Pub Date : 2026-01-12DOI: 10.1016/j.cmpb.2026.109241
Shilun Du, Yingda Hu, Fan Wei, Yong Lei
Background and Objectives:
Ultrasonically activated surgical devices (UASDs) are widely used in surgery due to their cutting, hemostatic, and thermal control capabilities. Modeling the UASD cutting process enhances understanding of these surgical procedures, aiding in surgery planning and design optimization. However, existing models lack consideration for the high-frequency cutting interactions, limiting their predictive accuracy. This study aims to develop a UASD-tissue interaction cutting model that considers high-frequency interactions and enhances prediction accuracy for multi-physical fields during cutting.
Methods:
This paper models the multi-field interaction process during the soft tissue cutting in UASD. First, a novel multi-field interaction cutting model is proposed, designed to predict cutting force, deformation, temperature, and tissue damage. Second, a LuGre-based interactive force module considering cellular rupture lubrication effects is developed for characterizing high-frequency UASD-tissue interactions. Third, a localized contact algorithm utilizing position-based dynamics and an adaptive time solver are proposed to achieve stable contact and solve the multi-time scale mechanism equations. Numerical experiments and physical experiments on phantoms and porcine livers are conducted.
Results:
The simulated force, temperature, damage, and deformation are consistent with the physical experimental results. The model captures the negative correlation between cutting speed and lubrication with temperature and friction, and shows increased vibration amplitude can lead to higher friction and heat generation, while maintaining stability across different cutting scenarios.
Conclusions:
The proposed model can robustly and accurately predict the multi-physical interactions during cutting, providing insights into the UASDs cutting process, thereby facilitating surgical planning and instrument design.
{"title":"A novel multi-field interaction cutting model for ultrasonically activated surgical devices","authors":"Shilun Du, Yingda Hu, Fan Wei, Yong Lei","doi":"10.1016/j.cmpb.2026.109241","DOIUrl":"10.1016/j.cmpb.2026.109241","url":null,"abstract":"<div><h3>Background and Objectives:</h3><div>Ultrasonically activated surgical devices (UASDs) are widely used in surgery due to their cutting, hemostatic, and thermal control capabilities. Modeling the UASD cutting process enhances understanding of these surgical procedures, aiding in surgery planning and design optimization. However, existing models lack consideration for the high-frequency cutting interactions, limiting their predictive accuracy. This study aims to develop a UASD-tissue interaction cutting model that considers high-frequency interactions and enhances prediction accuracy for multi-physical fields during cutting.</div></div><div><h3>Methods:</h3><div>This paper models the multi-field interaction process during the soft tissue cutting in UASD. First, a novel multi-field interaction cutting model is proposed, designed to predict cutting force, deformation, temperature, and tissue damage. Second, a LuGre-based interactive force module considering cellular rupture lubrication effects is developed for characterizing high-frequency UASD-tissue interactions. Third, a localized contact algorithm utilizing position-based dynamics and an adaptive time solver are proposed to achieve stable contact and solve the multi-time scale mechanism equations. Numerical experiments and physical experiments on phantoms and porcine livers are conducted.</div></div><div><h3>Results:</h3><div>The simulated force, temperature, damage, and deformation are consistent with the physical experimental results. The model captures the negative correlation between cutting speed and lubrication with temperature and friction, and shows increased vibration amplitude can lead to higher friction and heat generation, while maintaining stability across different cutting scenarios.</div></div><div><h3>Conclusions:</h3><div>The proposed model can robustly and accurately predict the multi-physical interactions during cutting, providing insights into the UASDs cutting process, thereby facilitating surgical planning and instrument design.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"277 ","pages":"Article 109241"},"PeriodicalIF":4.8,"publicationDate":"2026-01-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145975306","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}
Although artificial-intelligence-enhanced electrocardiograms (AI-ECGs) offer prediction and diagnosis capabilities superior to those of humans, they exhibit poor explainability and interpretability because of their complex-neural-network-derived black-box characteristics. To augment the explainability of artificial intelligence-enhanced electrocardiogram (AI-ECG) analysis, we proposed a method combining AI-ECG and synthetic ECG database created by a multi-scale heart simulator.
Methods
Using the “UT-Heart” multi-scale heart simulator running on the supercomputer Fugaku, we simulated 30,720 12-lead ECG recordings. This dataset comprises an exhaustive combination of 12 cellular and subcellular pathologies reportedly associated with heart failure and was analysed using a previously developed AI-ECG system that accurately classifies ECGs into New York Heart Association (NYHA) functional classes. By analysing the relationship between HF severity and labelled pathology in each simulated ECG recording, we elucidated the origin of abnormalities detected using AI-ECG.
Results
AI-ECG classified 30,618 ECGs (excluding 102 arrhythmia cases) into 2234 control and 28,384 HF cases. A separate three-group classification identified 2234 control, 18,444 NYHA I/II, and 9940 NYHA III/IV cases. In the two-group classification, significant differences (p < 0.01) were observed in sodium (Na) and Na–calcium exchanger currents and the transmural distribution of distinct cell types. Although the three-group classification revealed a severity-dependent progression of the Na current abnormality, the cell distribution in NYHA III/IV was closer to that of normal cases than to that of NYHA I/II. These findings did not explain the changes in the ECG waveform that the AI-ECG identified as notable features of heart failure in the heatmap analysis.
Conclusions
The ECG dataset generated using the multi-scale heart simulator can enhance the explainability of AI-ECGs by elucidating the mechanisms underlying HF-severity-specific changes in ECGs of heart failure.
{"title":"Multi-scale heart simulation augments the explainability of artificial intelligence-enabled electrocardiogram through provision of an electrocardiogram database labelled with cellular pathologies","authors":"Jun-ichi Okada , Katsuhito Fujiu , Eriko Hasumi , Ying Chen , Takumi Washio , Toshiaki Hisada , Seiryo Sugiura","doi":"10.1016/j.cmpb.2026.109247","DOIUrl":"10.1016/j.cmpb.2026.109247","url":null,"abstract":"<div><h3>Background and Objectives</h3><div>Although artificial-intelligence-enhanced electrocardiograms (AI-ECGs) offer prediction and diagnosis capabilities superior to those of humans, they exhibit poor explainability and interpretability because of their complex-neural-network-derived black-box characteristics. To augment the explainability of artificial intelligence-enhanced electrocardiogram (AI-ECG) analysis, we proposed a method combining AI-ECG and synthetic ECG database created by a multi-scale heart simulator.</div></div><div><h3>Methods</h3><div>Using the “UT-Heart” multi-scale heart simulator running on the supercomputer Fugaku, we simulated 30,720 12-lead ECG recordings. This dataset comprises an exhaustive combination of 12 cellular and subcellular pathologies reportedly associated with heart failure and was analysed using a previously developed AI-ECG system that accurately classifies ECGs into New York Heart Association (NYHA) functional classes. By analysing the relationship between HF severity and labelled pathology in each simulated ECG recording, we elucidated the origin of abnormalities detected using AI-ECG.</div></div><div><h3>Results</h3><div>AI-ECG classified 30,618 ECGs (excluding 102 arrhythmia cases) into 2234 control and 28,384 HF cases. A separate three-group classification identified 2234 control, 18,444 NYHA I/II, and 9940 NYHA III/IV cases. In the two-group classification, significant differences (<em>p</em> < 0.01) were observed in sodium (Na) and Na–calcium exchanger currents and the transmural distribution of distinct cell types. Although the three-group classification revealed a severity-dependent progression of the Na current abnormality, the cell distribution in NYHA III/IV was closer to that of normal cases than to that of NYHA I/II. These findings did not explain the changes in the ECG waveform that the AI-ECG identified as notable features of heart failure in the heatmap analysis.</div></div><div><h3>Conclusions</h3><div>The ECG dataset generated using the multi-scale heart simulator can enhance the explainability of AI-ECGs by elucidating the mechanisms underlying HF-severity-specific changes in ECGs of heart failure.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"277 ","pages":"Article 109247"},"PeriodicalIF":4.8,"publicationDate":"2026-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145975339","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 : 2026-01-10DOI: 10.1016/j.cmpb.2026.109249
Sofia M. Monteiro , Patrícia Bota , Pedro S. Cunha , Mário M. Oliveira , Sérgio Laranjo , Hugo Plácido da Silva
Background and Objective:
This systematic review evaluates the current state of Machine Learning (ML) methods for predicting Atrial Fibrillation (AF) recurrence following catheter ablation. With the growing use of ML, a systematic evaluation of performance and key influencing factors such as study design, data types, and reporting is needed. The main objectives are to provide an updated overview of current achievements of ML in this field, anticipate future challenges and opportunities, and derive methodological recommendations based on the findings.
Methods:
Seven databases were systematically searched, and studies proposing ML algorithms with well-documented implementation, testing, and reporting of performance metrics underwent a qualitative synthesis and risk-of-bias assessment. A meta-analysis of 17 studies was conducted using the Area Under the receiver operating characteristic Curve (AUC) as the most commonly reported performance metric.
Results:
The mean overall AUC was 0.81, indicating reasonable predictive accuracy, although there was substantial inter-study heterogeneity. Meta-regression identified sample size and input data type (clinical, imaging, or electrophysiological) as significant contributors to this heterogeneity. Subgroup analysis demonstrated that models incorporating complex data modalities achieved higher predictive accuracy and lower heterogeneity compared to those relying solely on simpler clinical variables.
Conclusion:
This review quantifies the performance of ML algorithms in predicting AF recurrence and establishes a benchmark for future research. It also highlights key challenges, including the lack of standardized datasets and limited generalizability. Incorporating more complex data sources may improve model performance, reduce inconsistencies, and enhance the potential clinical applicability of ML models in guiding patient management.
{"title":"Machine learning for the prediction of atrial fibrillation recurrence after catheter ablation: A systematic review and meta-analysis","authors":"Sofia M. Monteiro , Patrícia Bota , Pedro S. Cunha , Mário M. Oliveira , Sérgio Laranjo , Hugo Plácido da Silva","doi":"10.1016/j.cmpb.2026.109249","DOIUrl":"10.1016/j.cmpb.2026.109249","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>This systematic review evaluates the current state of Machine Learning (ML) methods for predicting Atrial Fibrillation (AF) recurrence following catheter ablation. With the growing use of ML, a systematic evaluation of performance and key influencing factors such as study design, data types, and reporting is needed. The main objectives are to provide an updated overview of current achievements of ML in this field, anticipate future challenges and opportunities, and derive methodological recommendations based on the findings.</div></div><div><h3>Methods:</h3><div>Seven databases were systematically searched, and studies proposing ML algorithms with well-documented implementation, testing, and reporting of performance metrics underwent a qualitative synthesis and risk-of-bias assessment. A meta-analysis of 17 studies was conducted using the Area Under the receiver operating characteristic Curve (AUC) as the most commonly reported performance metric.</div></div><div><h3>Results:</h3><div>The mean overall AUC was 0.81, indicating reasonable predictive accuracy, although there was substantial inter-study heterogeneity. Meta-regression identified sample size and input data type (clinical, imaging, or electrophysiological) as significant contributors to this heterogeneity. Subgroup analysis demonstrated that models incorporating complex data modalities achieved higher predictive accuracy and lower heterogeneity compared to those relying solely on simpler clinical variables.</div></div><div><h3>Conclusion:</h3><div>This review quantifies the performance of ML algorithms in predicting AF recurrence and establishes a benchmark for future research. It also highlights key challenges, including the lack of standardized datasets and limited generalizability. Incorporating more complex data sources may improve model performance, reduce inconsistencies, and enhance the potential clinical applicability of ML models in guiding patient management.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"277 ","pages":"Article 109249"},"PeriodicalIF":4.8,"publicationDate":"2026-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145965449","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 : 2026-01-09DOI: 10.1016/j.cmpb.2026.109244
Jonghun Kim, Hyunjin Park
Background and objective
: Medical image synthesis has broad applications in modality-to-modality translation, denoising, and super-resolution, when specific modalities are missing, various types of noise occur, or resolution discrepancies exist across modalities. The traditional approach requires a separate model for each task, making it inefficient, and nearly impossible to accommodate various tasks in medical image synthesis.
Methods:
We introduce a task-agnostic medical image-synthesis model utilizing prompt tuning that leverages a diffusion model and prompt tuning to fine-tune large capacity pretrained models efficiently. Our method can handle multiple tasks that cover various input–output combinations in a single model.
Results:
Our method can perform denoising, translation, super-resolution, and tumor inpainting tasks for brain MRI and abdominal CT. Through quantitative and qualitative evaluations, we demonstrate that our model achieves the best performance in terms of FID scores across all evaluated tasks. Our multi-task model achieves a PSNR of 25.76 and an SSIM of 0.908 for T1-to-T2 translation; a PSNR of 30.30 and an SSIM of 0.932 for denoising; a PSNR of 29.24 and an SSIM of 0.874 for super-resolution; and an FID of 16.18 with an LPIPS of 0.090 for tumor inpainting.
Conclusions:
We proposed a method that enables task-agnostic medical image synthesis, allowing for the specification of the desired synthesis task, modality, and organ of the target image via prompt tuning. Our method can be extended to other modalities and organs. The code is available at https://github.com/jongdory/VPT-Med.
{"title":"Visual prompt tuning for task-flexible medical image synthesis","authors":"Jonghun Kim, Hyunjin Park","doi":"10.1016/j.cmpb.2026.109244","DOIUrl":"10.1016/j.cmpb.2026.109244","url":null,"abstract":"<div><h3>Background and objective</h3><div>: Medical image synthesis has broad applications in modality-to-modality translation, denoising, and super-resolution, when specific modalities are missing, various types of noise occur, or resolution discrepancies exist across modalities. The traditional approach requires a separate model for each task, making it inefficient, and nearly impossible to accommodate various tasks in medical image synthesis.</div></div><div><h3>Methods:</h3><div>We introduce a task-agnostic medical image-synthesis model utilizing prompt tuning that leverages a diffusion model and prompt tuning to fine-tune large capacity pretrained models efficiently. Our method can handle multiple tasks that cover various input–output combinations in a single model.</div></div><div><h3>Results:</h3><div>Our method can perform denoising, translation, super-resolution, and tumor inpainting tasks for brain MRI and abdominal CT. Through quantitative and qualitative evaluations, we demonstrate that our model achieves the best performance in terms of FID scores across all evaluated tasks. Our multi-task model achieves a PSNR of 25.76 and an SSIM of 0.908 for T1-to-T2 translation; a PSNR of 30.30 and an SSIM of 0.932 for denoising; a PSNR of 29.24 and an SSIM of 0.874 for super-resolution; and an FID of 16.18 with an LPIPS of 0.090 for tumor inpainting.</div></div><div><h3>Conclusions:</h3><div>We proposed a method that enables task-agnostic medical image synthesis, allowing for the specification of the desired synthesis task, modality, and organ of the target image via prompt tuning. Our method can be extended to other modalities and organs. The code is available at <span><span>https://github.com/jongdory/VPT-Med</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"277 ","pages":"Article 109244"},"PeriodicalIF":4.8,"publicationDate":"2026-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145924500","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 : 2026-01-09DOI: 10.1016/j.cmpb.2026.109245
Adi Konsens , Alejandro F. Frangi , Gil Marom
Background and objective
Intracranial aneurysms (IA) cause hundreds of thousands of deaths annually, yet most remain undiagnosed until rupture due to their asymptomatic nature. Improved prediction of aneurysm initiation could enable earlier detection and intervention. While computational hemodynamic models can identify high-risk regions, previous studies were limited to small cohorts due to labor-intensive manual workflows. We developed the first semi-automated workflow to enable large-scale, patient-specific hemodynamic analysis of IA initiation.
Methods
Our workflow integrates automated centerline extraction for quantitative morphological characterization with computational fluid dynamics (CFD) simulations to derive wall shear stress patterns and hemodynamic markers. We tested the workflow's robustness across multiple IA types and anatomical locations, focusing primarily on sidewall aneurysms of the internal carotid artery (ICA).
Results
Our semi-automated workflow successfully processed 42 diverse cases, 5 of them initially failed but were subsequently resolved through manual reconstruction, demonstrating robust performance across sidewall ICA aneurysms (16 cases), bifurcation aneurysms (6 cases), and validation cohorts. Validation against published data showed consistent trends with mean normalized TAWSS values of 1.31±0.09 in aneurysmal cases versus 1.14±0.07 in controls, aligning with previous findings despite methodological differences.
Conclusions
The workflow's adaptability was confirmed across multiple anatomical configurations and region of interest selection methods. This scalable approach enables the statistical analysis necessary to identify reliable hemodynamic biomarkers for IA initiation, representing a critical advancement towards evidence-based prediction models for clinical risk stratification.
{"title":"Automated hemodynamic modeling to explore arterial curvature effects on intracranial aneurysm initiation","authors":"Adi Konsens , Alejandro F. Frangi , Gil Marom","doi":"10.1016/j.cmpb.2026.109245","DOIUrl":"10.1016/j.cmpb.2026.109245","url":null,"abstract":"<div><h3>Background and objective</h3><div>Intracranial aneurysms (IA) cause hundreds of thousands of deaths annually, yet most remain undiagnosed until rupture due to their asymptomatic nature. Improved prediction of aneurysm initiation could enable earlier detection and intervention. While computational hemodynamic models can identify high-risk regions, previous studies were limited to small cohorts due to labor-intensive manual workflows. We developed the first semi-automated workflow to enable large-scale, patient-specific hemodynamic analysis of IA initiation.</div></div><div><h3>Methods</h3><div>Our workflow integrates automated centerline extraction for quantitative morphological characterization with computational fluid dynamics (CFD) simulations to derive wall shear stress patterns and hemodynamic markers. We tested the workflow's robustness across multiple IA types and anatomical locations, focusing primarily on sidewall aneurysms of the internal carotid artery (ICA).</div></div><div><h3>Results</h3><div>Our semi-automated workflow successfully processed 42 diverse cases, 5 of them initially failed but were subsequently resolved through manual reconstruction, demonstrating robust performance across sidewall ICA aneurysms (16 cases), bifurcation aneurysms (6 cases), and validation cohorts. Validation against published data showed consistent trends with mean normalized TAWSS values of 1.31±0.09 in aneurysmal cases versus 1.14±0.07 in controls, aligning with previous findings despite methodological differences.</div></div><div><h3>Conclusions</h3><div>The workflow's adaptability was confirmed across multiple anatomical configurations and region of interest selection methods. This scalable approach enables the statistical analysis necessary to identify reliable hemodynamic biomarkers for IA initiation, representing a critical advancement towards evidence-based prediction models for clinical risk stratification.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"277 ","pages":"Article 109245"},"PeriodicalIF":4.8,"publicationDate":"2026-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145975305","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 : 2026-01-08DOI: 10.1016/j.cmpb.2026.109243
Siyu Mu, Wei Xuan Chan, Choon Hwai Yap
Background and Objective:
The unloaded cardiac geometry, representing the zero-stress and zero-strain reference state of the heart, is fundamental for personalized biomechanical modeling of cardiac function. However, this state cannot be directly observed in vivo, as clinical imaging only captures pressure-loaded geometries such as those at end-diastole. Traditional inverse finite element solvers are commonly used to reconstruct the unloaded geometry, but they require iterative optimization, are computationally expensive, and may suffer from convergence issues. The objective of this study was to develop an efficient and accurate deep learning framework to predict the unloaded left ventricular geometry directly from clinical end-diastolic states.
Methods:
We propose HeartUnloadNet, a graph attention-based neural network that incorporates both mesh topology and physiological parameters, including pressure, myocardial stiffness, and fiber orientation. The framework employs a cycle-consistent bidirectional training strategy, allowing reduced supervision by enforcing that the predicted unloaded state can reconstruct the original end-diastolic geometry. The model was trained and validated on 10,350 finite element simulations generated across diverse anatomical shapes and physiological conditions. Performance was evaluated using geometric metrics such as Dice similarity coefficient, Hausdorff distance, mean distance, and standard deviation of nodal errors.
Results:
HeartUnloadNet achieved sub-millimeter accuracy, with a Dice similarity coefficient of 0.986 0.023 and a Hausdorff distance of 0.083 0.028 cm. Compared to conventional inverse finite element solvers, the framework was over 100,000 times faster, with an average inference time of 0.02 seconds per case. Ablation studies demonstrated that cycle consistency enabled the model to maintain high accuracy even when only 3% of the training data were labeled. The method consistently outperformed baseline architectures across all evaluation metrics.
Conclusions:
HeartUnloadNet provides a scalable and accurate alternative to traditional inverse finite element approaches for estimating the unloaded cardiac geometry. By combining mesh-aware learning with physiological conditioning and reduced supervision, the framework achieves real-time performance while maintaining biomechanical fidelity. This work establishes a foundation for future integration of learning-based surrogates into clinical workflows, supporting patient-specific cardiac modeling and real-time functional assessment.
{"title":"HeartUnloadNet: A cycle-consistent graph network with reduced supervision for predicting unloaded cardiac geometry from diastolic states","authors":"Siyu Mu, Wei Xuan Chan, Choon Hwai Yap","doi":"10.1016/j.cmpb.2026.109243","DOIUrl":"10.1016/j.cmpb.2026.109243","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>The unloaded cardiac geometry, representing the zero-stress and zero-strain reference state of the heart, is fundamental for personalized biomechanical modeling of cardiac function. However, this state cannot be directly observed in vivo, as clinical imaging only captures pressure-loaded geometries such as those at end-diastole. Traditional inverse finite element solvers are commonly used to reconstruct the unloaded geometry, but they require iterative optimization, are computationally expensive, and may suffer from convergence issues. The objective of this study was to develop an efficient and accurate deep learning framework to predict the unloaded left ventricular geometry directly from clinical end-diastolic states.</div></div><div><h3>Methods:</h3><div>We propose HeartUnloadNet, a graph attention-based neural network that incorporates both mesh topology and physiological parameters, including pressure, myocardial stiffness, and fiber orientation. The framework employs a cycle-consistent bidirectional training strategy, allowing reduced supervision by enforcing that the predicted unloaded state can reconstruct the original end-diastolic geometry. The model was trained and validated on 10,350 finite element simulations generated across diverse anatomical shapes and physiological conditions. Performance was evaluated using geometric metrics such as Dice similarity coefficient, Hausdorff distance, mean distance, and standard deviation of nodal errors.</div></div><div><h3>Results:</h3><div>HeartUnloadNet achieved sub-millimeter accuracy, with a Dice similarity coefficient of 0.986 <span><math><mo>±</mo></math></span> 0.023 and a Hausdorff distance of 0.083 <span><math><mo>±</mo></math></span> 0.028 cm. Compared to conventional inverse finite element solvers, the framework was over 100,000 times faster, with an average inference time of 0.02 seconds per case. Ablation studies demonstrated that cycle consistency enabled the model to maintain high accuracy even when only 3% of the training data were labeled. The method consistently outperformed baseline architectures across all evaluation metrics.</div></div><div><h3>Conclusions:</h3><div>HeartUnloadNet provides a scalable and accurate alternative to traditional inverse finite element approaches for estimating the unloaded cardiac geometry. By combining mesh-aware learning with physiological conditioning and reduced supervision, the framework achieves real-time performance while maintaining biomechanical fidelity. This work establishes a foundation for future integration of learning-based surrogates into clinical workflows, supporting patient-specific cardiac modeling and real-time functional assessment.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"277 ","pages":"Article 109243"},"PeriodicalIF":4.8,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145965515","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 : 2026-01-08DOI: 10.1016/j.cmpb.2026.109240
Yaowen Zhang , Libera Fresiello , Peter H. Veltink , Dirk W. Donker , Ying Wang
<div><h3>Background and Objective:</h3><div>Heart failure (HF) poses a significant global health challenge, with early detection offering opportunities for improved outcomes. Abnormalities in heart rate (HR), particularly during daily activities, may serve as early indicators of HF risk. However, existing HR monitoring tools for HF detection are limited by their reliability on population-based averages. The estimation of individualized HR serves as a dynamic digital twin, enabling precise tracking of cardiac health biomarkers.</div></div><div><h3>Methods:</h3><div>This study introduces a novel physiological-model-based neural network (PMB-NN) framework to model the oxygen uptake (<span><math><mrow><mover><mrow><mtext>V</mtext></mrow><mrow><mo>̇</mo></mrow></mover><msub><mrow><mtext>O</mtext></mrow><mrow><mn>2</mn></mrow></msub></mrow></math></span>)-HR relationship during physical activities, establishing a physiology-grounded intermediate module for future daily life HR estimation from body movement signals. PMB-NN embeds physiological constraints, derived from our proposed simplified metabolic–HR physiological model (PM), into the neural network training process. The framework was trained and tested on individual datasets from 25 participants engaged in activities including resting, cycling, and running. HR estimation performance was evaluated for PMB-NN, comparing with benchmark fully connected neural network (FCNN) and PM, across three dimensions: numerical accuracy, physiological plausibility, and physiological interpretability. Furthermore, sensitivity analysis was conducted to verify the model’s robustness against input uncertainty.</div></div><div><h3>Results:</h3><div>The PMB-NN model adheres to human physiological principles while achieving high estimation accuracy, with median R<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span> score of 0.88, RMSE of 9.96 bpm and MAE of 8.87 bpm, even in the presence of intermittent data. Comparative statistical analysis demonstrates that the PMB-NN achieves performance on par with FCNN while significantly outperforming PM (<span><math><mrow><mi>p</mi><mo><</mo><mn>0</mn><mo>.</mo><mn>001</mn></mrow></math></span>). Meanwhile, PMB-NN reaches higher plausibility for HR-<span><math><mrow><mover><mrow><mtext>V</mtext></mrow><mrow><mo>̇</mo></mrow></mover><msub><mrow><mtext>O</mtext></mrow><mrow><mn>2</mn></mrow></msub></mrow></math></span> coupling (<span><math><mi>ρ</mi></math></span> = 1) than both FCNN (p = 0.028) and PM (<span><math><mrow><mi>p</mi><mo><</mo><mn>0</mn><mo>.</mo><mn>001</mn></mrow></math></span>). Furthermore, PMB-NN is adept at identifying personalized parameters of the PM, enabling reasonable HR reconstruction. Sensitivity analysis reveals that PMB-NN yields an RMSE within 15 bpm despite input uncertainties of up to 20% Gaussian noise, 4% outliers, and an 18 s time lag.</div></div><div><h3>Conclusion:</h3><div>This study confirms the validity of the PMB
{"title":"Physiological-model-based neural network for modeling the metabolic–heart rate relationship during physical activities","authors":"Yaowen Zhang , Libera Fresiello , Peter H. Veltink , Dirk W. Donker , Ying Wang","doi":"10.1016/j.cmpb.2026.109240","DOIUrl":"10.1016/j.cmpb.2026.109240","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Heart failure (HF) poses a significant global health challenge, with early detection offering opportunities for improved outcomes. Abnormalities in heart rate (HR), particularly during daily activities, may serve as early indicators of HF risk. However, existing HR monitoring tools for HF detection are limited by their reliability on population-based averages. The estimation of individualized HR serves as a dynamic digital twin, enabling precise tracking of cardiac health biomarkers.</div></div><div><h3>Methods:</h3><div>This study introduces a novel physiological-model-based neural network (PMB-NN) framework to model the oxygen uptake (<span><math><mrow><mover><mrow><mtext>V</mtext></mrow><mrow><mo>̇</mo></mrow></mover><msub><mrow><mtext>O</mtext></mrow><mrow><mn>2</mn></mrow></msub></mrow></math></span>)-HR relationship during physical activities, establishing a physiology-grounded intermediate module for future daily life HR estimation from body movement signals. PMB-NN embeds physiological constraints, derived from our proposed simplified metabolic–HR physiological model (PM), into the neural network training process. The framework was trained and tested on individual datasets from 25 participants engaged in activities including resting, cycling, and running. HR estimation performance was evaluated for PMB-NN, comparing with benchmark fully connected neural network (FCNN) and PM, across three dimensions: numerical accuracy, physiological plausibility, and physiological interpretability. Furthermore, sensitivity analysis was conducted to verify the model’s robustness against input uncertainty.</div></div><div><h3>Results:</h3><div>The PMB-NN model adheres to human physiological principles while achieving high estimation accuracy, with median R<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span> score of 0.88, RMSE of 9.96 bpm and MAE of 8.87 bpm, even in the presence of intermittent data. Comparative statistical analysis demonstrates that the PMB-NN achieves performance on par with FCNN while significantly outperforming PM (<span><math><mrow><mi>p</mi><mo><</mo><mn>0</mn><mo>.</mo><mn>001</mn></mrow></math></span>). Meanwhile, PMB-NN reaches higher plausibility for HR-<span><math><mrow><mover><mrow><mtext>V</mtext></mrow><mrow><mo>̇</mo></mrow></mover><msub><mrow><mtext>O</mtext></mrow><mrow><mn>2</mn></mrow></msub></mrow></math></span> coupling (<span><math><mi>ρ</mi></math></span> = 1) than both FCNN (p = 0.028) and PM (<span><math><mrow><mi>p</mi><mo><</mo><mn>0</mn><mo>.</mo><mn>001</mn></mrow></math></span>). Furthermore, PMB-NN is adept at identifying personalized parameters of the PM, enabling reasonable HR reconstruction. Sensitivity analysis reveals that PMB-NN yields an RMSE within 15 bpm despite input uncertainties of up to 20% Gaussian noise, 4% outliers, and an 18 s time lag.</div></div><div><h3>Conclusion:</h3><div>This study confirms the validity of the PMB","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"277 ","pages":"Article 109240"},"PeriodicalIF":4.8,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145965421","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 : 2026-01-06DOI: 10.1016/j.cmpb.2026.109242
Teng Lu , Zhongwei Sun , Xijing He
Background and objective
The elastic modulus of cage material (cage-E) is a key determinant of fusion outcomes in oblique lateral interbody fusion (OLIF), as it modulates the efficiency of mechanically induced osteogenesis (EMIO). Here, we establish a logarithmic predictive model linking cage-E to EMIO and delineate the underlying biomechanical mechanisms via computational biomechanical analysis.
Methods
A customized mechano-regulation algorithm was applied to finite element models of the L4/5 OLIF construct to simulate the iteration of tissue differentiation and regeneration, which was driven by mechanical stimulation (MechSt). The regenerative bone fraction at the final iteration was defined as EMIO. A total of 23 cage-E values ranging from 0.1 GPa to 110 GPa were evaluated.
Results
As cage-E increased from 0.1 GPa to 110 GPa, the OLIF construct stiffness increased from 3.29 to 6.02 N/mm to 4.95–6.13 N/mm across iterations; the stress-shielding MechSt region expanded from 0 to 0.92% to 9.75–53.67%, whereas the stress-growth MechSt region contracted from 100 to 99.08% to 90.25–46.33%. Correspondingly, EMIO declined from 92.05% to 55.44%. Logarithmic regression revealed strong correlations (R²=0.72–0.89) between cage-E and construct stiffness, MechSt distribution, and tissue regeneration.
Conclusions
Reduced cage-E enhances OLIF EMIO via a defined cascade biomechanical mechanism: cage-E logarithmically regulates construct stiffness, with lower cage-E mitigating stress shielding and preserving the osteogenic MechSt domain, in turn promoting osteoblastic differentiation of mesenchymal stem cells and bone regeneration. The established logarithmic model characterizes the cage-E–EMIO relationship and serves as a potential tool for cage-E screening to optimize OLIF fusion outcomes.
{"title":"Oblique lateral interbody fusion: role of the elastic modulus of the cage material in mechanically induced osteogenesis","authors":"Teng Lu , Zhongwei Sun , Xijing He","doi":"10.1016/j.cmpb.2026.109242","DOIUrl":"10.1016/j.cmpb.2026.109242","url":null,"abstract":"<div><h3>Background and objective</h3><div>The elastic modulus of cage material (cage-E) is a key determinant of fusion outcomes in oblique lateral interbody fusion (OLIF), as it modulates the efficiency of mechanically induced osteogenesis (EMIO). Here, we establish a logarithmic predictive model linking cage-E to EMIO and delineate the underlying biomechanical mechanisms via computational biomechanical analysis.</div></div><div><h3>Methods</h3><div>A customized mechano-regulation algorithm was applied to finite element models of the L4/5 OLIF construct to simulate the iteration of tissue differentiation and regeneration, which was driven by mechanical stimulation (MechSt). The regenerative bone fraction at the final iteration was defined as EMIO. A total of 23 cage-E values ranging from 0.1 GPa to 110 GPa were evaluated.</div></div><div><h3>Results</h3><div>As cage-E increased from 0.1 GPa to 110 GPa, the OLIF construct stiffness increased from 3.29 to 6.02 N/mm to 4.95–6.13 N/mm across iterations; the stress-shielding MechSt region expanded from 0 to 0.92% to 9.75–53.67%, whereas the stress-growth MechSt region contracted from 100 to 99.08% to 90.25–46.33%. Correspondingly, EMIO declined from 92.05% to 55.44%. Logarithmic regression revealed strong correlations (R²=0.72–0.89) between cage-E and construct stiffness, MechSt distribution, and tissue regeneration.</div></div><div><h3>Conclusions</h3><div>Reduced cage-E enhances OLIF EMIO via a defined cascade biomechanical mechanism: cage-E logarithmically regulates construct stiffness, with lower cage-E mitigating stress shielding and preserving the osteogenic MechSt domain, in turn promoting osteoblastic differentiation of mesenchymal stem cells and bone regeneration. The established logarithmic model characterizes the cage-E–EMIO relationship and serves as a potential tool for cage-E screening to optimize OLIF fusion outcomes.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"276 ","pages":"Article 109242"},"PeriodicalIF":4.8,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145921989","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 : 2026-01-05DOI: 10.1016/j.cmpb.2026.109238
Fan Zhang , Bingchen Yu , Jianwei Zuo , Rui Xu , Kai Pang , Wei Jin , Jiajia Luo
Background and Objective
White blood cells (WBCs) are key biomarkers of immune status, but current monitoring still relies on intermittent blood sampling and hematology analyzers, which are invasive and lack real-time, dynamic information. This work aims to develop a noninvasive system that continuously monitors WBC dynamics in nailfold microcirculation by combining a compact optical imaging device with deep learning–based detection and tracking.
Methods
We designed a portable microscopic imaging system that records high-frame-rate videos of nailfold capillaries under 532 nm illumination, where WBCs appear as bright optical gaps against the red blood cell column. From videos of 22 volunteers, we constructed dedicated vessel and WBC datasets and trained a two-stage YOLOv8-based detection framework that first localizes vascular regions and then detects WBCs within these regions. To enhance temporal consistency, we integrated a Flow-Guided Feature Aggregation module, and employed the ByteTrack multi-object tracking algorithm to assign unique IDs to WBCs and achieve real-time counting from streaming video. System performance was evaluated using mean average precision (mAP), precision, recall and F1-score.
Results
The proposed framework achieved accurate and stable vessel and WBC detection, with detection results closely matching manual annotations and maintaining robustness under motion blur and partial occlusion. The complete “detect–track–count” pipeline supports real-time analysis on a general computing platform while using only a compact optical device.
Conclusions
This study demonstrates a portable, noninvasive AI system that enables continuous in vivo monitoring of WBC dynamics in nailfold microcirculation without blood sampling. The approach provides a promising tool for scenarios requiring frequent WBC surveillance, such as chemotherapy monitoring and immune function assessment, and offers a transferable framework for other cell detection and microcirculation studies in medical imaging.
{"title":"Noninvasive real-time dynamic monitoring of white blood cells based on microscopic imaging and deep learning","authors":"Fan Zhang , Bingchen Yu , Jianwei Zuo , Rui Xu , Kai Pang , Wei Jin , Jiajia Luo","doi":"10.1016/j.cmpb.2026.109238","DOIUrl":"10.1016/j.cmpb.2026.109238","url":null,"abstract":"<div><h3>Background and Objective</h3><div>White blood cells (WBCs) are key biomarkers of immune status, but current monitoring still relies on intermittent blood sampling and hematology analyzers, which are invasive and lack real-time, dynamic information. This work aims to develop a noninvasive system that continuously monitors WBC dynamics in nailfold microcirculation by combining a compact optical imaging device with deep learning–based detection and tracking.</div></div><div><h3>Methods</h3><div>We designed a portable microscopic imaging system that records high-frame-rate videos of nailfold capillaries under 532 nm illumination, where WBCs appear as bright optical gaps against the red blood cell column. From videos of 22 volunteers, we constructed dedicated vessel and WBC datasets and trained a two-stage YOLOv8-based detection framework that first localizes vascular regions and then detects WBCs within these regions. To enhance temporal consistency, we integrated a Flow-Guided Feature Aggregation module, and employed the ByteTrack multi-object tracking algorithm to assign unique IDs to WBCs and achieve real-time counting from streaming video. System performance was evaluated using mean average precision (mAP), precision, recall and F1-score.</div></div><div><h3>Results</h3><div>The proposed framework achieved accurate and stable vessel and WBC detection, with detection results closely matching manual annotations and maintaining robustness under motion blur and partial occlusion. The complete “detect–track–count” pipeline supports real-time analysis on a general computing platform while using only a compact optical device.</div></div><div><h3>Conclusions</h3><div>This study demonstrates a portable, noninvasive AI system that enables continuous in vivo monitoring of WBC dynamics in nailfold microcirculation without blood sampling. The approach provides a promising tool for scenarios requiring frequent WBC surveillance, such as chemotherapy monitoring and immune function assessment, and offers a transferable framework for other cell detection and microcirculation studies in medical imaging.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"276 ","pages":"Article 109238"},"PeriodicalIF":4.8,"publicationDate":"2026-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145921974","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 : 2026-01-05DOI: 10.1016/j.cmpb.2026.109236
Xingji Fu , Xiaofang Yang , Feilong Hei , Anqiang Sun , Zengsheng Chen
Background and Objective
Pulsatile blood flow is considered more potential for delivering hemodynamic energy and enhancing microcirculatory perfusion in patients compared to non-pulsatile flow in ECMO. This study aims to systematically evaluate the effects of different blood flow modes on the hemodynamic environment, thrombosis risk, and oxygen transport within oxygenators.
Methods
QUADROX-i Adult Oxygenator was investigated using CFD to simulate its hemodynamic environment under different blood flow modes, comprising one non-pulsatile condition and nine pulsatile conditions with varying frequencies and amplitudes. The stasis (ART) and hypercoagulability (C[FXIa]) were used to assess the thrombosis risk and oxygen transport (PO2) was also analyzed. The dynamic blood volume (DBV) were calculated to reflect the effective volume within the oxygenator.
Results
Under all blood flow modes, the velocity distribution is more uneven in inlet-side and outlet-side transition regions, and the high value is near the inlet and outlet, and becomes lower away from the inlet and outlet. In gas exchange region, the velocity is low and evenly distributed. The region with the highest ART and C[FXIa] are located at the north corner region close to the outlet. The highest PO2 evenly appears in the region near the outlet. Under pulsatile conditions, As the flow rate increases, the distribution of velocity, ART and C[FXIa] becomes more uneven, and vice verse. Compared to non-pulsatile condition, the period-averaged ART, C[FXIa] and PO2 become higher, while the DBV decreases under pulsatile conditions. Amplitude has a more significant effect on all parameters than frequency. Higher amplitude results in the higher period-averaged ART, C[FXIa] and PO2, alongside a lower DBV.
Conclusions
Uneven flow field mainly occurs in the inlet-side and outlet-side transition region, and the uneven degree increases with the higher flow rate, and vice verse. The highest thrombosis risk locates in the north corner region close to the outlet and the highest oxygen transport occurs in the region close to the outlet. Pulsatile flow can enhance oxygen transport but increase thrombosis risk than non-pulsatile flow. Higher amplitude can increase thrombosis risk but improve oxygen transport in the oxygenator. The frequency variation exhibits minimal influence.
{"title":"Impact of different blood flow modes on hemodynamic environment, thrombosis risk and oxygen transport of oxygenators: A numerical simulation study","authors":"Xingji Fu , Xiaofang Yang , Feilong Hei , Anqiang Sun , Zengsheng Chen","doi":"10.1016/j.cmpb.2026.109236","DOIUrl":"10.1016/j.cmpb.2026.109236","url":null,"abstract":"<div><h3>Background and Objective</h3><div>Pulsatile blood flow is considered more potential for delivering hemodynamic energy and enhancing microcirculatory perfusion in patients compared to non-pulsatile flow in ECMO. This study aims to systematically evaluate the effects of different blood flow modes on the hemodynamic environment, thrombosis risk, and oxygen transport within oxygenators.</div></div><div><h3>Methods</h3><div>QUADROX-i Adult Oxygenator was investigated using CFD to simulate its hemodynamic environment under different blood flow modes, comprising one non-pulsatile condition and nine pulsatile conditions with varying frequencies and amplitudes. The stasis (ART) and hypercoagulability (C[FXIa]) were used to assess the thrombosis risk and oxygen transport (PO<sub>2</sub>) was also analyzed. The dynamic blood volume (DBV) were calculated to reflect the effective volume within the oxygenator.</div></div><div><h3>Results</h3><div>Under all blood flow modes, the velocity distribution is more uneven in inlet-side and outlet-side transition regions, and the high value is near the inlet and outlet, and becomes lower away from the inlet and outlet. In gas exchange region, the velocity is low and evenly distributed. The region with the highest ART and C[FXIa] are located at the north corner region close to the outlet. The highest PO<sub>2</sub> evenly appears in the region near the outlet. Under pulsatile conditions, As the flow rate increases, the distribution of velocity, ART and C[FXIa] becomes more uneven, and vice verse. Compared to non-pulsatile condition, the period-averaged ART, C[FXIa] and PO<sub>2</sub> become higher, while the DBV decreases under pulsatile conditions. Amplitude has a more significant effect on all parameters than frequency. Higher amplitude results in the higher period-averaged ART, C[FXIa] and PO<sub>2</sub>, alongside a lower DBV.</div></div><div><h3>Conclusions</h3><div>Uneven flow field mainly occurs in the inlet-side and outlet-side transition region, and the uneven degree increases with the higher flow rate, and vice verse. The highest thrombosis risk locates in the north corner region close to the outlet and the highest oxygen transport occurs in the region close to the outlet. Pulsatile flow can enhance oxygen transport but increase thrombosis risk than non-pulsatile flow. Higher amplitude can increase thrombosis risk but improve oxygen transport in the oxygenator. The frequency variation exhibits minimal influence.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"276 ","pages":"Article 109236"},"PeriodicalIF":4.8,"publicationDate":"2026-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145973710","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