection system, was trained on prospectively collected video recordings from flexible cystoscopies and transurethral resections of bladder tumors. Diagnostic accuracy was evaluated using a retrospective external validation dataset representative of routine clinical practice, in accordance with STARD-AI recommendations.</p><p><strong>Results: </strong>In the external validation cohort, CystoAID achieved a sensitivity of 1.00 (95% CI 0.95-1.00). Precision was 88.1% (95% CI 81.3-92.7), exceeding published estimates for WLC. Precision-recall analysis showed consistently high precision (>0.8) across clinically relevant recall levels, with declining precision at higher recall, reflecting the expected trade-off between sensitivity and false-positive detections. The system operated with low processing latency, supporting feasibility for real-time clinical use. Sensitivity was prioritized to mitigate the clinical risk associated with false-negative findings.</p><p><strong>Conclusions: </strong>CystoAID is a real-time, AI-based decision support tool for cystoscopy that demonstrated high sensitivity and favorable precision in external validation. These findings support its potential role as an assistive technology in routine urologic practice. Prospective studies are warranted to evaluate clinical impact, workflow integration, and performance in detecting challenging lesion subtypes, including flat lesions and carcinoma in situ.</p>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"213 ","pages":"106390"},"PeriodicalIF":4.1,"publicationDate":"2026-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147464393","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-13DOI: 10.1016/j.ijmedinf.2026.106384
Gerard Chung Siew Keong
Objectives: This study evaluates privacy policies and terms of service agreements from digital mental health platforms, focusing on accessibility, comprehensibility, and alignment with informed consent principles in healthcare informatics.
Materials and methods: We applied mixed methods combining content analysis and computational linguistic assessment to 139 user agreements from international mental health applications and Singaporean providers, including commercial platforms and social service agencies serving vulnerable populations. We evaluated readability, communicative practices, regulatory compliance, and power asymmetries. Only 1.67% of services implemented comprehension verification for informed consent. User agreements required approximately 16 years of education for comprehension and exhibited significant linguistic power asymmetries favoring providers. Privacy policies comprehensively addressed data collection but systematically neglected post-service communication regarding data retention and deletion. Among local services, only 8.33% adequately communicated data breach notification procedures as required by Singapore's Personal Data Protection Act. Terms of service failed to establish bidirectional communicative exchange necessary for meaningful healthcare informed consent. Findings reveal fundamental misalignment between digital mental health agreements and collaborative communication principles essential to therapeutic relationships and healthcare informatics best practices. Communication barriers pose particular risks for individuals with serious mental illness requiring accessible health information for decision-making. Results have implications for health informatics policy, consumer health technology design, and digital health regulatory frameworks. Digital mental health platforms demonstrate significant user communication deficiencies. Our findings point to the need for user agreements that are written in plain language, that incorporate essential informed consent components, that balance linguistic power between providers and users, and that accommodate the cognitive needs of vulnerable populations seeking mental health support.
{"title":"\"Clicking without understanding\": A mixed-methods analysis of user agreements in digital mental health services.","authors":"Gerard Chung Siew Keong","doi":"10.1016/j.ijmedinf.2026.106384","DOIUrl":"https://doi.org/10.1016/j.ijmedinf.2026.106384","url":null,"abstract":"<p><strong>Objectives: </strong>This study evaluates privacy policies and terms of service agreements from digital mental health platforms, focusing on accessibility, comprehensibility, and alignment with informed consent principles in healthcare informatics.</p><p><strong>Materials and methods: </strong>We applied mixed methods combining content analysis and computational linguistic assessment to 139 user agreements from international mental health applications and Singaporean providers, including commercial platforms and social service agencies serving vulnerable populations. We evaluated readability, communicative practices, regulatory compliance, and power asymmetries. Only 1.67% of services implemented comprehension verification for informed consent. User agreements required approximately 16 years of education for comprehension and exhibited significant linguistic power asymmetries favoring providers. Privacy policies comprehensively addressed data collection but systematically neglected post-service communication regarding data retention and deletion. Among local services, only 8.33% adequately communicated data breach notification procedures as required by Singapore's Personal Data Protection Act. Terms of service failed to establish bidirectional communicative exchange necessary for meaningful healthcare informed consent. Findings reveal fundamental misalignment between digital mental health agreements and collaborative communication principles essential to therapeutic relationships and healthcare informatics best practices. Communication barriers pose particular risks for individuals with serious mental illness requiring accessible health information for decision-making. Results have implications for health informatics policy, consumer health technology design, and digital health regulatory frameworks. Digital mental health platforms demonstrate significant user communication deficiencies. Our findings point to the need for user agreements that are written in plain language, that incorporate essential informed consent components, that balance linguistic power between providers and users, and that accommodate the cognitive needs of vulnerable populations seeking mental health support.</p>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"213 ","pages":"106384"},"PeriodicalIF":4.1,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147464283","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-13DOI: 10.1016/j.ijmedinf.2026.106398
Lo Conte Sofia, Bufano Annalisa, Cevenini Gabriele, Barbini Paolo, Castagna Maria Grazia, Cartocci Alessandra
Background: Sarcopenia is characterized by progressive loss of skeletal muscle mass and strength and is associated with increased disability and mortality. However, the diagnosis of sarcopenia remains challenging due to the absence of a universally accepted gold standard and validated cut-off values for skeletal muscle indices. Data-driven approaches based on unsupervised clustering may overcome these limitations by identifying muscle-related phenotypes directly from anthropometric and body composition data.
Methods: In this study, 600 adults with obesity were analyzed and stratified by sex. The dataset was randomly divided into a training set (80%) and a testing set (20%). After data standardization, principal component analysis (PCA) was applied separately in males and females. Unsupervised clustering was then performed on the preserved principal components, and the optimal number of clusters was determined using internal validation indices. Linear Discriminant Analysis (LDA) was applied to assign patients in the test set, and posterior probabilities were correlated with Skeletal Muscle Index (SMI).
Results: Clustering consistently identified two distinct groups in both sexes: one with higher SMI and another with lower SMI, consistent with reduced muscle status. Stepwise LDA accurately classified individuals, and posterior probabilities of belonging to the pathological cluster were negatively correlated with SMI in both sexes, despite SMI not being used in clustering or classification. Individuals in the pathological group exhibited significantly lower SMI, particularly among females.
Conclusions: The combined use of unsupervised clustering and LDA allows reliable identification of distinct muscle-related phenotypes in adults with obesity. This framework provides reproducible classifications, correlates with skeletal muscle index, and offers a quantitative approach to stratify patients by muscle status, even in the absence of predefined diagnostic criteria. These findings support the potential of data-driven phenotyping to improve early detection of sarcopenic obesity.
{"title":"Unsupervised identification of muscle phenotypes in adults with obesity: a data-driven framework for the identification of sarcopenia in absence of a gold standard.","authors":"Lo Conte Sofia, Bufano Annalisa, Cevenini Gabriele, Barbini Paolo, Castagna Maria Grazia, Cartocci Alessandra","doi":"10.1016/j.ijmedinf.2026.106398","DOIUrl":"https://doi.org/10.1016/j.ijmedinf.2026.106398","url":null,"abstract":"<p><strong>Background: </strong>Sarcopenia is characterized by progressive loss of skeletal muscle mass and strength and is associated with increased disability and mortality. However, the diagnosis of sarcopenia remains challenging due to the absence of a universally accepted gold standard and validated cut-off values for skeletal muscle indices. Data-driven approaches based on unsupervised clustering may overcome these limitations by identifying muscle-related phenotypes directly from anthropometric and body composition data.</p><p><strong>Methods: </strong>In this study, 600 adults with obesity were analyzed and stratified by sex. The dataset was randomly divided into a training set (80%) and a testing set (20%). After data standardization, principal component analysis (PCA) was applied separately in males and females. Unsupervised clustering was then performed on the preserved principal components, and the optimal number of clusters was determined using internal validation indices. Linear Discriminant Analysis (LDA) was applied to assign patients in the test set, and posterior probabilities were correlated with Skeletal Muscle Index (SMI).</p><p><strong>Results: </strong>Clustering consistently identified two distinct groups in both sexes: one with higher SMI and another with lower SMI, consistent with reduced muscle status. Stepwise LDA accurately classified individuals, and posterior probabilities of belonging to the pathological cluster were negatively correlated with SMI in both sexes, despite SMI not being used in clustering or classification. Individuals in the pathological group exhibited significantly lower SMI, particularly among females.</p><p><strong>Conclusions: </strong>The combined use of unsupervised clustering and LDA allows reliable identification of distinct muscle-related phenotypes in adults with obesity. This framework provides reproducible classifications, correlates with skeletal muscle index, and offers a quantitative approach to stratify patients by muscle status, even in the absence of predefined diagnostic criteria. These findings support the potential of data-driven phenotyping to improve early detection of sarcopenic obesity.</p>","PeriodicalId":54950,"journal":{"name":"International Journal of Medical Informatics","volume":"213 ","pages":"106398"},"PeriodicalIF":4.1,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147476218","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号