Computer methods and programs in biomedicine最新文献

{"title":"Early detection of Multidrug Resistance using Multivariate Time Series analysis and interpretable patient-similarity representations","authors":"Óscar Escudero-Arnanz ,&nbsp;Antonio G. Marques ,&nbsp;Inmaculada Mora-Jiménez ,&nbsp;Joaquín Álvarez-Rodríguez ,&nbsp;Cristina Soguero-Ruiz","doi":"10.1016/j.cmpb.2025.108920","DOIUrl":"10.1016/j.cmpb.2025.108920","url":null,"abstract":"<div><h3>Background and Objectives:</h3><div>Multidrug Resistance has been identified by the World Health Organization as a major global health threat. It leads to severe social and economic consequences, including extended hospital stays, increased healthcare costs, and higher mortality rates. In response to this challenge, this study proposes a novel interpretable Machine Learning (ML) approach for predicting MDR, developed with two primary objectives: accurate inference and enhanced explainability.</div></div><div><h3>Methods:</h3><div><em>For inference</em>, the proposed method is based on patient-to-patient similarity representations to predict MDR outcomes. Each patient is modeled as a Multivariate Time Series (MTS), capturing both clinical progression and interactions with similar patients. To quantify these relationships, we employ MTS-based similarity metrics, including feature engineering using descriptive statistics, Dynamic Time Warping, and the Time Cluster Kernel. These methods are used as inputs for MDR classification through Logistic Regression, Random Forest, and Support Vector Machines, with dimensionality reduction and kernel transformations applied to enhance model performance. <em>For explainability</em>, we employ graph-based methods to extract meaningful patterns from the data. Patient similarity networks are generated using the MTS-based similarity metrics mentioned above, while spectral clustering and t-SNE are applied to identify MDR-related subgroups, uncover clinically relevant patterns, and visualize high-risk clusters. These insights improve interpretability and support more informed decision-making in critical care settings.</div></div><div><h3>Results:</h3><div>We validate our architecture on real-world Electronic Health Records from the Intensive Care Unit (ICU) dataset at the University Hospital of Fuenlabrada, achieving a Receiver Operating Characteristic Area Under the Curve of 81%. Our framework surpasses ML and deep learning models on the same dataset by leveraging graph-based patient similarity. In addition, it offers a simple yet effective interpretability mechanism that facilitates the identification of key risk factors—such as prolonged antibiotic exposure, invasive procedures, co-infections, and extended ICU stays—and the discovery of clinically meaningful patient clusters. For transparency, all results and code are available at <span><span>https://github.com/oscarescuderoarnanz/DM4MTS</span><svg><path></path></svg></span>.</div></div><div><h3>Conclusions:</h3><div>This study demonstrates the effectiveness of patient similarity representations and graph-based methods for MDR prediction and interpretability. The approach enhances prediction, identifies key risk factors, and improves patient stratification, enabling early detection and targeted interventions, highlighting the potential of interpretable ML in critical care.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108920"},"PeriodicalIF":4.9,"publicationDate":"2025-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144633137","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":"ExtractPDF: A data extraction tool for scientific papers applied to a systematic scoping review in public health","authors":"Gunn E. Vist ,&nbsp;Trine Husøy ,&nbsp;Michael Guy Diemar ,&nbsp;Hubert Dirven ,&nbsp;Erwin L. Roggen ,&nbsp;Maria E. Kalyva","doi":"10.1016/j.cmpb.2025.108962","DOIUrl":"10.1016/j.cmpb.2025.108962","url":null,"abstract":"<div><h3>Background and Objectives</h3><div>Systematic reviews are widely used to identify the evidence and get an overview of the available knowledge for various questions related to public health and medical topics. They can provide a summary of all the available data and can be used to make knowledge-based decisions about policy, practice, and academic research. The conduct of systematic reviews can often be time‐consuming and costly.</div></div><div><h3>Methods</h3><div>We have developed a command-line based code in R to extract data in an automated manner from full-text scientific papers. ExtractPDF is a data extraction tool/software that provides a reliable computational workflow for extracting words or combinations of words from numerous portable document format (PDF) files.</div></div><div><h3>Results</h3><div>The software was applied to extract information from 299 papers that have been screened as included for a published systematic scoping review study within the field of risk assessment in public health. The output of the software is tables of extracted information per type of information of interest per PDF file. The tables were used during the data extraction stage as a second reviewer alongside a human reviewer to assist and/or validate data extraction items.</div></div><div><h3>Conclusions</h3><div>ExtractPDF tool has a novel pipeline architecture to automate extraction of information from unstructured format types, such as PDF files. ExtractPDF tool assisted in expediting the task of data extraction stage and reducing human related resources as well as errors. The tool’s performance and reliability were found to be very good with metrics of averagely 0.89 for precision, 0.92 for recall, 0.86 for accuracy and 0.91for F1-score.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108962"},"PeriodicalIF":4.9,"publicationDate":"2025-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144633138","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":"Malignant melanoma fractional-order mathematical model with stabilized fuzzy sliding mode control","authors":"David Amilo,&nbsp;Khadijeh Sadri,&nbsp;Evren Hincal","doi":"10.1016/j.cmpb.2025.108912","DOIUrl":"10.1016/j.cmpb.2025.108912","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Malignant melanoma, an aggressive form of skin cancer, poses significant challenges due to its rapid progression, metastatic potential, and resistance to therapies. This study aims to develop a fractional-order mathematical model capturing melanoma dynamics (tumor-immune interactions, extracellular matrix remodeling, nutrient dynamics) and introduce a Stabilized Fuzzy Sliding Mode Control (SFSMC) strategy to suppress tumor growth and restore microenvironmental homeostasis.</div></div><div><h3>Methods:</h3><div>A fractional-order model was derived using Caputo derivatives to incorporate memory effects and long-range dependencies. The SFSMC combines sliding mode control with fuzzy logic to manage uncertainties. Theoretical analysis included well-posedness, stability (via Lyapunov functions), and computation of the reproduction number <span><math><msub><mrow><mi>R</mi></mrow><mrow><mn>0</mn></mrow></msub></math></span>. Numerical simulations were performed using a predictor–corrector method with parameters calibrated from clinical data.</div></div><div><h3>Results:</h3><div>The model demonstrated stability when <span><math><mrow><msub><mrow><mi>R</mi></mrow><mrow><mn>0</mn></mrow></msub><mo>&lt;</mo><mn>1</mn></mrow></math></span>, indicating tumor suppression. SFSMC reduced tumor cell populations by 78% and circulating tumor cells by 65% while improving immune response (45% increase in immune cells) and nutrient availability (30% recovery). Sensitivity analysis revealed <span><math><msub><mrow><mi>R</mi></mrow><mrow><mn>0</mn></mrow></msub></math></span> is mostly influenced by tumor growth rate, natural degradation rate of extracellular matrix (ECM), rate of ECM degradation by tumor cells, and ECM production rate, suggesting their potential role in suppressing tumor growth.</div></div><div><h3>Conclusions:</h3><div>The fractional-order framework and SFSMC offer a robust approach to modeling and controlling melanoma, with potential clinical implications for adaptive therapy.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108912"},"PeriodicalIF":4.9,"publicationDate":"2025-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144613865","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":"Handwritten signature verification using a wearable surface-EMG armband","authors":"Jing Zheng ,&nbsp;Minwei Zhou ,&nbsp;Zhehao Zhou ,&nbsp;Jieyi Ge ,&nbsp;Hang Chen ,&nbsp;Xiaobai Li ,&nbsp;Wanlin Chen ,&nbsp;Shulin Chen","doi":"10.1016/j.cmpb.2025.108908","DOIUrl":"10.1016/j.cmpb.2025.108908","url":null,"abstract":"<div><div>The growing demand for remote authentication underscores the importance of robust signature verification systems. A major challenge in this domain is the substantial intra-class variability inherent in handwritten signatures. This study investigates the use of surface electromyography (sEMG) for signature verification through wearable armbands, aiming to address this issue. We introduce a dual-model deep learning framework that integrates muscle co-activation patterns with raw sEMG signal waveforms. A 4-channel armband was employed to collect sEMG data from 20 individuals signing Chinese characters, resulting in the first sEMG signature dataset centered on wearable acquisition. Experimental results show that conventional feature-based machine learning methods are limited in performance, yielding 80.90% accuracy and a 12.82% equal error rate (EER), primarily due to high intra-class variability. The proposed framework comprises: (1) a CNN-LSTM architecture that processes encoded muscle activation sequences, and (2) a multi-branch CNN designed to learn from raw sEMG signals. Fusion at the decision level between these models achieves 91.65% accuracy and 5.25% EER, reflecting a 10.75% improvement in accuracy compared with traditional techniques. These findings confirm the effectiveness of the proposed approach in reducing intra-class variability while preserving the usability of wearable devices, offering a practical and secure biometric authentication solution.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108908"},"PeriodicalIF":4.9,"publicationDate":"2025-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144623567","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":"Design a bistable polymeric vascular stent (BPVS) and evaluate the biomechanical properties","authors":"Chen Pan ,&nbsp;Zhifang Fan ,&nbsp;Jingjing Cao ,&nbsp;Hezong Li","doi":"10.1016/j.cmpb.2025.108960","DOIUrl":"10.1016/j.cmpb.2025.108960","url":null,"abstract":"<div><h3>Background and objectives</h3><div>Polymeric vascular stents generally have the disadvantages of poor biomechanical properties, which may not achieve the therapeutic purpose of supporting the blocked vascular vessels to restore normal blood flow. The bistable structure depending on the two stable configurations seems to improve the weak strength of stents. This paper mainly designs a polymeric vascular stent with bistable structure to enhance the radial force, and reduce the radial recoil and wall shear stress.</div></div><div><h3>Methods</h3><div>The bistable stents were derived from the bistable property of the tilted strut and the planar cell systematically. The mapping relationship between the tilted struts with different geometries and the bistable performance was revealed by finite element method, and then the bistable characteristics of the planar cells were further explored. Furthermore, the biomechanical performance involving radial force and radial recoil of bistable polymeric stents, and wall shear stress of vascular vessels were analyzed and evaluate by combining numerical simulation and experiments.</div></div><div><h3>Results</h3><div>The mapping relation between geometries and bistable properties of tilted struts was that the (<em>t/L, θ</em>) = (0.03, 10° ∼ 60°), and (<em>t/L, θ</em>) = (0.03 ∼ 0.1, 30° ∼ 40°) were the widest ranges of optional parameters. When the bistable evaluation factor <em>B</em> = <em>T</em>/<em>t</em> ≥ 4, the BREH cells had outstanding bistable properties. The finite element results of polymeric stents indicated that the bistable structure obviously greatened the radial force (2.52 N), and lessened the radial recoil (1.69 %) of the polymeric stent. Besides, the bistable structure minified the wall shear stress of vascular vessels to 0.04177 MPa.</div></div><div><h3>Conclusions</h3><div>It could be concluded that the bistable structure not only endowed polymeric stents with strong biomechanical properties, but also reduces the risk of secondary injury after its being implanted into vascular vessels. The bistable polymeric stents have the potential to support the blocked vessels and restore the blood flow.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108960"},"PeriodicalIF":4.9,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144632497","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":"Computational methods used to investigate atherosclerosis progression in coronary arteries: structural FEA, CFD or FSI","authors":"Vittorio Lissoni ,&nbsp;Giulia Luraghi ,&nbsp;Marco Stefanati ,&nbsp;Jose Felix Rodriguez Matas ,&nbsp;Francesco Migliavacca","doi":"10.1016/j.cmpb.2025.108959","DOIUrl":"10.1016/j.cmpb.2025.108959","url":null,"abstract":"<div><h3>Background and objectives</h3><div>In recent years, computational simulations have emerged as valuable tools for the evaluation of atherosclerosis progression in coronary anatomies, although only a few studies have utilized more realistic Fluid-Structure Interaction (FSI) simulations. This work aims to compare the results of Computational Fluid Dynamics (CFD), Structural Finite Element Analysis (structural FEA) and FSI simulations in order to assess differences in plaque progression indices estimation.</div></div><div><h3>Methods</h3><div>We performed structural FEA, CFD and FSI on five patient-specific epicardial coronary anatomies using the commercial software LS-Dyna. To account for the vessel pre-stress, the zero-pressure configuration was calculated for each anatomy with an inverse elastostatic algorithm. CFD, structural FEA and FSI simulations were performed applying boundary conditions based on physiological values.</div></div><div><h3>Results</h3><div>The comparison between structural FEA and FSI showed similar stress distribution and vessel expansions, with differences found only in the distal parts of the coronaries, where pressure reduction due to pressure loss affects the vessel walls. The elastic walls of the coronaries impact blood flow, resulting in a more disturbed flow. However, time averaged wall shear stress (TAWSS) and oscillatory shear index (OSI) distributions are similar across each coronary between CFD and FSI; TAWSS is slightly higher in CFD while OSI peaks are higher in FSI.</div></div><div><h3>Conclusion</h3><div>In conclusion, given the significantly higher computational costs of FSI, we believe that CFD and structural FEA offer a more practical and cost-effective approach, providing results comparable to those of FSI, making them preferable options.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108959"},"PeriodicalIF":4.9,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144614151","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":"Modeling changes in genetic heterogeneity using games with resources","authors":"Katarzyna Hajdowska ,&nbsp;Andrzej Swierniak ,&nbsp;Damian Borys","doi":"10.1016/j.cmpb.2025.108916","DOIUrl":"10.1016/j.cmpb.2025.108916","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>This study explores an extension of the classic Hawk and Dove evolutionary game model by considering the influence of environmental or external resources on the players’ fitness. This allows us to model the resulting heterogeneous population dynamics, which is of great importance for simulating cancer population growth and optimizing anti-cancer therapies.</div></div><div><h3>Methods:</h3><div>To model population heterogeneity, we are using an extension of classical spatial evolutionary game theory by introducing multidimensional spatial evolutionary games (MSEG). This allows for the study of genetic heterogeneity on a multidimensional lattice. The classic Hawk and Dove model is modified to reflect the impact of external resources. Various types and shapes of resource functions were included in the payoff matrix and then simulated to examine their impact on the model’s dynamics and population heterogeneity.</div></div><div><h3>Results:</h3><div>The results are presented in time-dependent plots for both mean-field and spatial models. Additionally, spatial 2D and 3D matrices are presented to show the spatial distribution of both phenotypes analyzed in the extended Hawk and Dove model. The results reveal significant differences between the mean-field and spatial models for the same parameter values. Furthermore, differences are observed when comparing models with different resource functions.</div></div><div><h3>Conclusion:</h3><div>The two-phenotype model was used to show the influence of external, time- and phenotype-specific resource functions on the dynamics of the game’s phenotypes. Moreover, the study highlights that spatial models, which provide more accurate information about population heterogeneity, can yield significantly different results compared to mean-field models.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108916"},"PeriodicalIF":4.9,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144613864","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":"Artificial intelligence in forensic pathology: Multi-organ postmortem pathomics for estimating postmortem interval","authors":"Guoshuai An ,&nbsp;Yu Gao ,&nbsp;Siyuan Cheng ,&nbsp;Na Li ,&nbsp;Kang Ren ,&nbsp;Qiuxiang Du ,&nbsp;Rufeng Bai ,&nbsp;Junhong Sun","doi":"10.1016/j.cmpb.2025.108949","DOIUrl":"10.1016/j.cmpb.2025.108949","url":null,"abstract":"<div><h3>Background</h3><div>Accurate estimation of the postmortem interval is crucial in forensic investigations. Pathomics presents a promising advancement by leveraging whole-slide images as a novel data modality for the diagnosis and prognosis of diseases in clinical situations. The extended application of this technology in forensic postmortem image analysis is expected to give rise to postmortem pathomics as an important subfield.</div></div><div><h3>Objective</h3><div>This study aimed to develop a three-level hierarchical strategy using pathomics to analyze postmortem histological images data, develop multi-organ integrated model for the postmortem interval estimation, and lay the foundation for postmortem pathomics.</div></div><div><h3>Methods</h3><div>Twelve Bama miniature pigs were euthanized, and liver, kidney, and skeletal muscle tissues were collected at 6, 24, 48, and 96 h postmortem. Hematoxylin and eosin stained whole slide images were divided into 512 × 512 pixel patches. Low-quality patches were excluded using Otsu thresholding, and color normalization was applied using the Vahadane algorithm to minimize staining variability. Deep learning models were trained on patch-level data using transfer learning and evaluated for interpretability with Grad-CAM. Slide-level predictions were obtained via organ-specific deep feature aggregation and machine learning models, while a multi-organ integrated model was developed using a stacking ensemble combining above machine learning models and a logistic regression. Four additional pigs were introduced for external validation at the multi-organ integrated individual-level.</div></div><div><h3>Results</h3><div>DenseNet121 demonstrated superior performance for liver and kidney, while VGG16 performed best for skeletal muscle tissue. These models were designated as postmortem-liver-net, postmortem-kidney-net, and postmortem-muscle-net, respectively, and employed to extract pathomics features from images. Slide-level models trained on these features vectors achieved accuracies of 81.25% (liver), 87.5% (kidney), and 62.5% (muscle). A stacking model integrating probability output of these three slide-level models achieved internal and external test accuracies at multi-organ integrated individual-level of 93.75% and 87.5%, respectively.</div></div><div><h3>Conclusion</h3><div>This study demonstrated the potential of pathomics and deep learning for postmortem interval estimation. The proposed three-level framework effectively integrated multi-organ features, introducing whole-slide images as a novel modality and offering an innovative strategy for postmortem interval estimation.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108949"},"PeriodicalIF":4.9,"publicationDate":"2025-07-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144596151","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":"Retrieval-based adaptive fusion strategy for medical report generation","authors":"Yingying Huang ,&nbsp;Yang Si ,&nbsp;Bingliang Hu ,&nbsp;Jiang Shen ,&nbsp;Linshen Xie ,&nbsp;Dongsheng Wu ,&nbsp;Quan Wang","doi":"10.1016/j.cmpb.2025.108907","DOIUrl":"10.1016/j.cmpb.2025.108907","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Retrieval-based medical report generation methods attempt to improve efficiency by reusing historical reports, but their fixed feature-concatenation strategies often introduce cross-case redundancy. Moreover, most methods are designed for low-resolution X-ray images, and their evaluation metrics rely on textual similarity and overlook the implications of misdescription. To address these, we first constructed a high-resolution CT-report dataset, comprising 9 categories of chest CT scans and corresponding reports from 505 patients. Then, we propose RAFS, a retrieval-based adaptive fusion strategy, to dynamically balance contributions from generation and retrieval modules. Finally, we propose DICE, a dual-perspective integrated clinical evaluation including consensus-based positive scoring and penalties of misdescription.</div></div><div><h3>Methods:</h3><div>RAFS integrates an attention module to calculate the similarity between the current generated word’s hidden state and the retrieved text, passing the result through a fully connected layer to obtain retrieval probabilities. After, obtained attention weights are feed the Sigmoid function and its result for fusing the generation probabilities and retrieval probabilities.</div></div><div><h3>Results:</h3><div>RAFS achieves superior performance with BLEU-4, METEOR, ROUGE_L, CIDEr and the average of DICE scores of 45.8, 32.9, 59.1, 79.3 and 64.6 in the CT report generation task, outperforming existing methods. methods.</div></div><div><h3>Conclusion:</h3><div>RAFS significantly enhances the clinical interpretability of generated reports, with future work dedicated to optimizing the characterization of local pathological lesions.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108907"},"PeriodicalIF":4.9,"publicationDate":"2025-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144579665","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":"TAC-ECG: A task-adaptive classification method for electrocardiogram based on cross-modal contrastive learning and low-rank convolutional adapter","authors":"Rongjia Wang ,&nbsp;Xunde Dong ,&nbsp;Xiuling Liu ,&nbsp;Yihai Fang ,&nbsp;Jianhong Dou ,&nbsp;Yupeng Qiang ,&nbsp;Yang Yang ,&nbsp;Fei Hu","doi":"10.1016/j.cmpb.2025.108918","DOIUrl":"10.1016/j.cmpb.2025.108918","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Cardiovascular diseases are one of the major health threats to humans. Researchers have proposed numerous deep learning-based methods for the automatic analysis of electrocardiogram (ECG), achieving encouraging results. However, many existing methods are limited to task-specific model training and require retraining or full fine-tuning when confronted with a new ECG classification task, thus lacking flexibility in clinical applications.</div></div><div><h3>Methods:</h3><div>In this study, we propose a <strong>T</strong>ask-<strong>A</strong>daptive <strong>C</strong>lassification method for ECG (TAC-ECG) based on cross-modal contrastive learning and low-rank convolutional adapters. TAC-ECG comprises two main phases. In the first phase, inspired by the Contrastive Language-Image Pre-training, we design the <strong>C</strong>ontrastive <strong>E</strong>CG-<strong>T</strong>ext <strong>P</strong>re-training (CETP) to pre-train a robust ECG encoder. In the second phase, the pre-trained ECG encoder is frozen and integrated with a lightweight plug-in, the <strong>L</strong>ow-<strong>R</strong>ank <strong>C</strong>onvolutional Adapter (LRC-Adapter), forming an extensible ECG classification model. The frozen encoder extracts more discriminative features from the ECG signal, while the LRC-Adapter enables task-specific adaptation. For diverse ECG classification tasks, TAC-ECG only requires training the LRC-Adapter. This mechanism enables TAC-ECG to efficiently perform different ECG classification tasks, significantly reducing resource consumption and deployment costs in multi-tasking scenarios compared to traditional fully fine-tuned methods.</div></div><div><h3>Results:</h3><div>We conducted extensive experiments using six different network architectures as ECG encoders. Specifically, we performed ECG classification experiments on four datasets: CPSC2018, Cinc2017, PTB-XL, and Chapman, targeting 9-category, 3-category, 5-category, and 4-category classifications respectively. The TAC-ECG achieved highly competitive results using only approximately 3% of the trainable parameters and approximately 25% of the total parameters compared to the fully fine-tuned method. These results demonstrates the effectiveness and practicality of the TAC-ECG method.</div></div><div><h3>Conclusion:</h3><div>The TAC-ECG offers a flexible and efficient method for ECG classification, enabling rapid adaptation to diverse tasks and enhancing clinical diagnostic practicality.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"270 ","pages":"Article 108918"},"PeriodicalIF":4.9,"publicationDate":"2025-07-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144587664","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}