Pub Date : 2026-01-15DOI: 10.1038/s41746-025-02061-8
Jiake Wu, Zongyu Wen, Hainan Zhou, Na Sun, Yuanyuan Zhang
Precise delineation and parametric modeling of curvilinear vascular architectures in volumetric medical imaging are pivotal for advancing clinical diagnostics and therapeutic planning. Prevailing methodologies predominantly adopt discrete voxel-wise representations, such as binary masks, which are prone to topological disruptions and artifact-induced fragmentation arising from inherent per-voxel classification biases. To address these challenges, we present FlowAxis, a pioneering continuous parameterization paradigm leveraging Adaptive Vessel Axes (AVA), wherein adaptive keypoints function as interconnected vertices to encapsulate intrinsic spatial topologies. FlowAxis distinguishes itself through superior topological coherence guaranteed by displacement convexity of the energy functional. Comprehensive empirical validations across four benchmark datasets for three-dimensional vascular segmentation substantiate FlowAxis’s performance, achieving significant improvements in both topological accuracy (clDice) and geometric fidelity (Hausdorff distance). Furthermore, qualitative assessments via curved planar reformations highlight its transformative potential in clinical workflows, while theoretical guarantees ensure reliability in safety-critical medical applications. Our work bridges the gap between mathematical rigor and practical medical imaging, providing the first complete theoretical framework for continuous vessel representation with provable optimality and convergence guarantees.
{"title":"Geometric-topological deep transfer learning for precise vessel segmentation in 3D medical volumes","authors":"Jiake Wu, Zongyu Wen, Hainan Zhou, Na Sun, Yuanyuan Zhang","doi":"10.1038/s41746-025-02061-8","DOIUrl":"https://doi.org/10.1038/s41746-025-02061-8","url":null,"abstract":"Precise delineation and parametric modeling of curvilinear vascular architectures in volumetric medical imaging are pivotal for advancing clinical diagnostics and therapeutic planning. Prevailing methodologies predominantly adopt discrete voxel-wise representations, such as binary masks, which are prone to topological disruptions and artifact-induced fragmentation arising from inherent per-voxel classification biases. To address these challenges, we present FlowAxis, a pioneering continuous parameterization paradigm leveraging Adaptive Vessel Axes (AVA), wherein adaptive keypoints function as interconnected vertices to encapsulate intrinsic spatial topologies. FlowAxis distinguishes itself through superior topological coherence guaranteed by displacement convexity of the energy functional. Comprehensive empirical validations across four benchmark datasets for three-dimensional vascular segmentation substantiate FlowAxis’s performance, achieving significant improvements in both topological accuracy (clDice) and geometric fidelity (Hausdorff distance). Furthermore, qualitative assessments via curved planar reformations highlight its transformative potential in clinical workflows, while theoretical guarantees ensure reliability in safety-critical medical applications. Our work bridges the gap between mathematical rigor and practical medical imaging, providing the first complete theoretical framework for continuous vessel representation with provable optimality and convergence guarantees.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"33 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968729","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}
Accurate and generalizable cancer diagnosis from whole slide images (WSIs) remains challenging due to limited fine-grained annotations, complex tumor architectures, and domain shifts across scanners and institutions1. We introduce StructMIL, a structure-aware and prototype-driven multiple instance learning framework designed for robust and interpretable cancer detection and grading2. StructMIL integrates graph-based topological priors with histological context, employs prototype-enhanced pooling for stable and transparent predictions, and incorporates a unified domain-generalization strategy that combines contrastive alignment, adversarial confusion, and consistency regularization. Evaluated on Camelyon16 for breast cancer metastasis detection and PANDA for prostate cancer Gleason grading, StructMIL achieves state-of-the-art performance. On Camelyon16, StructMIL improves cross-center AUC by +3.2% over standard MIL baselines, reaching an AUC of 0.967. On PANDA, it improves cross-scanner Gleason grading robustness with a +7.4% Cohen’s Kappa gain compared with prior MIL models, demonstrating substantially reduced performance degradation under domain shift. StructMIL further provides interpretable prototype-based attribution maps that highlight biologically meaningful structures more reliably than conventional MIL and graph-free approaches3. By jointly improving accuracy, interpretability, and generalization across scanners and medical centers, StructMIL offers a practical and clinically aligned solution for large-scale deployment in multi-center computational pathology workflows4.
{"title":"Structure-aware generalization for heterogeneous histopathology via prototype-based multiple instance learning","authors":"Zhenjun Yu, Zhelin Xia, Donghao Xu, Zhiyuan Zhang, Lingling Zhang, Peng Zhang, Liang Wu, Bibo Wang, Helin Wang, Zhenxiong Zhao","doi":"10.1038/s41746-025-02289-4","DOIUrl":"https://doi.org/10.1038/s41746-025-02289-4","url":null,"abstract":"Accurate and generalizable cancer diagnosis from whole slide images (WSIs) remains challenging due to limited fine-grained annotations, complex tumor architectures, and domain shifts across scanners and institutions1. We introduce StructMIL, a structure-aware and prototype-driven multiple instance learning framework designed for robust and interpretable cancer detection and grading2. StructMIL integrates graph-based topological priors with histological context, employs prototype-enhanced pooling for stable and transparent predictions, and incorporates a unified domain-generalization strategy that combines contrastive alignment, adversarial confusion, and consistency regularization. Evaluated on Camelyon16 for breast cancer metastasis detection and PANDA for prostate cancer Gleason grading, StructMIL achieves state-of-the-art performance. On Camelyon16, StructMIL improves cross-center AUC by +3.2% over standard MIL baselines, reaching an AUC of 0.967. On PANDA, it improves cross-scanner Gleason grading robustness with a +7.4% Cohen’s Kappa gain compared with prior MIL models, demonstrating substantially reduced performance degradation under domain shift. StructMIL further provides interpretable prototype-based attribution maps that highlight biologically meaningful structures more reliably than conventional MIL and graph-free approaches3. By jointly improving accuracy, interpretability, and generalization across scanners and medical centers, StructMIL offers a practical and clinically aligned solution for large-scale deployment in multi-center computational pathology workflows4.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"15 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968747","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 : 2026-01-14DOI: 10.1038/s41746-026-02338-6
Wenhao Zhang, Jacek Kwiecinski, Aakash Shanbhag, Robert J. H. Miller, Shiva Mostafavi, Giselle Ramirez, Jirong Yi, Donghee Han, Damini Dey, Dominika Grodecka, Kajetan Grodecki, Mark Lemley, Paul Kavanagh, Joanna X. Liang, Jianhang Zhou, Valerie Builoff, Jon Hainer, Sylvain Carre, Leanne Barrett, Andrew J. Einstein, Stacey Knight, Steve Mason, Viet T. Le, Wanda Acampa, Samuel Wopperer, Panithaya Chareonthaitawee, Daniel S. Berman, Marcelo F. Di Carli, Piotr J. Slomka
Positron emission tomography (PET)/computed tomography (CT) for myocardial perfusion imaging (MPI) provides multiple imaging biomarkers, often evaluated separately. We developed an artificial intelligence (AI) model integrating key clinical PET MPI parameters to improve the diagnosis of obstructive coronary artery disease (CAD). From 17,348 patients undergoing cardiac PET/CT across four sites, 1664 with invasive coronary angiography and no prior CAD were retrospectively analyzed. Coronary artery calcium (CAC) scores were derived from CT attenuation correction maps, and XGBoost model was trained on one site using 10 image-derived parameters: CAC, stress/rest left ventricular ejection fraction, stress myocardial blood flow (MBF), myocardial flow reserve (MFR), ischemic and stress total perfusion deficit (TPD), transient ischemic dilation ratio, rate pressure product, and sex. External validation was performed across three independent sites. In the testing cohort (n = 1278; CAD prevalence 53%), the AI model achieved an area under the receiver operating characteristic curve (AUC) of 0.83 (95% CI: 0.81–0.85), outperforming experienced physicians (0.80, p = 0.02) and individual biomarkers such as ischemic TPD (0.79, p < 0.001) and MFR (0.75, p < 0.001). Performance was consistent across sex, body mass index, and age. AI integrating perfusion, flow, and CAC scoring improves PET MPI diagnostic accuracy, offering automated and interpretable predictions for CAD diagnosis.
正电子发射断层扫描(PET)/计算机断层扫描（CT）用于心肌灌注成像（MPI）提供多种成像生物标志物，通常单独评估。我们开发了一个人工智能（AI）模型，整合了临床PET MPI关键参数，以提高阻塞性冠状动脉疾病（CAD）的诊断。回顾性分析了来自四个部位的17,348例接受心脏PET/CT检查的患者，其中1664例接受有创冠状动脉造影且没有CAD病史。冠状动脉钙（CAC）评分来自CT衰减校正图，XGBoost模型使用10个图像衍生参数：CAC、应激/休息左心室射血分数、应激心肌血流量（MBF）、心肌血流储备（MFR）、缺血和应激总灌注缺陷（TPD）、短暂缺血扩张比、心率压积和性别在一个位点上训练。外部验证在三个独立的站点进行。在测试队列（n = 1278， CAD患病率53%）中，AI模型的受试者工作特征曲线下面积（AUC）为0.83 (95% CI: 0.81-0.85)，优于经验丰富的医生（0.80,p = 0.02）和个体生物标志物，如缺血性TPD （0.79, p < 0.001）和MFR （0.75, p < 0.001）。表现在性别、体重指数和年龄上是一致的。集成灌注、血流和CAC评分的AI提高了PET MPI诊断的准确性，为CAD诊断提供了自动化和可解释的预测。
{"title":"Multicenter evaluation of interpretable AI for coronary artery disease diagnosis from PET biomarkers","authors":"Wenhao Zhang, Jacek Kwiecinski, Aakash Shanbhag, Robert J. H. Miller, Shiva Mostafavi, Giselle Ramirez, Jirong Yi, Donghee Han, Damini Dey, Dominika Grodecka, Kajetan Grodecki, Mark Lemley, Paul Kavanagh, Joanna X. Liang, Jianhang Zhou, Valerie Builoff, Jon Hainer, Sylvain Carre, Leanne Barrett, Andrew J. Einstein, Stacey Knight, Steve Mason, Viet T. Le, Wanda Acampa, Samuel Wopperer, Panithaya Chareonthaitawee, Daniel S. Berman, Marcelo F. Di Carli, Piotr J. Slomka","doi":"10.1038/s41746-026-02338-6","DOIUrl":"https://doi.org/10.1038/s41746-026-02338-6","url":null,"abstract":"Positron emission tomography (PET)/computed tomography (CT) for myocardial perfusion imaging (MPI) provides multiple imaging biomarkers, often evaluated separately. We developed an artificial intelligence (AI) model integrating key clinical PET MPI parameters to improve the diagnosis of obstructive coronary artery disease (CAD). From 17,348 patients undergoing cardiac PET/CT across four sites, 1664 with invasive coronary angiography and no prior CAD were retrospectively analyzed. Coronary artery calcium (CAC) scores were derived from CT attenuation correction maps, and XGBoost model was trained on one site using 10 image-derived parameters: CAC, stress/rest left ventricular ejection fraction, stress myocardial blood flow (MBF), myocardial flow reserve (MFR), ischemic and stress total perfusion deficit (TPD), transient ischemic dilation ratio, rate pressure product, and sex. External validation was performed across three independent sites. In the testing cohort (n = 1278; CAD prevalence 53%), the AI model achieved an area under the receiver operating characteristic curve (AUC) of 0.83 (95% CI: 0.81–0.85), outperforming experienced physicians (0.80, p = 0.02) and individual biomarkers such as ischemic TPD (0.79, p < 0.001) and MFR (0.75, p < 0.001). Performance was consistent across sex, body mass index, and age. AI integrating perfusion, flow, and CAC scoring improves PET MPI diagnostic accuracy, offering automated and interpretable predictions for CAD diagnosis.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"26 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968749","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 : 2026-01-14DOI: 10.1038/s41746-026-02340-y
Igor Matias,Maximilian Haas,Eric J Daza,Matthias Kliegel,Katarzyna Wac
Continuous and scalable monitoring of cognition and affective states is critical for the early detection of brain health, which is currently limited by the burden of active assessments. This study investigated the potential of consumer-grade wearable and mobile technologies to passively predict 21 cognitive and mental health outcomes in real-world conditions. We collected data from 82 cognitively healthy adults, including passively measured behaviour, physiology, and environmental exposures longitudinally, for 10 months. Active data were gathered in four waves using validated patient- and performance-reported outcomes. Data quality assurance involved a data filtering resulting in average wearable data coverage of 96% per day. Artificial Intelligence-powered prediction was applied, and performance was assessed using subject- and wave-dependent cross-validation. Cognitive and affective outcomes were predicted with low scaled errors. Patient-reported outcomes were more predictable than performance-based ones. Environmental and physiological metrics emerged as the most informative predictors. Passive multimodal data captured meaningful variability in cognition and affect, demonstrating the feasibility of low-burden, scalable approaches to continuous brain-health monitoring. Feature-importance analyses suggested that environmental exposures better explained inter-individual differences, whereas physiological and behavioural rhythms captured within-person changes. These findings highlight the potential of everyday technologies for population-level tracking of brain-health and deviations from expected trajectories.
{"title":"Digital biomarkers for brain health: passive and continuous assessment from wearable sensors.","authors":"Igor Matias,Maximilian Haas,Eric J Daza,Matthias Kliegel,Katarzyna Wac","doi":"10.1038/s41746-026-02340-y","DOIUrl":"https://doi.org/10.1038/s41746-026-02340-y","url":null,"abstract":"Continuous and scalable monitoring of cognition and affective states is critical for the early detection of brain health, which is currently limited by the burden of active assessments. This study investigated the potential of consumer-grade wearable and mobile technologies to passively predict 21 cognitive and mental health outcomes in real-world conditions. We collected data from 82 cognitively healthy adults, including passively measured behaviour, physiology, and environmental exposures longitudinally, for 10 months. Active data were gathered in four waves using validated patient- and performance-reported outcomes. Data quality assurance involved a data filtering resulting in average wearable data coverage of 96% per day. Artificial Intelligence-powered prediction was applied, and performance was assessed using subject- and wave-dependent cross-validation. Cognitive and affective outcomes were predicted with low scaled errors. Patient-reported outcomes were more predictable than performance-based ones. Environmental and physiological metrics emerged as the most informative predictors. Passive multimodal data captured meaningful variability in cognition and affect, demonstrating the feasibility of low-burden, scalable approaches to continuous brain-health monitoring. Feature-importance analyses suggested that environmental exposures better explained inter-individual differences, whereas physiological and behavioural rhythms captured within-person changes. These findings highlight the potential of everyday technologies for population-level tracking of brain-health and deviations from expected trajectories.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"390 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145961320","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 : 2026-01-14DOI: 10.1038/s41746-025-02243-4
Curtis A. Merriweather, Jr., Kalle Lyytinen, David Aron, Michael R. Cauley
Electronic health record (EHR) systems were designed to enhance clinical decision-making, yet the way data is organized and displayed can create significant cognitive demands for physicians. This study examines how EHR data usability (data quality, data completeness, and data-driven use) and system usability jointly shape physicians’ cognitive load. Using survey responses from 564 physicians across 32 specialties, we tested a mediated model with covariance-based structural equation modeling. Reliability and validity were assessed through standard psychometric criteria. Findings show that stronger data usability increases germane cognitive load, promoting deeper engagement with clinically meaningful information. In contrast, higher system usability reduces extraneous cognitive load by aligning interface design with clinical workflow and minimizing navigation-related effort. Information overload partially mediated these effects, suggesting that better data usability helps physicians better filter irrelevant data and stay focused on diagnostically relevant cues. Overall, the results highlight two levers for improving cognitive performance: enhancing system usability lowers unnecessary cognitive effort and documentation-related errors, while improving data usability supports reasoning-intensive diagnostic work. Optimizing both fosters balanced cognitive load and more sustainable, error-resilient clinical decision-making.
{"title":"When better data meets better design: How EHR data usability and system usability shape physicians’ cognitive load","authors":"Curtis A. Merriweather, Jr., Kalle Lyytinen, David Aron, Michael R. Cauley","doi":"10.1038/s41746-025-02243-4","DOIUrl":"https://doi.org/10.1038/s41746-025-02243-4","url":null,"abstract":"Electronic health record (EHR) systems were designed to enhance clinical decision-making, yet the way data is organized and displayed can create significant cognitive demands for physicians. This study examines how EHR data usability (data quality, data completeness, and data-driven use) and system usability jointly shape physicians’ cognitive load. Using survey responses from 564 physicians across 32 specialties, we tested a mediated model with covariance-based structural equation modeling. Reliability and validity were assessed through standard psychometric criteria. Findings show that stronger data usability increases germane cognitive load, promoting deeper engagement with clinically meaningful information. In contrast, higher system usability reduces extraneous cognitive load by aligning interface design with clinical workflow and minimizing navigation-related effort. Information overload partially mediated these effects, suggesting that better data usability helps physicians better filter irrelevant data and stay focused on diagnostically relevant cues. Overall, the results highlight two levers for improving cognitive performance: enhancing system usability lowers unnecessary cognitive effort and documentation-related errors, while improving data usability supports reasoning-intensive diagnostic work. Optimizing both fosters balanced cognitive load and more sustainable, error-resilient clinical decision-making.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"21 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968731","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 : 2026-01-14DOI: 10.1038/s41746-025-02315-5
Alireza Gharabaghi, Sergiu Groppa, Elena Casas, Alfons Schnitzler, Laura Muñoz-Delgado, Vicky L. Marshall, Jessica Karl, Lin Zhang, Ramiro Alvarez, Mary S. Feldman, Michael J. Soileau, Lan Luo, Benjamin L. Walter, Chengyuan Wu, Hong Lei, Damian M. Herz, Devyani Nanduri, Claudia A. Salazar, Corneliu Luca, Daniel Weiss
Remote, internet-based deep brain stimulation programming for Parkinson’s disease accelerates clinical benefits postoperatively by improving access to therapy adjustments compared to in-clinic optimization. After completion of the initial digital programming phase, we show that clinical outcomes, quality of life, and safety remain sustained over at least twelve months under routine care conditions. Embedding a randomized trial within a larger cohort study enables long-term, real-world evaluation, offering a scalable and pragmatic model for assessing complex digital interventions in routine clinical care. (NCT05269862 registered on 2022-03-08 and NCT04071847 registered on 2019-08-28).
{"title":"Real-world multicenter assessment of sustained clinical outcomes after digital deep brain stimulation","authors":"Alireza Gharabaghi, Sergiu Groppa, Elena Casas, Alfons Schnitzler, Laura Muñoz-Delgado, Vicky L. Marshall, Jessica Karl, Lin Zhang, Ramiro Alvarez, Mary S. Feldman, Michael J. Soileau, Lan Luo, Benjamin L. Walter, Chengyuan Wu, Hong Lei, Damian M. Herz, Devyani Nanduri, Claudia A. Salazar, Corneliu Luca, Daniel Weiss","doi":"10.1038/s41746-025-02315-5","DOIUrl":"https://doi.org/10.1038/s41746-025-02315-5","url":null,"abstract":"Remote, internet-based deep brain stimulation programming for Parkinson’s disease accelerates clinical benefits postoperatively by improving access to therapy adjustments compared to in-clinic optimization. After completion of the initial digital programming phase, we show that clinical outcomes, quality of life, and safety remain sustained over at least twelve months under routine care conditions. Embedding a randomized trial within a larger cohort study enables long-term, real-world evaluation, offering a scalable and pragmatic model for assessing complex digital interventions in routine clinical care. (NCT05269862 registered on 2022-03-08 and NCT04071847 registered on 2019-08-28).","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"48 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968748","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 : 2026-01-14DOI: 10.1038/s41746-025-02327-1
Elias Stenhede, Agnar Martin Bjørnstad, Arian Ranjbar
Billions of clinical ECGs exist only as paper scans, making them unusable for modern automated diagnostics. We introduce a fully automated, modular framework that converts scanned or photographed ECGs into digital signals, suitable for both clinical and research applications. The framework is validated on 37,191 ECG images with 1596 collected at Akershus University Hospital, where the algorithm obtains a mean signal-to-noise ratio of 19.65 dB on scanned papers with common artifacts. It is further evaluated on the Emory Paper Digitization ECG Dataset, comprising 35,595 images, including images with perspective distortion, wrinkles, and stains. The model improves on the state-of-the-art in all subcategories. The full software is released as open-source, promoting reproducibility and further development. We hope the software will contribute to unlocking retrospective ECG archives and democratize access to AI-driven diagnostics.
{"title":"Digitizing paper ECGs at scale: an open-source algorithm for clinical research","authors":"Elias Stenhede, Agnar Martin Bjørnstad, Arian Ranjbar","doi":"10.1038/s41746-025-02327-1","DOIUrl":"https://doi.org/10.1038/s41746-025-02327-1","url":null,"abstract":"Billions of clinical ECGs exist only as paper scans, making them unusable for modern automated diagnostics. We introduce a fully automated, modular framework that converts scanned or photographed ECGs into digital signals, suitable for both clinical and research applications. The framework is validated on 37,191 ECG images with 1596 collected at Akershus University Hospital, where the algorithm obtains a mean signal-to-noise ratio of 19.65 dB on scanned papers with common artifacts. It is further evaluated on the Emory Paper Digitization ECG Dataset, comprising 35,595 images, including images with perspective distortion, wrinkles, and stains. The model improves on the state-of-the-art in all subcategories. The full software is released as open-source, promoting reproducibility and further development. We hope the software will contribute to unlocking retrospective ECG archives and democratize access to AI-driven diagnostics.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"57 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145968739","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}
This study addresses the critical challenge of master-slave latency in robot-assisted telesurgery by introducing a Digital Twin Visual Assistance (DTVA) system. DTVA integrates parametric 3D modeling and virtual endoscopic visualization within a tri-layered architecture to enable real-time bidirectional synchronization. The system was evaluated on a geographically distributed robotic platform using programmable latency emulation. Results demonstrated that DTVA maintained spatial precision within 2 mm error under typical conditions and reduced peg-transfer completion time by 13.6% under 900 ms communication latency while lowering operator workload by 27.2%. Clinical validation through teleoperated radical nephrectomy under 300 ms communication latency confirmed feasibility, with all procedures completed successfully without complications and favorable perioperative outcomes. The study establishes DTVA’s capacity to mitigate latency effects and demonstrates preliminary clinical feasibility for telesurgical procedures.
{"title":"Enhancing telesurgical safety with predictive digital twin synchronization: a framework for latency compensation in robotic surgery","authors":"Hang Yuan, Junjie Li, Bo Guan, Guangdi Chu, Wei Jiao, Hongzhi Zheng, Xingchi Liu, Jianchang Zhao, Jinhua Li, Jianmin Li, Xuecheng Yang, Haitao Niu","doi":"10.1038/s41746-025-02283-w","DOIUrl":"https://doi.org/10.1038/s41746-025-02283-w","url":null,"abstract":"This study addresses the critical challenge of master-slave latency in robot-assisted telesurgery by introducing a Digital Twin Visual Assistance (DTVA) system. DTVA integrates parametric 3D modeling and virtual endoscopic visualization within a tri-layered architecture to enable real-time bidirectional synchronization. The system was evaluated on a geographically distributed robotic platform using programmable latency emulation. Results demonstrated that DTVA maintained spatial precision within 2 mm error under typical conditions and reduced peg-transfer completion time by 13.6% under 900 ms communication latency while lowering operator workload by 27.2%. Clinical validation through teleoperated radical nephrectomy under 300 ms communication latency confirmed feasibility, with all procedures completed successfully without complications and favorable perioperative outcomes. The study establishes DTVA’s capacity to mitigate latency effects and demonstrates preliminary clinical feasibility for telesurgical procedures.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"51 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145956266","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 : 2026-01-13DOI: 10.1038/s41746-025-02287-6
Chenshen Huang, Haoyun Xia, Xi Xiao, Hong Chen, Yiqing Jiang, Yahui Lyu, Zhizhan Ni, Tianyang Wang, Ning Wang, Qi Huang
Weakly supervised segmentation of cancerous regions in whole-slide images (WSIs) is a crucial task in computational pathology, but it is severely hampered by the need for expensive pixel-level annotations. Existing Multiple Instance Learning (MIL) frameworks, while popular, typically fail to produce accurate segmentation masks because they treat WSIs as an unordered ’bag-of-patches’, ignoring the critical tissue topology and architectural patterns that define malignancy. In this paper, we address this fundamental limitation by proposing Geometric Multi-Instance Learning (Geo-MIL), a novel graph-based framework that explicitly models the spatial relationships between tissue patches. At the core of our method is a new topological attention mechanism that operates on the WSI graph, learning to identify and prioritize entire diagnostically relevant tissue structures over isolated patch features. Through extensive experiments on three public gastric cancer datasets, we demonstrate that Geo-MIL significantly outperforms a wide array of state-of-the-art baselines, achieving a new benchmark in both segmentation accuracy and classification performance. Our work represents a significant step towards bridging the gap between weak slide-level labels and precise, pixel-level predictions, paving the way for scalable and accurate quantitative analysis in digital pathology.
{"title":"Geometric multi-instance learning for weakly supervised gastric cancer segmentation","authors":"Chenshen Huang, Haoyun Xia, Xi Xiao, Hong Chen, Yiqing Jiang, Yahui Lyu, Zhizhan Ni, Tianyang Wang, Ning Wang, Qi Huang","doi":"10.1038/s41746-025-02287-6","DOIUrl":"https://doi.org/10.1038/s41746-025-02287-6","url":null,"abstract":"Weakly supervised segmentation of cancerous regions in whole-slide images (WSIs) is a crucial task in computational pathology, but it is severely hampered by the need for expensive pixel-level annotations. Existing Multiple Instance Learning (MIL) frameworks, while popular, typically fail to produce accurate segmentation masks because they treat WSIs as an unordered ’bag-of-patches’, ignoring the critical tissue topology and architectural patterns that define malignancy. In this paper, we address this fundamental limitation by proposing Geometric Multi-Instance Learning (Geo-MIL), a novel graph-based framework that explicitly models the spatial relationships between tissue patches. At the core of our method is a new topological attention mechanism that operates on the WSI graph, learning to identify and prioritize entire diagnostically relevant tissue structures over isolated patch features. Through extensive experiments on three public gastric cancer datasets, we demonstrate that Geo-MIL significantly outperforms a wide array of state-of-the-art baselines, achieving a new benchmark in both segmentation accuracy and classification performance. Our work represents a significant step towards bridging the gap between weak slide-level labels and precise, pixel-level predictions, paving the way for scalable and accurate quantitative analysis in digital pathology.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"10 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145956267","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}
Myopia is a major global health concern. To enable precision myopia management, we developed a Transformer-based artificial intelligence (AI) model, the Myopia Progression Predictive Model (MPPM), comprising two modules: the Natural Progression Module (NPM) for predicting untreated myopia progression and the Intervention Progression Module (IPM) for forecasting progression under specific interventions. NPM was trained on 1,109,827 refractive records from 304,353 children and adolescents, achieving high predictive accuracy for future spherical equivalent (SE) and axial length (AL) over a 10-year period. In the internal test set, SE prediction reached R² = 0.94, MAE = 0.35D; for AL, R² = 0.91, MAE = 0.16 mm. Comparable performance was observed in external validation. IPM was trained on four intervention cohorts (0.01% atropine, orthokeratology, peripheral defocus spectacles, and repeated low-level red light [RLRL] therapy) using a Transformer-based causal machine learning framework, enabling individualized estimation of treatment effects. It accurately predicted myopia changes under each intervention (SE: R² > 0.88, MAE < 0.45D; AL: R² > 0.80, MAE < 0.31 mm). Among the interventions, RLRL slightly reversed myopia progression, whereas the others slowed myopia progression. MPPM demonstrates strong promise as an AI-driven platform for personalized prediction and optimization of pediatric myopia management.
{"title":"AI-guided personalized predictions on myopia progression and interventions.","authors":"Sian Liu,Yuxing Lu,Xiaoman Li,Xiaoniao Chen,Zhuo Sun,Gen Li,Kai Wang,Wei Wu,Hui Xu,Hongyi Li,Changxi Hu,Zixing Zou,Miao Zhang,Xuan Zhang,Wenyang Lu,Yun Yin,Jia Qu,Kang Zhang,Jie Chen","doi":"10.1038/s41746-025-02308-4","DOIUrl":"https://doi.org/10.1038/s41746-025-02308-4","url":null,"abstract":"Myopia is a major global health concern. To enable precision myopia management, we developed a Transformer-based artificial intelligence (AI) model, the Myopia Progression Predictive Model (MPPM), comprising two modules: the Natural Progression Module (NPM) for predicting untreated myopia progression and the Intervention Progression Module (IPM) for forecasting progression under specific interventions. NPM was trained on 1,109,827 refractive records from 304,353 children and adolescents, achieving high predictive accuracy for future spherical equivalent (SE) and axial length (AL) over a 10-year period. In the internal test set, SE prediction reached R² = 0.94, MAE = 0.35D; for AL, R² = 0.91, MAE = 0.16 mm. Comparable performance was observed in external validation. IPM was trained on four intervention cohorts (0.01% atropine, orthokeratology, peripheral defocus spectacles, and repeated low-level red light [RLRL] therapy) using a Transformer-based causal machine learning framework, enabling individualized estimation of treatment effects. It accurately predicted myopia changes under each intervention (SE: R² > 0.88, MAE < 0.45D; AL: R² > 0.80, MAE < 0.31 mm). Among the interventions, RLRL slightly reversed myopia progression, whereas the others slowed myopia progression. MPPM demonstrates strong promise as an AI-driven platform for personalized prediction and optimization of pediatric myopia management.","PeriodicalId":19349,"journal":{"name":"NPJ Digital Medicine","volume":"9 1","pages":""},"PeriodicalIF":15.2,"publicationDate":"2026-01-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145956024","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：