Computers in biology and medicine最新文献

{"title":"Time-series deep learning and conformal prediction for improved sepsis diagnosis in primarily Non-ICU hospitalized patients","authors":"Shaunak Dalal ,&nbsp;Ahad Khaleghi Ardabili ,&nbsp;Anthony S. Bonavia","doi":"10.1016/j.compbiomed.2025.110497","DOIUrl":"10.1016/j.compbiomed.2025.110497","url":null,"abstract":"<div><h3>Purpose</h3><div>Sepsis, a life-threatening condition from an uncontrolled immune response to infection, is a leading cause of in-hospital mortality. Early detection is crucial, yet traditional diagnostic methods, like SIRS and SOFA, often fail to identify sepsis in non-ICU settings where monitoring is less frequent. Recent machine learning models offer new possibilities but lack generalizability and suffer from high false alarm rates.</div></div><div><h3>Methods</h3><div>We developed a deep learning (DL) model tailored for non-ICU environments, using MIMIC-IV data with a conformal prediction framework to handle uncertainty. The model was trained on 83,813 patients and validated with the eICU-CRD dataset to test performance across hospital settings.</div></div><div><h3>Results</h3><div>Our model predicted sepsis at 24, 12, and 6 h before onset, achieving AUROCs of 0.96, 0.98, and 0.99, respectively. The conformal approach reduced false positives and improved specificity. External validation confirmed similar performance, with a 57 % reduction in false alarms at the 6 h window, supporting practical use in low-monitoring environments.</div></div><div><h3>Conclusions</h3><div>This DL-based model enables accurate, early sepsis prediction with minimal data, addressing key clinical challenges and potentially improving resource allocation in hospital settings by reducing unnecessary ICU admissions and enhancing timely interventions.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110497"},"PeriodicalIF":7.0,"publicationDate":"2025-05-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144178307","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":"A comparative analysis of metamodels for 0D cardiovascular models, and pipeline for sensitivity analysis, parameter estimation, and uncertainty quantification","authors":"John M. Hanna ,&nbsp;Pavlos Varsos ,&nbsp;Jérôme Kowalski ,&nbsp;Lorenzo Sala ,&nbsp;Roel Meiburg ,&nbsp;Irene E. Vignon-Clementel","doi":"10.1016/j.compbiomed.2025.110381","DOIUrl":"10.1016/j.compbiomed.2025.110381","url":null,"abstract":"<div><div>Zero-dimensional (0D) cardiovascular models are reduced-order models aimed at studying the global dynamics of the whole circulation system or transport within it. They are employed to obtain estimates of important biomarkers for surgery planning and assessment applications (such as pressures, volumes, flow rates, and concentrations in the circulation) and can provide boundary conditions for high-fidelity three-dimensional models. Despite their low computational cost, tasks such as parameter estimation or uncertainty quantification require a large number of model evaluations, which is still a computationally expensive task. This motivates the building of metamodels in an offline stage, which can be evaluated significantly faster than 0D models. In this work, a pipeline going from 0D cardiovascular models to the building of metamodels and showcasing their use for tasks such as sensitivity analysis, parameter estimation, or uncertainty quantification is proposed. Three different strategies are assessed to build metamodels for 0D cardiovascular models, namely Neural Networks, Polynomial Chaos Expansion, and Gaussian Processes. The metamodels are assessed for three different 0D models. The first is a lumped model aimed at predicting the pressure in the portal vein after surgery. Due to the strong interaction between local liver hemodynamics and global circulation, the full circulation is modeled. The second one is simulating the whole-body circulation under the conditions of pulmonary arterial hypertension before and after shunt insertion. The final model is aimed at assessing the blood perfusion of an organ after a revascularization surgery. The transport of a contrast agent is modeled on top of a simplified 0D hemodynamics model. This model is chosen due to the different nature of the output which is a signal (concentration of the contrast agent over time), which requires a different treatment from the metamodeling point of view. The metamodels are trained and tested on synthetic data generated from the 0D models. It was found that neural networks offer the most convenient way of building metamodels in terms of the quality of the results, computational time, and practical ease of performing parameter estimation, sensitivity analysis, or uncertainty quantification tasks. Finally, we demonstrate a full pipeline of sensitivity analysis, inverse problem and (patient-specific) UQ, with a neural network as emulator.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110381"},"PeriodicalIF":7.0,"publicationDate":"2025-05-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144178308","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":"Adaptive motor unit decomposition using a cross-validation-based update policy","authors":"Tianze Ma ,&nbsp;Xiaogang Hu","doi":"10.1016/j.compbiomed.2025.110479","DOIUrl":"10.1016/j.compbiomed.2025.110479","url":null,"abstract":"<div><div>Extraction of motor unit (MU) information from electromyographic (EMG) signals has shown promise in neurophysiology and rehabilitation. However, the low accuracy of MU spike train firing information remains a major issue when the signals have stochastic variations. The objective of this study was to develop a new adaptive MU spike train decomposition algorithm with a deterministic pool of MU spike trains update policy. We first identified common MU spike trains, which were proven to be accurate, from two groups of concurrently recorded EMG signals. We then updated the common pool of MU spike trains with a flag policy, when we periodically updated the MU spike train separation matrix, which could add newly identified MU spike trains and remove inaccurate MU spike trains from the MU spike train pool. The flags of individual MU spike trains captured the consistency of MU active state and the likelihood of being extracted by the decomposition algorithm repetitively. We systematically evaluated the new algorithm on simulated datasets with 1-h pseudorandom activation levels under various conditions, including different degrees of amplitude drift of action potentials, different rates of MU rotation, and different levels of signal-to-noise ratios. The results demonstrated that our adaptive algorithm could identify and retain MU spike trains with 28 % higher accuracy compared with the conventional decomposition method. We also found consistently high decomposition accuracy across various signal conditions. These findings highlight the robustness of our decomposition approach. The outcomes have the potential to enhance neural decoding performance and could be applied to different scenarios, such as evaluating neurophysiological mechanisms during sustained muscle activations and assessing motor recovery during rehabilitation.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110479"},"PeriodicalIF":7.0,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144178306","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":"Multimodal AI framework for lung cancer diagnosis: Integrating CNN and ANN models for imaging and clinical data analysis","authors":"Emir Oncu ,&nbsp;Fatih Ciftci","doi":"10.1016/j.compbiomed.2025.110488","DOIUrl":"10.1016/j.compbiomed.2025.110488","url":null,"abstract":"<div><div>Lung cancer remains a leading cause of cancer-related mortality worldwide, emphasizing the critical need for accurate and early diagnostic solutions. This study introduces a novel multimodal artificial intelligence (AI) framework that integrates Convolutional Neural Networks (CNNs) and Artificial Neural Networks (ANNs) to improve lung cancer classification and severity assessment. The CNN model, trained on 1019 preprocessed CT images, classified lung tissue into four histological categories, adenocarcinoma, large cell carcinoma, squamous cell carcinoma, and normal, with a weighted accuracy of 92 %. Interpretability is enhanced using Gradient-weighted Class Activation Mapping (Grad-CAM), which highlights the salient image regions influencing the model's predictions. In parallel, an ANN trained on clinical data from 999 patients—spanning 24 key features such as demographic, symptomatic, and genetic factors—achieves 99 % accuracy in predicting cancer severity (low, medium, high). SHapley Additive exPlanations (SHAP) are employed to provide both global and local interpretability of the ANN model, enabling transparent decision-making. Both models were rigorously validated using k-fold cross-validation to ensure robustness and reduce overfitting. This hybrid approach effectively combines spatial imaging data and structured clinical information, demonstrating strong predictive performance and offering an interpretable and comprehensive AI-based solution for lung cancer diagnosis and management.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110488"},"PeriodicalIF":7.0,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144167182","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":"Multiclass ensemble framework for enhanced prostate gland Segmentation: Integrating Self-ONN decoders with EfficientNet","authors":"Md. Shaheenur Islam Sumon ,&nbsp;Muhammad E.H. Chowdhury ,&nbsp;Enamul Hoque Bhuiyan ,&nbsp;Md. Sohanur Rahman ,&nbsp;Muntakim Mahmud Khan ,&nbsp;Israa Al-Hashimi ,&nbsp;Adam Mushtak ,&nbsp;Sohaib Bassam Zoghoul","doi":"10.1016/j.compbiomed.2025.110459","DOIUrl":"10.1016/j.compbiomed.2025.110459","url":null,"abstract":"<div><div>Digital pathology relies on the morphological architecture of prostate glands to recognize cancerous tissue. Prostate cancer (PCa) originates in walnut shaped prostate gland in the male reproductive system. Deep learning (DL) pipelines can assist in identifying these regions with advanced segmentation techniques which are effective in diagnosing and treating prostate diseases. This facilitates early detection, targeted biopsy, and accurate treatment planning, ensuring consistent, reproducible results while minimizing human error. Automated segmentation techniques trained on MRI datasets can aid in monitoring disease progression which leads to clinical support by developing patient-specific models for personalized medicine. In this study, we present multiclass segmentation models designed to localize the prostate gland and its zonal regions—specifically the peripheral zone (PZ), transition zone (TZ), and the whole gland—by combining EfficientNetB4 encoders with Self-organized Operational Neural Network (Self-ONN)-based decoders. Traditional convolutional neural networks (CNNs) rely on linear neuron models, which limit their ability to capture the complex dynamics of biological neural systems. In contrast, Operational Neural Networks (ONNs), particularly Self-ONNs, address this limitation by incorporating nonlinear and adaptive operations at the neuron level. We evaluated various encoder-decoder configurations and identified that the combination of an EfficientNet-based encoder with a Self-ONN-based decoder yielded the best performance. To further enhance segmentation accuracy, we employed the STAPLE method to ensemble the top three performing models. Our approach was tested on the large-scale, recently updated PI-CAI Challenge dataset using 5-fold cross-validation, achieving Dice scores of 95.33 % for the whole gland and 92.32 % for the combined PZ and TZ regions. These advanced segmentation techniques significantly improve the quality of PCa diagnosis and treatment, contributing to better patient care and outcomes.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110459"},"PeriodicalIF":7.0,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144178305","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":"Real-time brain tumor detection in intraoperative ultrasound: From model training to deployment in the operating room","authors":"Santiago Cepeda ,&nbsp;Olga Esteban-Sinovas ,&nbsp;Roberto Romero ,&nbsp;Vikas Singh ,&nbsp;Prakash Shett ,&nbsp;Aliasgar Moiyadi ,&nbsp;Ilyess Zemmoura ,&nbsp;Giuseppe Roberto Giammalva ,&nbsp;Massimiliano Del Bene ,&nbsp;Arianna Barbotti ,&nbsp;Francesco DiMeco ,&nbsp;Timothy R. West ,&nbsp;Brian V. Nahed ,&nbsp;Ignacio Arrese ,&nbsp;Roberto Hornero ,&nbsp;Rosario Sarabia","doi":"10.1016/j.compbiomed.2025.110481","DOIUrl":"10.1016/j.compbiomed.2025.110481","url":null,"abstract":"<div><div>Intraoperative ultrasound (ioUS) is a valuable tool in brain tumor surgery due to its versatility, affordability, and seamless integration into the surgical workflow. However, its adoption remains limited, primarily because of the challenges associated with image interpretation and the steep learning curve required for effective use. This study aimed to enhance the interpretability of ioUS images by developing a real-time brain tumor detection system deployable in the operating room. We collected 2D ioUS images from the BraTioUS and ReMIND datasets, annotated with expert-refined tumor labels. Using the YOLO11 architecture and its variants, we trained object detection models to identify brain tumors. The dataset included 1732 images from 192 patients, divided into training, validation, and test sets. Data augmentation expanded the training set to 11,570 images. In the test dataset, YOLO11s achieved the best balance of precision and computational efficiency, with a mAP@50 of 0.95, mAP@50–95 of 0.65, and a processing speed of 34.16 frames per second. The proposed solution was prospectively validated in a cohort of 20 consecutively operated patients diagnosed with brain tumors. Neurosurgeons confirmed its seamless integration into the surgical workflow, with real-time predictions accurately delineating tumor regions. These findings highlight the potential of real-time object detection algorithms to enhance ioUS-guided brain tumor surgery, addressing key challenges in interpretation and providing a foundation for future development of computer vision-based tools for neuro-oncological surgery.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110481"},"PeriodicalIF":7.0,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144167181","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":"Higher-order EEG microstate syntax and surrogate testing","authors":"Frederic von Wegner ,&nbsp;Gesine Hermann ,&nbsp;Inken Tödt ,&nbsp;Inga Karin Todtenhaupt ,&nbsp;Helmut Laufs","doi":"10.1016/j.compbiomed.2025.110367","DOIUrl":"10.1016/j.compbiomed.2025.110367","url":null,"abstract":"<div><div>Higher-order syntax properties of EEG microstate sequences offer insight into the transition dynamics of functional brain networks. We here define higher-order syntax as microstate sequence properties that are not explained by the first-order transition matrix, and we postulate three requirements that surrogate data should fulfill to provide a null hypothesis for higher-order syntax tests. We then compare two general approaches to surrogate data generation that have been used in microstate research, (a) surrogates from a first-order Markov chain model, and, (b) surrogates obtained from sequence shuffling.</div><div>There are two different ways of representing microstate sequences, and syntax analyses can be applied to both, continuous microstate sequences, where each time sample is assigned the nearest microstate cluster, or to jump sequences which record only non-identical transitions by removing adjacent duplicates. We show that jump sequences have at least first-order syntax properties, whereas continuous sequences allow for zero-order and first-order surrogates. Markov chain generated surrogates fulfill the three requirements, i.e. they preserve the microstate distribution and transition matrix, and have no higher-order properties. Jump sequence shuffling, on the other hand, yields first-order surrogates whose first-order parameters are markedly different from the original sequence. Using a large open-access resting-state EEG dataset we show that jump sequence shuffling almost certainly produces microstate word probabilities that are significantly different from first-order expected word frequencies, erroneously indicating higher-order syntax properties. Markov chain surrogates reproduce the expected word probabilities of first-order sequences and correctly reject higher-order syntax properties in these cases.</div><div>We conclude that jump sequence shuffling does not produce adequate surrogates for higher-order syntax investigations. The proposed Markov chain generative method for surrogate data synthesis is computationally efficient and allows the generation of surrogate sequences of arbitrary length, whereas shuffling can lead to sequences that are shorter than the original sequence and have variable length. Sample code in Python and MATLAB is provided.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110367"},"PeriodicalIF":7.0,"publicationDate":"2025-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144167185","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":"Computer-aided unveiling molecular mechanisms of Xylocarpus granatum against colorectal cancer: therapeutic intervention targeting P13K-AKT signaling pathway","authors":"Md Shakil Ahamed,&nbsp;Sheikh Abdullah Al Ashik","doi":"10.1016/j.compbiomed.2025.110441","DOIUrl":"10.1016/j.compbiomed.2025.110441","url":null,"abstract":"<div><div>Colorectal cancer (CRC) is a leading cause of cancer-related morbidity and mortality globally, with increasing incidence rates, particularly in early-onset cases. Despite advances in treatment, many developing countries face affordability and safety issues, emphasizing the need for more cost-effective, natural, and safer alternatives. While numerous <em>in vitro</em> studies have reported the anticancer efficacy of mangrove plant <em>Xylocarpus granatum</em>, the molecular mechanisms underlying its effects on CRC remain unexplored. Therefore, our study aimed to uncover the potential therapeutic impact of compounds of <em>X. granatum</em> on CRC to gain insight into novel therapeutic interventions through comprehensive bioinformatics and computational analyses. We aimed to investigate the molecular interactions and mechanisms of <em>Xylocarpus granatum</em> in CRC treatment. Using network pharmacology, patient survival, and cancer hallmarks analysis, we identified 2 significant proteins (AKT1 and ESR1) associated with CRC. On the other hand, utilizing ADMET, quantum chemistry, molecular docking, machine learning, and molecular dynamics simulation, we explored deacetylgedunin (CID 3034112) as the most promising compound derived from <em>Xylocarpus granatum</em>. Our findings revealed that deacetylgedunin exhibited strong binding affinities to AKT1, with docking binding affinity −11.1 kcal/mol and MM-GBSA binding free energy −90.04 kcal/mol. Additionally, molecular dynamics analysis confirmed the stability of the AKT1-CID 3034112 complex, while machine learning (ML) estimation suggested potent biological activity (IC₅₀: 114.02 nM) against AKT1, reaffirming its therapeutic potential against CRC, particularly through modulation of the PI3K-AKT signaling pathway. Therefore, our study highlights the promising role of <em>Xylocarpus granatum</em> as a novel therapeutic intervention against CRC by modulating the P13K-AKT signaling pathway. However, our study was limited to computer-aided studies only, and therefore, further experimental validation is necessary to establish the therapeutic efficacy of deacetylgedunin from the <em>Xylocarpus granatum</em> plant in clinical settings.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110441"},"PeriodicalIF":7.0,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144166480","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":"Exploring best-performing radiomic features with combined multilevel discrete wavelet decompositions for multiclass COVID-19 classification using chest X-ray images","authors":"Hakan Özcan","doi":"10.1016/j.compbiomed.2025.110392","DOIUrl":"10.1016/j.compbiomed.2025.110392","url":null,"abstract":"<div><div>Discrete wavelet transforms have been applied in many machine learning models for the analysis of COVID-19; however, little is known about the impact of combined multilevel wavelet decompositions for the disease identification. This study proposes a computer-aided diagnosis system for addressing the combined multilevel effects of multiscale radiomic features on multiclass COVID-19 classification using chest X-ray images. A two-level discrete wavelet transform was applied to an optimal region of interest to obtain multiscale decompositions. Both approximation and detail coefficients were extensively investigated in varying frequency bands through 1240 experimental models. High dimensionality in the feature space was managed using a proposed filter- and wrapper-based feature selection approach. A comprehensive comparison was conducted between the bands and features to explore best-performing ensemble algorithm models. The results indicated that incorporating multilevel decompositions could lead to improved model performance. An inclusive region of interest, encompassing both lungs and the mediastinal regions, was identified to enhance feature representation. The light gradient-boosting machine, applied on combined bands with the features of basic, gray-level, Gabor, histogram of oriented gradients and local binary patterns, achieved the highest weighted precision, sensitivity, specificity, and accuracy of 97.50 %, 97.50 %, 98.75 %, and 97.50 %, respectively. The COVID-19-versus-the-rest receiver operating characteristic area under the curve was 0.9979. These results underscore the potential of combining decomposition levels with the original signals and employing an inclusive region of interest for effective COVID-19 detection, while the feature selection and training processes remain efficient within a practical computational time.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110392"},"PeriodicalIF":7.0,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144166484","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":"A medical information extraction model with contrastive tuning and tagging layer training","authors":"Xiaowei Wang","doi":"10.1016/j.compbiomed.2025.110465","DOIUrl":"10.1016/j.compbiomed.2025.110465","url":null,"abstract":"<div><div>Medical information extraction, as a core task in medical intelligent systems, focuses on extracting necessary structured information from clinical texts. In recent years, deep learning-based methods have become mainstream and often achieve superior extraction results. However, these existing methods have not fully tapped into the semantic potential of medical information categories, and most rely on a large amount of annotated data. This study proposes a novel semantic guided representation training model for medical information, which trains the representation of medical texts and medical information categories in the same semantic space by contrasting loss mechanisms, effectively reducing the need for annotated data. The experimental results show that our method objectives F1 value of 88.29 on CCKS2019 and 90.68 on CMeEE. Our method also exceeds the baseline by 4.07 on CCKS2019 and 4.95 on CMeEE.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110465"},"PeriodicalIF":7.0,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144166483","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}