Computer methods and programs in biomedicine最新文献

{"title":"Predicting synergistic drug combinations via hierarchical molecular representation and cell line latent space fusion","authors":"Can Bai ,&nbsp;Xianjun Han ,&nbsp;Siqi Li ,&nbsp;Yue Zhang","doi":"10.1016/j.cmpb.2025.108933","DOIUrl":"10.1016/j.cmpb.2025.108933","url":null,"abstract":"<div><div>Cancer treatment often benefits from the synergistic effects of drug combinations. Predicting these synergies is critical for developing effective combination therapies. Existing deep learning models typically represent drugs using a single graph structure and use cell line gene expression profiles directly, potentially leading to loss of detailed molecular features and introduction of noise. This study aims to improve the prediction of drug synergy by developing a novel deep learning model that captures both local and global features of drug molecules at multiple levels and reduces noise in cell line data. The proposed model hierarchically represents drug molecules at the node, motif, and graph levels to capture comprehensive feature information. The Mamba module and graph attention-based convolution are employed to effectively extract deep feature information from drug pairs. An encoder–decoder structure projects cell lines into a latent space, minimizing noise and enhancing the integration with drug pair data through star operations and an attention mechanism. The model was trained and validated on benchmark datasets containing drug response data from various cancer cell lines. The evaluation of the model against benchmark datasets demonstrated superior performance compared to existing methods. These results indicate that the model can more accurately predict synergistic anticancer drug combinations, providing reliable support for the design of combination therapies. The enhanced predictive accuracy can facilitate the discovery of effective drug combinations, potentially accelerating the development of personalized cancer treatments.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108933"},"PeriodicalIF":4.9,"publicationDate":"2025-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144704098","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}

{"title":"Dynamic hypergraph representation for bone metastasis analysis","authors":"Yuxuan Chen ,&nbsp;Jiawen Li ,&nbsp;Lianghui Zhu ,&nbsp;Yang Xu ,&nbsp;Tian Guan ,&nbsp;Huijuan Shi ,&nbsp;Yonghong He ,&nbsp;Anjia Han","doi":"10.1016/j.cmpb.2025.108966","DOIUrl":"10.1016/j.cmpb.2025.108966","url":null,"abstract":"<div><h3>Background and objective:</h3><div>Bone metastasis cancer analysis is a significant challenge in pathology and plays a critical role in determining patient quality of life and treatment strategies. The microenvironment and specific tissue structures are essential for pathologists to predict the primary bone cancer origins and primary bone cancer subtyping. By digitizing bone tissue sections into whole slide images (WSIs) and leveraging deep learning to model slide embeddings, this analysis can be enhanced. However, tumor metastasis involves complex multivariate interactions with diverse bone tissue structures, which traditional WSI analysis methods such as multiple instance learning (MIL) fail to capture. Moreover, graph neural networks (GNNs), limited to modeling pairwise relationships, are hard to represent high-order biological associations.</div></div><div><h3>Methods:</h3><div>In this paper, we propose a dynamic hypergraph neural network (DyHG) to overcome conventional graph limitations by connecting multiple nodes via hyperedges. A learnable hypergraph structure is obtained through nonlinear transformation, while a Gumbel-Softmax sampling strategy optimizes patch distribution across hyperedges. An MIL aggregator then derives a graph-level embedding for downstream tasks.</div></div><div><h3>Results:</h3><div>Two large-scale datasets for primary bone cancer origins and subtyping classification are constructed from real-world bone metastasis scenarios. Extensive experiments show that DyHG outperforms state-of-the-art (SOTA) baselines by up to 1.28%, demonstrating its capability to model complex biological interactions and enhance analysis accuracy.</div></div><div><h3>Conclusion:</h3><div>We believe that the proposed DyHG can provide auxiliary diagnostic information for bone metastasis analysis and has potential for clinical application.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"271 ","pages":"Article 108966"},"PeriodicalIF":4.8,"publicationDate":"2025-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144720924","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}

{"title":"Decoding muscle activity via CNN-LSTM from 3D spatiotemporal EEG","authors":"Golnaz Amiri,&nbsp;Vahid Shalchyan","doi":"10.1016/j.cmpb.2025.108983","DOIUrl":"10.1016/j.cmpb.2025.108983","url":null,"abstract":"<div><div><em>Objective</em>. Reconstructing muscle activity from electromyogram (EMG) data using non-invasive electroencephalogram (EEG) signals could lead to significant advancements in brain-computer interfaces (BCIs). However, extracting muscle-related signals from EEG poses considerable challenges due to the mixed nature of signals captured by EEG sensors from various cortical regions. <em>Approach</em>. This study introduces a new method for estimating muscle activity from non-invasive EEG signals while participants performed the grasp and lift (GAL) task. Envelopes of the delta, theta, alpha, beta, and gamma frequency bands were chosen as EEG features for the decoding models, computed similarly to muscle activity (EMG envelopes). These were converted into three-dimensional spatiotemporal matrices based on EEG electrode locations. A deep learning model, combining convolutional neural networks (CNN) for spatial and long short-term memory (LSTM) network for temporal EEG information extraction, was applied. This model was compared with two linear and nonlinear decoding methods: multivariate linear regression (mLR) and multilayer perceptron (MLP). <em>Main Results</em>. The average ± standard deviation of the normalized root mean square error (nRMSE), coefficient of determination (R²), and correlation coefficient (CC) between the estimated and actual muscle activity of two muscles in five participants were 0.21 ± 0.05, 0.54 ± 0.17, and 0.76 ± 0.10, respectively. The CNN-LSTM model outperformed both mLR and MLP approaches (<em>p</em>-value &lt; 0.016), with higher frequencies proving more effective for decoding. <em>Significance</em>. The proposed model effectively captures nonlinear relationships between brain and muscle activities, indicating its potential to enhance the accuracy and reliability of non-invasive BCIs.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"271 ","pages":"Article 108983"},"PeriodicalIF":4.8,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144723019","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}

{"title":"Numerical estimation of optimal limits of cerebral blood flow and oxygen partial pressure in brain tissue of preterm infants","authors":"Irina Sidorenko ,&nbsp;Andrey Kovtanyuk ,&nbsp;Nadia Feddahi ,&nbsp;Sabine Jahn ,&nbsp;Eva Lück ,&nbsp;Christian Brickmann ,&nbsp;Ursula Felderhoff-Müser ,&nbsp;Marcus Krüger ,&nbsp;Hubert Kerschbaum ,&nbsp;Renée Lampe","doi":"10.1016/j.cmpb.2025.108984","DOIUrl":"10.1016/j.cmpb.2025.108984","url":null,"abstract":"<div><div><em>Background and Objective:</em> For normal brain function, a sufficient supply of oxygen to resident cells is essential. A lack of oxygen delivery can cause irreversible neuronal damage within minutes. This situation is especially critical for preterm infants and can lead to lifelong mental and motor disabilities. Although cerebral blood flow (CBF) and tissue partial pressure of oxygen (PtO₂) provide important information on oxygen delivery and consumption in the brain, appropriate clinically suitable measuring techniques are still lacking. <em>Methods:</em> Mathematical modeling of cerebral oxygen transport is a promising complementary approach for the assessment of PtO₂ in the brain tissue of newborns. A nonlinear model of oxygen transport from capillaries to tissue was coupled with the cerebrovascular model for CBF calculation in preterm infants. The advantage of this mathematical method lies in its ability to estimate CBF and PtO₂ from routinely measured parameters such as mean arterial blood pressure, hematocrit, arterial carbon dioxide and oxygen partial pressure. <em>Results:</em> The combined model was applied to estimate optimal limits of CBF and PtO₂ in the brain tissue of preterm infants with gestational ages between 23 and 32 weeks of pregnancy using optimal values of medical parameters from clinical routine. <em>Conclusions:</em> The calculated optimal values can be used as additional parameters to assess the health condition of patients during the neonatal period.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108984"},"PeriodicalIF":4.9,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144711654","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}

{"title":"An integrated optimization and deep learning pipeline for predicting live birth success in IVF using feature optimization and transformer-based models","authors":"Arezoo Borji ,&nbsp;Hossam Haick ,&nbsp;Birgit Pohn ,&nbsp;Antonia Graf ,&nbsp;Jana Zakall ,&nbsp;S M Ragib Shahriar Islam ,&nbsp;Gernot Kronreif ,&nbsp;Daniel Kovatchki ,&nbsp;Heinz Strohmer ,&nbsp;Sepideh Hatamikia","doi":"10.1016/j.cmpb.2025.108979","DOIUrl":"10.1016/j.cmpb.2025.108979","url":null,"abstract":"<div><div>The complicated interplay of clinical, demographic, and procedural factors makes it difficult to predict the success of in vitro fertilization (IVF), a commonly used assisted reproductive technology. The goal of this research was to create an artificial intelligence (AI) pipeline that could predict live birth outcomes in IVF treatments with high accuracy.</div></div><div><h3>Design</h3><div>We evaluated prediction performance by integrating different feature selection methods, such as principal component analysis (PCA) and particle swarm optimization (PSO), with different machine learning-based classifiers, including random forest (RF) and decision tree, as well as deep learning-based classifiers, including a custom transformer-based model and a Tab_transformer model with an attention mechanism. Additionally, this study analyzes confounding factors like patient age and previous IVF cycles and explores the influence of different perturbation and preprocessing techniques and validates the model’s robustness under varied scenarios. In addition, Shapley Additive Explanations (SHAP) analysis was performed to enhance interpretability of methods.</div></div><div><h3>Results</h3><div>This research demonstrated that the best performance was achieved by combining PSO for feature selection with the Tab_transformer-based deep learning model, yielding an accuracy of 97 % and an AUC of 98.4 %, highlighting its significant performance in prediction live births. By identifying the most significant predictors of infertility and guaranteeing clinical significance, SHAP analysis significantly improved interpretability.</div></div><div><h3>Conclusion</h3><div>With the accuracy and interpretability, this study develops a strong AI pipeline for predicting live birth outcomes in IVF. This study establishes a highly accurate AI pipeline for predicting live birth outcomes in IVF, demonstrating its potential to enhance personalized fertility treatments.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"271 ","pages":"Article 108979"},"PeriodicalIF":4.8,"publicationDate":"2025-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144720923","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}

{"title":"Interpretable framework for predicting preoperative cardiorespiratory fitness using wearable data","authors":"Iqram Hussain ,&nbsp;Julianna Zeepvat ,&nbsp;M․Cary Reid ,&nbsp;Sara Czaja ,&nbsp;Kane Pryor ,&nbsp;Richard Boyer","doi":"10.1016/j.cmpb.2025.108980","DOIUrl":"10.1016/j.cmpb.2025.108980","url":null,"abstract":"<div><h3>Objectives</h3><div>Predicting preoperative cardiorespiratory fitness (CRF) is crucial for assessing the risk of complications and adverse outcomes in patients undergoing surgery. CRF is formally evaluated through submaximal exercise testing with cardiopulmonary exercise testing (CPET) or the 6-minute walk test (6MWT). However, formal CRF testing is impractical as a preoperative screening tool. Wrist-worn devices with actigraphy and heart rate monitoring have become increasingly capable of predicting physiological measurements. Our aim was to develop a clinically interpretable machine learning (ML) model using wearable-derived physiological data to predict CRF for older adults, and to access whether this model can accurately estimate the 6MWT distances for preoperative risk evaluation.</div></div><div><h3>Methods</h3><div>We examined heart rate and activity data collected from Fitbit devices worn by older adults (<em>N</em> = 65) who were scheduled to undergo major noncardiac surgery. Data collection took place over a 1-week period prior to surgery while participants engaged in their typical daily activities. Our primary aim was to leverage this wearable technology to forecast CRF among this group. We employed a machine-learning ensemble regression model to predict CRF, using 6MWT outcomes as an index. Further, we applied the shapley feature attribution approach to gain insights into how specific features derived from wearable data contribute to CRF prediction within the model, aiding in personalized fitness prediction.</div></div><div><h3>Results</h3><div>Adults with higher CRF exhibited elevated levels of moderate-to-vigorous physical activity (MVPA), maximal activity energy expenditure (aEE_max), heart rate recovery (HRR), and non-linear heart rate variability (HRV). These measures increased concurrently with improvements in 6MWT outcomes. Our regression models, employing random forest and linear regression techniques, demonstrated strong predictive capabilities, with coefficient of determination values of 0.91 and 0.81, respectively, for estimating CRF. The shapley feature attribution approach elucidated those greater levels of MVPA, aEE_max, HRR, and nonlinear dynamics of HRV serve as reliable indicators of enhanced CRF test performance.</div></div><div><h3>Conclusion</h3><div>The integration of wearable data-driven activity and heart rate metrics forms the basis for utilizing wearables to provide preoperative cardiorespiratory fitness assessments, supporting surgical risk stratification, personalized prehabilitation, and improved patient outcomes.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"271 ","pages":"Article 108980"},"PeriodicalIF":4.9,"publicationDate":"2025-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144704707","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}

{"title":"Personalized and uncertainty-aware coronary hemodynamics simulations: From Bayesian estimation to improved multi-fidelity uncertainty quantification","authors":"Karthik Menon ,&nbsp;Andrea Zanoni ,&nbsp;M. Owais Khan ,&nbsp;Gianluca Geraci ,&nbsp;Koen Nieman ,&nbsp;Daniele E. Schiavazzi ,&nbsp;Alison L. Marsden","doi":"10.1016/j.cmpb.2025.108951","DOIUrl":"10.1016/j.cmpb.2025.108951","url":null,"abstract":"<div><h3>Background:</h3><div>Non-invasive simulations of coronary hemodynamics have improved clinical risk stratification and treatment outcomes for coronary artery disease, compared to relying on anatomical imaging alone. However, simulations typically use empirical approaches to distribute total coronary flow amongst the arteries in the coronary tree, which ignores patient variability, the presence of disease, and other clinical factors. Further, uncertainty in the clinical data often remains unaccounted for in the modeling pipeline.</div></div><div><h3>Objective:</h3><div>We present an end-to-end uncertainty-aware pipeline to (1) personalize coronary flow simulations by incorporating vessel-specific coronary flows as well as cardiac function; and (2) predict clinical and biomechanical quantities of interest with improved precision, while accounting for uncertainty in the clinical data.</div></div><div><h3>Methods:</h3><div>We assimilate patient-specific measurements of myocardial blood flow from clinical CT myocardial perfusion imaging to estimate branch-specific coronary artery flows. Simulated noise in the clinical data is used to estimate the joint posterior distributions of the model parameters using adaptive Markov Chain Monte Carlo sampling. Additionally, the posterior predictive distribution for the relevant quantities of interest is determined using a new approach combining multi-fidelity Monte Carlo estimation with non-linear, data-driven dimensionality reduction. This leads to improved correlations between high- and low-fidelity model outputs.</div></div><div><h3>Results:</h3><div>Our framework accurately recapitulates clinically measured cardiac function as well as branch-specific coronary flows under measurement noise uncertainty. We observe substantial reductions in confidence intervals for estimated quantities of interest compared to single-fidelity Monte Carlo estimation and state-of-the-art multi-fidelity Monte Carlo methods. This holds especially true for quantities of interest that showed limited correlation between the low- and high-fidelity model predictions. In addition, the proposed multi-fidelity Monte Carlo estimators are significantly cheaper to compute than traditional estimators, under a specified confidence level or variance.</div></div><div><h3>Conclusions:</h3><div>The proposed pipeline for personalized and uncertainty-aware predictions of coronary hemodynamics is based on routine clinical measurements and recently developed techniques for CT myocardial perfusion imaging. The proposed pipeline offers significant improvements in precision and reduction in computational cost.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"271 ","pages":"Article 108951"},"PeriodicalIF":4.8,"publicationDate":"2025-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144720886","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}

{"title":"Multi-scale simulation of red blood cell trauma in large-scale high-shear flows after Norwood operation","authors":"Saba Mansour ,&nbsp;Emily Logan ,&nbsp;James F. Antaki ,&nbsp;Mahdi Esmaily","doi":"10.1016/j.cmpb.2025.108947","DOIUrl":"10.1016/j.cmpb.2025.108947","url":null,"abstract":"<div><h3>Background and objective:</h3><div>Cardiovascular surgeries and mechanical circulatory support devices create non-physiological blood flow conditions that can be detrimental, especially for pediatric patients. A source of complications is mechanical red blood cell (RBC) damage induced by localized supraphysiological shear fields. To understand such complications in single ventricle patients, we introduce a multi-scale numerical model to predict hemolysis risk in idealized anatomies.</div></div><div><h3>Methods:</h3><div>We employed our in-house CFD solver coupled with Lagrangian tracking and cell-resolved fluid–structure interaction to measure flow-induced stresses and strains on the RBC membrane. The Norwood procedure, known for its high mortality rate, is selected for its importance to single-ventricle population survival. We simulated three anatomies including 2.5 mm and 4.0 mm diameter modified Blalock–Taussig shunts (mBTS) and a 2.5 mm central shunt (CS), with hundreds of RBCs per case for statistical analysis.</div></div><div><h3>Results:</h3><div>The results show that the conditions created by these surgeries can elongate RBCs by more than two-fold (3.1% of RBCs for 2.5 mm mBTS, 1.4% for 4 mm mBTS, and 8.8% for CS). Shear and areal strain metrics also reveal that CS creates the greatest deformations on the RBC membrane. These conclusions are further confirmed when strain history and different damage thresholds are considered.</div></div><div><h3>Conclusions:</h3><div>The central shunt is more hemolytic in comparison to the modified Blalock–Taussig shunt. Between the two mBTSs, the smaller diameter is slightly more prone to hemolysis. Spatial damage maps produced based on the studied metrics, highlighted hot zones that match the clinical images of shunt thrombosis, demonstrating their potential to enhance cardiac surgery outcomes.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"271 ","pages":"Article 108947"},"PeriodicalIF":4.8,"publicationDate":"2025-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144768438","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}

{"title":"Regularized multi-task learning with individual-feature-based task correlations for Alzheimer’s cognitive score prediction","authors":"Shanshan Tang ,&nbsp;Qi Chen ,&nbsp;Bing Xue ,&nbsp;Min Huang ,&nbsp;Mengjie Zhang","doi":"10.1016/j.cmpb.2025.108954","DOIUrl":"10.1016/j.cmpb.2025.108954","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Predicting multiple cognitive scores from brain features for Alzheimer’s disease (AD) patients can aid in early intervention treatments and enhance disease management. Regularized multi-task sparse learning has become an important approach since it can predict multiple cognitive scores and identify biomarkers in one process. However, existing methods often assign the same correlation coefficient for a pair of tasks on all features, even though their relationships at different features are usually different. Introducing inaccurate task correlations into multi-task learning can hinder the improvement of models’ prediction performance. This study overcomes the above limitation by introducing a novel multi-task learning framework that captures task correlations at a fine-grained, feature-specific level.</div></div><div><h3>Methods:</h3><div>We propose a novel individual-feature-based task correlation matrices guided multi-task learning (IFTMTL) method. The method constructs a non-smooth convex objective function that jointly learns regression models for multiple cognitive scores. This objective function integrates task and feature correlations to enhance predictive performance. Specifically, the fine-grained inter-task correlations are modeled at the feature level using a set of task correlation matrices, while feature correlations are captured via the Pearson coefficient. An iterative optimization algorithm is developed to jointly update the task correlation structures and model parameters.</div></div><div><h3>Results:</h3><div>The proposed IFTMTL significantly outperforms the 11 competitive methods in the normalized mean squared error (nMSE) and correlation coefficient (CC) metrics. Specifically, IFTMTL improves the nMSE metric of the Multi-Task Feature Learning method by 4.09%, and the CC metric by 1.68%. Furthermore, IFTMTL identifies the most severely affected brain regions in AD, including the left hippocampus, left middle temporal, and right entorhinal.</div></div><div><h3>Conclusion:</h3><div>IFTMTL improves AD prediction performance by effectively incorporating unequal task correlations at the feature level, enabling better knowledge transfer across tasks. The method outperforms existing approaches in predicting cognitive scores and identifying significant biomarkers, such as the hippocampus and middle temporal regions, which are crucial for AD prediction and clinical analysis.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108954"},"PeriodicalIF":4.9,"publicationDate":"2025-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144680521","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}

{"title":"EndoUSScan: Keyframe detection in transvaginal ultrasound imaging for measuring endometrial thickness","authors":"Yiyang Liu ,&nbsp;Boyuan Peng ,&nbsp;Qin Zhou ,&nbsp;Suzhen Yuan ,&nbsp;Wei Yan ,&nbsp;Li Fang ,&nbsp;Jingjing Jiang ,&nbsp;Shixuan Wang ,&nbsp;Xin Zhu ,&nbsp;Wenwen Wang","doi":"10.1016/j.cmpb.2025.108957","DOIUrl":"10.1016/j.cmpb.2025.108957","url":null,"abstract":"<div><h3>Background:</h3><div>Accurate measurement of endometrial thickness (ET) using transvaginal ultrasound (TVUS) imaging is essential for diagnosing various gynecological conditions. However, manual ET measurement remains challenging, especially for junior physicians, due to variability in image quality and patient characteristics.</div></div><div><h3>Methods:</h3><div>A prospective observational study was performed using a dataset of 976 uterine ultrasound videos (82,063 images) measured in 2014-2019 in Tongji Hospital, Huazhong University of Science and Technology. We developed EndoUSScan, a comprehensive system for automated image selection and keyframe identification. The system incorporates MSNet, an improved DenseNet169-based system, to select candidate images with accurate endometrial representation. We also designed a keyframe detection system to assist junior medical staff in identifying frames with the largest ET from the candidate images. Comparative evaluations involved six junior sonographers, who assessed both speed and accuracy.</div></div><div><h3>Findings:</h3><div>MSNet achieved an accuracy of 94.7% and a specificity of 96.7% in selecting candidate images, outperforming conventional models including ResNet50, ResNet101, DenseNet121, and DenseNet169. The automatically selected keyframes were consistent with the expert-defined gold standard. Compared with manual procedures by junior sonographers, EndoUSScan significantly improved both the speed and accuracy of keyframe selection.</div></div><div><h3>Interpretation:</h3><div>This study presents the first fully automated and clinically validated system for keyframe detection in TVUS videos to support ET measurement. By standardizing the image selection process and assisting junior sonographers, EndoUSScan enhances diagnostic efficiency and accuracy, ultimately contributing to improved patient care.</div></div><div><h3>Funding:</h3><div>This study was funded by the <span>National Key Research and Development Program of China</span> (grant number <span><span>2022YFC2704100)</span></span> and <span>Knowledge Innovation Program of Wuhan -Basic Research</span> (No. <span><span>2023020201010041</span></span>).</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"271 ","pages":"Article 108957"},"PeriodicalIF":4.9,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144713983","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}