Pub Date : 2026-03-06DOI: 10.1038/s41746-026-02461-4
Isabel Voigt,Lars Masanneck,Marc Pawlitzki,Hernan Inojosa,Sven G Meuth,Tjalf Ziemssen
This perspective introduces MS360°, a conceptual hybrid care model for the management of multiple sclerosis (MS). It integrates traditional on-site assessments with digital health technologies (DHT) to enable more continuous, personalised, and proactive disease management. Current MS care is often fragmented, limiting timely interventions and patient engagement. MS360° addresses these challenges by introducing a digital-first hybrid framework for continuous data collection through remote monitoring, wearable sensors, and telemedicine. This data can be used to dynamically steer structured patient pathways and trigger targeted on-site assessments and interventions such as neurological examinations, imaging, laboratory assessments, and standardised functional tests based on predefined thresholds and patient profiles. The interaction of multidisciplinary teams, structured care pathways and bidirectional data flow enables timely clinical decision-making, stratified patient management and early detection of disease progression. Digital tools can further enhance patient engagement and lifestyle management, promoting adherence and outcomes. New technologies, including artificial intelligence and digital twins, are being discussed as potential future extensions for precision care, workflow optimisation, and risk prediction. MS360° provides a quality-driven conceptual framework, offering a roadmap for integrating digital innovations into patient-centred MS care.
{"title":"MS360°: a conceptual digital-first, data-driven hybrid care framework for personalised multiple sclerosis management.","authors":"Isabel Voigt,Lars Masanneck,Marc Pawlitzki,Hernan Inojosa,Sven G Meuth,Tjalf Ziemssen","doi":"10.1038/s41746-026-02461-4","DOIUrl":"https://doi.org/10.1038/s41746-026-02461-4","url":null,"abstract":"This perspective introduces MS360°, a conceptual hybrid care model for the management of multiple sclerosis (MS). It integrates traditional on-site assessments with digital health technologies (DHT) to enable more continuous, personalised, and proactive disease management. Current MS care is often fragmented, limiting timely interventions and patient engagement. MS360° addresses these challenges by introducing a digital-first hybrid framework for continuous data collection through remote monitoring, wearable sensors, and telemedicine. This data can be used to dynamically steer structured patient pathways and trigger targeted on-site assessments and interventions such as neurological examinations, imaging, laboratory assessments, and standardised functional tests based on predefined thresholds and patient profiles. The interaction of multidisciplinary teams, structured care pathways and bidirectional data flow enables timely clinical decision-making, stratified patient management and early detection of disease progression. Digital tools can further enhance patient engagement and lifestyle management, promoting adherence and outcomes. New technologies, including artificial intelligence and digital twins, are being discussed as potential future extensions for precision care, workflow optimisation, and risk prediction. MS360° provides a quality-driven conceptual framework, offering a roadmap for integrating digital innovations into patient-centred MS care.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"1 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147359263","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Digital decision aids significantly improve shared decision-making outcomes, but barriers to implementation in clinical settings remain. We conducted a Hybrid Type 2 implementation-effectiveness trial of an atrial fibrillation rhythm control decision aid (clinicaltrials.gov NCT04993807; registered 08/06/2021) among 75 older adults across two sites. Guided by the RE-AIM framework, we assessed decision quality and implementation outcomes. While the decision aid was highly acceptable and broadly adopted, changes in decisional conflict and self-efficacy varied widely, with no significant average improvement across the cohort. Subgroup and qualitative analyses revealed that the decision aid was most effective when delivered to the right patient, at the right time, and in the right clinical context. Barriers included variability in health literacy, digital access, and timing of delivery relative to the clinical decision-making process. Findings underscore the challenges of deploying digital interventions within real-world workflows and highlight the importance of targeting decision support tools based on patient readiness, literacy, and care context.
{"title":"Evaluating a digital decision aid for atrial fibrillation rhythm control in a hybrid implementation-effectiveness trial.","authors":"Meghan Reading Turchioe,Afra Shamnath,David Slotwiner,Yihong Zhao,Deepak Saluja,Seth Goldbarg,JoonHyuk Kim,Paul Varosy,Angelo Biviano","doi":"10.1038/s41746-026-02405-y","DOIUrl":"https://doi.org/10.1038/s41746-026-02405-y","url":null,"abstract":"Digital decision aids significantly improve shared decision-making outcomes, but barriers to implementation in clinical settings remain. We conducted a Hybrid Type 2 implementation-effectiveness trial of an atrial fibrillation rhythm control decision aid (clinicaltrials.gov NCT04993807; registered 08/06/2021) among 75 older adults across two sites. Guided by the RE-AIM framework, we assessed decision quality and implementation outcomes. While the decision aid was highly acceptable and broadly adopted, changes in decisional conflict and self-efficacy varied widely, with no significant average improvement across the cohort. Subgroup and qualitative analyses revealed that the decision aid was most effective when delivered to the right patient, at the right time, and in the right clinical context. Barriers included variability in health literacy, digital access, and timing of delivery relative to the clinical decision-making process. Findings underscore the challenges of deploying digital interventions within real-world workflows and highlight the importance of targeting decision support tools based on patient readiness, literacy, and care context.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"225 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147368378","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-05DOI: 10.1038/s41746-026-02471-2
Amanda Dy, Sandra M. Buetow, Andrew J. Bredemeyer, Monika Lamba Saini, Fabienne Lucas, Shannon Bennett, Kim R. M. Blenman, Keith Wharton Jr., Sunil Singhal, M. E. de Baca, Kevin Schap, Matthew G. Hanna, Staci J. Kearney, Norman Zerbe, Roberto Salgado, Jithesh Veetil, Jansen N. Seheult, David S. McClintock, April Khademi, Jochen K. Lennerz
Validation is a cornerstone of reliability and trust in diagnostics, yet discipline-specific assumptions and unspoken contextual differences often lead to miscommunication, misalignment, and avoidable delays. As AI/ML becomes more integrated into healthcare, there is a growing necessity to re-examine how the term validation is used and understood. We highlight inconsistencies in the use of the term validation through an analysis of 94 themes across five domains, including Communication Science (n = 12), AI/ML (n = 26), Clinical and Laboratory Practice (n = 19), Regulatory Science (n = 22), and Business (n = 15). We emphasize how persistent reliance on domain-specific implied definitions impedes interdisciplinary alignment. Rather than advocating for a single definition, we derived five consensus proposals that collectively advocate for more specific and context-aware additions to the term validation to support clarity, reliability, and compliance across disciplines. Our goal is to support clearer communication and provide useful strategies that inform the development, regulation, and use of digital health technologies.
{"title":"Clarifying validation terminologies in healthcare","authors":"Amanda Dy, Sandra M. Buetow, Andrew J. Bredemeyer, Monika Lamba Saini, Fabienne Lucas, Shannon Bennett, Kim R. M. Blenman, Keith Wharton Jr., Sunil Singhal, M. E. de Baca, Kevin Schap, Matthew G. Hanna, Staci J. Kearney, Norman Zerbe, Roberto Salgado, Jithesh Veetil, Jansen N. Seheult, David S. McClintock, April Khademi, Jochen K. Lennerz","doi":"10.1038/s41746-026-02471-2","DOIUrl":"https://doi.org/10.1038/s41746-026-02471-2","url":null,"abstract":"Validation is a cornerstone of reliability and trust in diagnostics, yet discipline-specific assumptions and unspoken contextual differences often lead to miscommunication, misalignment, and avoidable delays. As AI/ML becomes more integrated into healthcare, there is a growing necessity to re-examine how the term validation is used and understood. We highlight inconsistencies in the use of the term validation through an analysis of 94 themes across five domains, including Communication Science (n = 12), AI/ML (n = 26), Clinical and Laboratory Practice (n = 19), Regulatory Science (n = 22), and Business (n = 15). We emphasize how persistent reliance on domain-specific implied definitions impedes interdisciplinary alignment. Rather than advocating for a single definition, we derived five consensus proposals that collectively advocate for more specific and context-aware additions to the term validation to support clarity, reliability, and compliance across disciplines. Our goal is to support clearer communication and provide useful strategies that inform the development, regulation, and use of digital health technologies.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"61 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147350655","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-05DOI: 10.1038/s41746-026-02511-x
Hwan-ho Cho, Joonwon Lee, Jeonghoon Bae, Dongwhane Lee, Hyung Chan Kim, Suk Yoon Lee, Jung Hwa Seo, Woo-Keun Seo, Jin-Man Jung, Hyunjin Park, Seongho Park
We developed and externally validated a deep learning model to automatically detect new ischemic lesions on serial FLAIR MRI scans in patients with stroke. Manual interpretation of follow-up imaging is labor-intensive and variable, and silent brain infarctions (SBIs) are frequently missed despite their prognostic importance. Using 25,451 paired slices from 1055 patients across two hospitals, we trained a convolutional neural network with supervised contrastive learning to classify new lesion occurrence. The model achieved an area under the receiver operating characteristic curve of 0.89 in both internal and external validation cohorts. To evaluate clinical relevance, we further analyzed an independent asymptomatic cohort of 307 patients with a median follow-up of two years. Patients classified as SBI-positive by the model showed a significantly higher risk of subsequent symptomatic stroke than those without SBI. In multivariable Cox regression adjusted for age and major vascular risk factors, model-positive patients had a 3.8-fold increased risk of stroke recurrence. These findings indicate that AI can identify clinically meaningful SBIs that are under-recognized in routine practice and independently associated with stroke recurrence. Automated lesion detection may provide a reproducible imaging biomarker for risk stratification, supporting standardized interpretation of follow-up MRI and informing secondary stroke prevention strategies.
{"title":"Automated detection of new cerebral infarctions and prognostic implications using deep learning on serial MRI","authors":"Hwan-ho Cho, Joonwon Lee, Jeonghoon Bae, Dongwhane Lee, Hyung Chan Kim, Suk Yoon Lee, Jung Hwa Seo, Woo-Keun Seo, Jin-Man Jung, Hyunjin Park, Seongho Park","doi":"10.1038/s41746-026-02511-x","DOIUrl":"https://doi.org/10.1038/s41746-026-02511-x","url":null,"abstract":"We developed and externally validated a deep learning model to automatically detect new ischemic lesions on serial FLAIR MRI scans in patients with stroke. Manual interpretation of follow-up imaging is labor-intensive and variable, and silent brain infarctions (SBIs) are frequently missed despite their prognostic importance. Using 25,451 paired slices from 1055 patients across two hospitals, we trained a convolutional neural network with supervised contrastive learning to classify new lesion occurrence. The model achieved an area under the receiver operating characteristic curve of 0.89 in both internal and external validation cohorts. To evaluate clinical relevance, we further analyzed an independent asymptomatic cohort of 307 patients with a median follow-up of two years. Patients classified as SBI-positive by the model showed a significantly higher risk of subsequent symptomatic stroke than those without SBI. In multivariable Cox regression adjusted for age and major vascular risk factors, model-positive patients had a 3.8-fold increased risk of stroke recurrence. These findings indicate that AI can identify clinically meaningful SBIs that are under-recognized in routine practice and independently associated with stroke recurrence. Automated lesion detection may provide a reproducible imaging biomarker for risk stratification, supporting standardized interpretation of follow-up MRI and informing secondary stroke prevention strategies.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"27 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147350611","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-05DOI: 10.1038/s41746-026-02496-7
Raphael Judkiewicz, Eran Berkowitz, Meishar Meisel, Tomer Michaeli, Joachim A. Behar
Optical Coherence Tomography (OCT) is essential in ophthalmology for cross-sectional imaging of the retina. Pretrained foundation models facilitate task-specific model development by enabling fine-tuning with limited labeled data. However, current foundation models rely on a single B-scan (usually the central slice), overlooking volumetric context. This research investigates video foundation models to capture full 3D retinal structure and improve diagnostic performance. V-JEPA, a state-of-the-art video foundation model, was benchmarked against retinal foundation models (RETFound, VisionFM) and a natural image foundation model (DINOv2). All were fine-tuned to detect Age-related Macular Degeneration or Glaucomatous Optic Neuropathy using five OCT datasets. V-JEPA consistently equaled or outperformed image-based models, achieving an average AUROC of 0.94 (0.80–0.99), versus 0.90 (0.76–0.98) for the best image model, a statistically significant improvement (p < 0.001). To our knowledge, this is the first application of transformer-based video models to volumetric OCT, highlighting their promise in 3D medical imaging.
{"title":"Shifting the retinal foundation models paradigm from slices to volumes for optical coherence tomography","authors":"Raphael Judkiewicz, Eran Berkowitz, Meishar Meisel, Tomer Michaeli, Joachim A. Behar","doi":"10.1038/s41746-026-02496-7","DOIUrl":"https://doi.org/10.1038/s41746-026-02496-7","url":null,"abstract":"Optical Coherence Tomography (OCT) is essential in ophthalmology for cross-sectional imaging of the retina. Pretrained foundation models facilitate task-specific model development by enabling fine-tuning with limited labeled data. However, current foundation models rely on a single B-scan (usually the central slice), overlooking volumetric context. This research investigates video foundation models to capture full 3D retinal structure and improve diagnostic performance. V-JEPA, a state-of-the-art video foundation model, was benchmarked against retinal foundation models (RETFound, VisionFM) and a natural image foundation model (DINOv2). All were fine-tuned to detect Age-related Macular Degeneration or Glaucomatous Optic Neuropathy using five OCT datasets. V-JEPA consistently equaled or outperformed image-based models, achieving an average AUROC of 0.94 (0.80–0.99), versus 0.90 (0.76–0.98) for the best image model, a statistically significant improvement (p < 0.001). To our knowledge, this is the first application of transformer-based video models to volumetric OCT, highlighting their promise in 3D medical imaging.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"130 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147346791","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-05DOI: 10.1038/s41746-026-02460-5
Ariel Leong, Risa M. Wolf, Roomasa Channa, Jiangxia Wang, Harold Lehmann, Michael D. Abramoff, T. Y. Alvin Liu
Adult patients with diabetes (n = 3745) seen at Johns Hopkins Medicine primary care sites were referred to the Wilmer Eye Institute either based on a primary care provider referral or autonomous AI diagnostic result (referral was made after a positive or non-diagnostic result). An inverse-probability-weighted regression, which incorporated propensity score matching on social determinants of health and relevant clinical variables, showed that implementation of an autonomous AI-assisted diabetic screening program in a primary care clinic was associated with increased presentation to eye care specialist care by African-Americans (p = 0.02). This is significant because African-Americans have traditionally been less likely to undergo annual screening exams and more likely to present with more severe forms of diabetic retinopathy (DR). The results suggest a potential association between office-based, AI-assisted DR screening and improved downstream ophthalmic access for African-American patients. However, given that the analysis was exploratory, this association should be interpreted cautiously and further validated.
{"title":"Autonomous AI-assisted diabetic retinopathy screening at primary care is associated with increased presentation to eye care by at risk patients","authors":"Ariel Leong, Risa M. Wolf, Roomasa Channa, Jiangxia Wang, Harold Lehmann, Michael D. Abramoff, T. Y. Alvin Liu","doi":"10.1038/s41746-026-02460-5","DOIUrl":"https://doi.org/10.1038/s41746-026-02460-5","url":null,"abstract":"Adult patients with diabetes (n = 3745) seen at Johns Hopkins Medicine primary care sites were referred to the Wilmer Eye Institute either based on a primary care provider referral or autonomous AI diagnostic result (referral was made after a positive or non-diagnostic result). An inverse-probability-weighted regression, which incorporated propensity score matching on social determinants of health and relevant clinical variables, showed that implementation of an autonomous AI-assisted diabetic screening program in a primary care clinic was associated with increased presentation to eye care specialist care by African-Americans (p = 0.02). This is significant because African-Americans have traditionally been less likely to undergo annual screening exams and more likely to present with more severe forms of diabetic retinopathy (DR). The results suggest a potential association between office-based, AI-assisted DR screening and improved downstream ophthalmic access for African-American patients. However, given that the analysis was exploratory, this association should be interpreted cautiously and further validated.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"11 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147346793","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-05DOI: 10.1038/s41746-026-02502-y
Linea Schmidt,Benedikt Langenberger,Felix Schirmann,Simon Reif,Ariel Dora Stern
Digital therapeutics (DTx) are increasingly established, yet their mechanisms of action (MoA) remain underexplored. This article defines and categorizes DTx MoA through a novel conceptual framework, distinguishing them from conventional treatments. We specifically conceptualize therapeutic elements, cognitive-affective and behavioral changes, and segment the notion of "dose" into distinct categories. This actionable framework provides a systematic basis for enhancing DTx research, design, and clinical effectiveness.
{"title":"Mechanisms of action for digital therapeutics.","authors":"Linea Schmidt,Benedikt Langenberger,Felix Schirmann,Simon Reif,Ariel Dora Stern","doi":"10.1038/s41746-026-02502-y","DOIUrl":"https://doi.org/10.1038/s41746-026-02502-y","url":null,"abstract":"Digital therapeutics (DTx) are increasingly established, yet their mechanisms of action (MoA) remain underexplored. This article defines and categorizes DTx MoA through a novel conceptual framework, distinguishing them from conventional treatments. We specifically conceptualize therapeutic elements, cognitive-affective and behavioral changes, and segment the notion of \"dose\" into distinct categories. This actionable framework provides a systematic basis for enhancing DTx research, design, and clinical effectiveness.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"264 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147359264","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-05DOI: 10.1038/s41746-026-02489-6
Jianming Li, Haoyan Zhang, Huan Zheng, Yuancheng Cang, Lin xue Qian, Ligang Cui, Xinping Wu, Baoding Chen, Man Lu, Yong Xu, Runqin Miao, Desheng Sun, Liping Liu, Ping Li, Changsong Xu, Li Ma, Guoyong Hua, Shengnan Huo, Yanjun Liu, Weide Dai, Kexin Lou, Xiang Xie, Liping Yang, Fang Mei, Bo Ping, Xin Yang, Jie Yu, Kun Wang, Ping Liang
Preoperatively distinguishing follicular thyroid carcinoma (FTC) from follicular thyroid adenoma (FTA) remains a significant clinical challenge. Current ultrasound risk stratification systems show limited efficacy for follicular neoplasms, and existing artificial intelligence (AI) approaches lack sufficient validation. We developed and validated a deep learning model using ultrasound images to differentiate FTC from FTA and classify FTC into invasion subtypes. This multicenter retrospective study incorporated data from 31 hospitals, using 1531 patients for model development and 900 across three external test sets for validation. The model demonstrated high diagnostic performance, with AUCs of 0.816–0.847 for FTC vs FTA discrimination across external test sets and robust performance across subtypes (AUC range 0.754–0.910), and generalized well to varied clinical settings. Triple-classification macro-AUCs were 0.818–0.861. It consistently outperformed radiologists and improved diagnostic accuracy as an assistive tool. Our AI model provides a reliable, non-invasive tool for preoperative diagnosis and risk stratification of follicular thyroid neoplasms.
{"title":"Artificial intelligence-enabled ultrasound diagnosis and stratification of follicular thyroid neoplasms: a multi-center study","authors":"Jianming Li, Haoyan Zhang, Huan Zheng, Yuancheng Cang, Lin xue Qian, Ligang Cui, Xinping Wu, Baoding Chen, Man Lu, Yong Xu, Runqin Miao, Desheng Sun, Liping Liu, Ping Li, Changsong Xu, Li Ma, Guoyong Hua, Shengnan Huo, Yanjun Liu, Weide Dai, Kexin Lou, Xiang Xie, Liping Yang, Fang Mei, Bo Ping, Xin Yang, Jie Yu, Kun Wang, Ping Liang","doi":"10.1038/s41746-026-02489-6","DOIUrl":"https://doi.org/10.1038/s41746-026-02489-6","url":null,"abstract":"Preoperatively distinguishing follicular thyroid carcinoma (FTC) from follicular thyroid adenoma (FTA) remains a significant clinical challenge. Current ultrasound risk stratification systems show limited efficacy for follicular neoplasms, and existing artificial intelligence (AI) approaches lack sufficient validation. We developed and validated a deep learning model using ultrasound images to differentiate FTC from FTA and classify FTC into invasion subtypes. This multicenter retrospective study incorporated data from 31 hospitals, using 1531 patients for model development and 900 across three external test sets for validation. The model demonstrated high diagnostic performance, with AUCs of 0.816–0.847 for FTC vs FTA discrimination across external test sets and robust performance across subtypes (AUC range 0.754–0.910), and generalized well to varied clinical settings. Triple-classification macro-AUCs were 0.818–0.861. It consistently outperformed radiologists and improved diagnostic accuracy as an assistive tool. Our AI model provides a reliable, non-invasive tool for preoperative diagnosis and risk stratification of follicular thyroid neoplasms.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"15 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147346832","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-05DOI: 10.1038/s41746-026-02456-1
Jussi A. Hernesniemi, Teemu Pukkila, Jani Rankinen, Antti Kallonen, Mikko Uimonen, Leo-Pekka Lyytikäinen, Kjell Nikus, Esa Räsänen, Juho Tynkkynen
We evaluated the performance of 12-channel ECG in predicting sudden cardiac death across different time intervals using a retrospective data set of 17,625 high-risk cardiac patients who underwent coronary angiography (2007–2018) with follow-up data until 2022. Extreme gradient boosting using 12SL Marquette software-derived parameters from digital ECG recording was used to train and validate models using a random 80/20 split. Model performance was evaluated in both unbalanced and risk-factor-balanced case-control sets. Using single ECG, both long-term (from baseline ECG) and short-term predictions (from the last recorded ECG) achieved a modest area under the curve (AUC) of 0.68 in the unbalanced validation and 0.59/0.63 in the balanced validation (long-/short-term). Adding clinical risk factor data resulted in AUC 0.70/0.71 (unbalanced) and 0.64/0.62 (balanced) for long- and short-term prediction. Adding data of observed ECG changes during follow-up for short-term prediction resulted in the best model performance (0.72/0.66; unbalanced/balanced).
{"title":"Performance of the 12-lead ECG in predicting short- and long-term risk of sudden cardiac death","authors":"Jussi A. Hernesniemi, Teemu Pukkila, Jani Rankinen, Antti Kallonen, Mikko Uimonen, Leo-Pekka Lyytikäinen, Kjell Nikus, Esa Räsänen, Juho Tynkkynen","doi":"10.1038/s41746-026-02456-1","DOIUrl":"https://doi.org/10.1038/s41746-026-02456-1","url":null,"abstract":"We evaluated the performance of 12-channel ECG in predicting sudden cardiac death across different time intervals using a retrospective data set of 17,625 high-risk cardiac patients who underwent coronary angiography (2007–2018) with follow-up data until 2022. Extreme gradient boosting using 12SL Marquette software-derived parameters from digital ECG recording was used to train and validate models using a random 80/20 split. Model performance was evaluated in both unbalanced and risk-factor-balanced case-control sets. Using single ECG, both long-term (from baseline ECG) and short-term predictions (from the last recorded ECG) achieved a modest area under the curve (AUC) of 0.68 in the unbalanced validation and 0.59/0.63 in the balanced validation (long-/short-term). Adding clinical risk factor data resulted in AUC 0.70/0.71 (unbalanced) and 0.64/0.62 (balanced) for long- and short-term prediction. Adding data of observed ECG changes during follow-up for short-term prediction resulted in the best model performance (0.72/0.66; unbalanced/balanced).","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"52 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147350666","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Cutaneous adverse drug reactions (CADRs) are the most common form of adverse drug reactions, ranging from mild rashes to life-threatening diseases, such as Stevens-Johnson syndrome and toxic epidermal necrolysis. However, there is no effective tool to predict antibiotic-associated CADRs. In this study, we propose an antibiotic-associated CADR prediction model using electronic health record (EHR) foundation models (FMs). EHR FMs are based on the pretraining-finetuning paradigms of language models, corresponding medical codes and their sequences to words and sentences. We included 802,131 inpatients across three tertiary hospitals in Korea, combining EHR data with nursing statements and reports to extract skin rash records. Our approach achieved the best predictive performance compared to all the other baseline models across all datasets. To enhance clinical relevance, we classified CADRs into immediate and delayed types and conducted a detailed sub-analysis. Finally, we found that properly configured EHR FMs can effectively predict the risk of developing antibiotics-associated CADRs, particularly for delayed-type reactions where predictive testing options are limited.
{"title":"Prediction of antibiotic-associated cutaneous adverse drug reactions using electronic health record foundation models.","authors":"Junmo Kim,Kyunghoon Kim,Jeong-Eun Yun,Yu-Kyoung Hwang,Min-Gyu Kang,Seok Kim,Sooyoung Yoo,Chaiho Shin,Suhyun Kim,Kwangsoo Kim,Sae-Hoon Kim","doi":"10.1038/s41746-026-02503-x","DOIUrl":"https://doi.org/10.1038/s41746-026-02503-x","url":null,"abstract":"Cutaneous adverse drug reactions (CADRs) are the most common form of adverse drug reactions, ranging from mild rashes to life-threatening diseases, such as Stevens-Johnson syndrome and toxic epidermal necrolysis. However, there is no effective tool to predict antibiotic-associated CADRs. In this study, we propose an antibiotic-associated CADR prediction model using electronic health record (EHR) foundation models (FMs). EHR FMs are based on the pretraining-finetuning paradigms of language models, corresponding medical codes and their sequences to words and sentences. We included 802,131 inpatients across three tertiary hospitals in Korea, combining EHR data with nursing statements and reports to extract skin rash records. Our approach achieved the best predictive performance compared to all the other baseline models across all datasets. To enhance clinical relevance, we classified CADRs into immediate and delayed types and conducted a detailed sub-analysis. Finally, we found that properly configured EHR FMs can effectively predict the risk of developing antibiotics-associated CADRs, particularly for delayed-type reactions where predictive testing options are limited.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"292 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147346300","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：