Computers in biology and medicine最新文献

{"title":"MIO: An ontology for annotating and integrating medical knowledge in myocardial infarction to enhance clinical decision making","authors":"Chaoying Zhan ,&nbsp;Shumin Ren ,&nbsp;Yuxin Zhang ,&nbsp;Xiaojun Lv ,&nbsp;Yalan Chen ,&nbsp;Xin Zheng ,&nbsp;Rongrong Wu ,&nbsp;Erman Wu ,&nbsp;Tong Tang ,&nbsp;Jiao Wang ,&nbsp;Cheng Bi ,&nbsp;Mengqiao He ,&nbsp;Xingyun Liu ,&nbsp;Ke Zhang ,&nbsp;Yingbo Zhang ,&nbsp;Bairong Shen","doi":"10.1016/j.compbiomed.2025.110107","DOIUrl":"10.1016/j.compbiomed.2025.110107","url":null,"abstract":"<div><div>As biotechnology and computer science continue to advance, there's a growing amount of biomedical data worldwide. However, standardizing and consolidating these data remains challenging, making analysis and comprehension more difficult. To enhance research on complex diseases like myocardial infarction (MI), an ontology is necessary to ensure consistent data labeling and knowledge representation. This will facilitate data management and the application of artificial intelligence techniques in this field, ultimately advancing precision medicine research for MI. This study introduced the MI Ontology (MIO), which was developed using Stanford's seven-step method and Protégé. MIO aims to support precision medicine research on MI by effectively modeling and representing MI-related concepts and relationships. The validation of the MIO model involved employing Ontology Web Language (OWL) reasoners and comparing it with other disease-specific ontologies. MIO is an ontology model comprising of 3090 classes, 14 object attributes, 3494 individuals, 9415 synonyms and 49263 axioms, which encompass knowledge related to MI such as anatomical entities, clinical findings, drugs, genes, influencing factors, pathogenesis, patients-related concepts, procedures, and disease types. Furthermore, MIO has passed logical consistency validation and exhibits a broader conceptual scope and deeper knowledge structure than other disease-specific ontologies. Additionally, clinical use scenarios for MIO were developed to help address specific clinical problems. This study constructed the first comprehensive disease-specific ontology in cardiovascular diseases, named MIO, to promote precision medicine research on MI. MIO integrates and standardizes medical data, addressing complexity and standardization challenges. This promotes the use of big data analysis, explainable AI, and deep phenotype research in precision medicine. Future efforts will focus on enhancing and expanding MIO's applicability and scalability for superior services in this field.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":"Article 110107"},"PeriodicalIF":7.0,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143747439","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":"Accelerating continuum-based protein dynamics simulation using three-dimensional mixed overlapping element","authors":"Giseok Yun ,&nbsp;Do-Nyun Kim","doi":"10.1016/j.compbiomed.2025.110114","DOIUrl":"10.1016/j.compbiomed.2025.110114","url":null,"abstract":"<div><div>The conformational dynamics of proteins are crucial for their biological functions, requiring accurate modeling of both proteins and the surrounding solvent. This study focuses on the finite element framework of continuum-based protein dynamics simulation which integrates protein and solvent environments to simulate protein dynamics with explicit consideration of solvent effects. The conventional approach uses high-order mixed finite elements for incompressible analysis due to stability constraints, which results in a computationally burdensome process. To address this issue, we utilize three-dimensional mixed overlapping element to construct the solvent finite element models. This approach maintains stability while significantly reducing computational costs compared to conventional high-order methods, enabling the acceleration of continuum-based protein dynamics simulations.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":"Article 110114"},"PeriodicalIF":7.0,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143747441","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 data-driven personalized approach to predict blood glucose levels in type-1 diabetes patients exercising in free-living conditions","authors":"Anas Neumann ,&nbsp;Yessine Zghal ,&nbsp;Marzia Angela Cremona ,&nbsp;Adnene Hajji ,&nbsp;Michael Morin ,&nbsp;Monia Rekik","doi":"10.1016/j.compbiomed.2025.110015","DOIUrl":"10.1016/j.compbiomed.2025.110015","url":null,"abstract":"&lt;div&gt;&lt;h3&gt;Objective:&lt;/h3&gt;&lt;div&gt;The development of new technologies has generated vast amount of data that can be analyzed to better understand and predict the glycemic behavior of people living with type 1 diabetes. This paper aims to assess whether a data-driven approach can accurately and safely predict blood glucose levels in patients with type 1 diabetes exercising in free-living conditions.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Methods:&lt;/h3&gt;&lt;div&gt;Multiple machine learning (XGBoost, Random Forest) and deep learning (LSTM, CNN-LSTM, Dual-encoder with Attention layer) regression models were considered. Each deep-learning model was implemented twice: first, as a personalized model trained solely on the target patient’s data, and second, as a fine-tuned model of a population-based training model. The datasets used for training and testing the models were derived from the Type 1 Diabetes Exercise Initiative (T1DEXI). A total of 79 patients in T1DEXI met our inclusion criteria. Our models used various features related to continuous glucose monitoring, insulin pumps, carbohydrate intake, exercise (intensity and duration), and physical activity-related information (steps and heart rate). This data was available for four weeks for each of the 79 included patients. Three prediction horizons (10, 20, and 30 min) were tested and analyzed.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Results:&lt;/h3&gt;&lt;div&gt;For each patient, there always exists either a machine learning or a deep learning model that conveniently predicts BGLs for up to 30 min. The best performing model differs from one patient to another. When considering the best performing model for each patient, the median and the mean Root Mean Squared Error (RMSE) values (across the 79 patients) for predictions made 10 min ahead were 6.99 mg/dL and 7.46 mg/dL, respectively. For predictions made 30 min ahead, the median and mean RMSE values were 16.85 mg/dL and 17.74 mg/dL, respectively. The majority of the predictions output by the best model of each patient fell within the clinically safe zones A and B of the Clarke Error Grid (CEG), with almost no predictions falling into the unsafe zone E. The most challenging patient to predict 30 min ahead achieved an RMSE value of 32.31 mg/dL (with the corresponding best performing model). The best-predicted patient had an RMSE value of 10.48 mg/dL. Predicting blood glucose levels was more difficult during and after exercise, resulting in higher RMSE values on average. Prediction errors during and after physical activity (two hours and four hours after) generally remained within the clinical safe zones of the CEG with less than 0.5% of predictions falling into the harmful zones D and E, regardless of the exercise category.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Conclusions:&lt;/h3&gt;&lt;div&gt;Data-driven approaches can accurately predict blood glucose levels in type 1 diabetes patients exercising in free-living conditions. The best-performing model varies across patients. Approaches in which a population-based model is initially trained and","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":""},"PeriodicalIF":7.0,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143737834","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Immune profile and routine laboratory indicator-based machine learning for prediction of lung cancer","authors":"Yi Huang ,&nbsp;Kaishan Jiang ,&nbsp;Xiaochen Wang ,&nbsp;Siyu Zou ,&nbsp;Ziyong Sun ,&nbsp;Shiji Wu ,&nbsp;Bin Wang ,&nbsp;Hongyan Hou ,&nbsp;Feng Wang","doi":"10.1016/j.compbiomed.2025.110111","DOIUrl":"10.1016/j.compbiomed.2025.110111","url":null,"abstract":"<div><h3>Introduction</h3><div>Early diagnosis of lung cancer is still a challenge by using current diagnostic methods.</div></div><div><h3>Objectives</h3><div>The study aims to explore the utilization of host immune parameters, in combination with conventional laboratory tests, for the early prediction of lung cancer.</div></div><div><h3>Methods</h3><div>Immune profiles were assessed by flow cytometry in 221 patients, and machine learning algorithms, utilizing either combined or routine indicators alone, were applied to classify lung cancer stages.</div></div><div><h3>Results</h3><div>The study revealed significant alterations in immune profiles across different stages of lung cancer. Notably, we observed a progressive increase in the percentages of effector memory CD8⁺ T cells and polymorphonuclear-MDSCs from healthy controls to patients with benign lesion, early-stage cancer, and late-stage cancer. Conversely, the percentages of naive CD8⁺ T cells, DCs, and NKG2D⁺ NK cells exhibited a decreasing trend throughout this progression. Accordingly, the gradual differentiation of effector CD8⁺ T cells and the accumulation of inhibitory polymorphonuclear-MDSCs, along with the progressive impairment of innate and adaptive immunity, were the most prominent immune features observed during lung cancer progression. Through in combination of selected conventional laboratory and immune indicators, we demonstrated the effectiveness of machine learning models, particularly SVC and logistic regression, in predicting the presence of lung cancer and its staging with high accuracy.</div></div><div><h3>Conclusion</h3><div>We depict the immune landscape in patients with benign disease and different stages of lung cancer. Combination of routine and immune indicators by using machine learning displays a potential in predicting the presence of lung cancer and its staging.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":"Article 110111"},"PeriodicalIF":7.0,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143747437","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 novel hybrid feature fusion approach using handcrafted features with transfer learning model for enhanced skin cancer classification","authors":"B. Soundarya,&nbsp;C. Poongodi","doi":"10.1016/j.compbiomed.2025.110104","DOIUrl":"10.1016/j.compbiomed.2025.110104","url":null,"abstract":"<div><div>Skin cancer is a deadly disease and has the highest rising rates globally. It arises from aberrant skin cells, which are often caused by prolonged exposure to ultraviolet rays from sunlight or artificial tanning devices. Dermatologists rely on visual inspection and need to identify suspicious lesions. Prompt and accurate diagnosis is pivotal for effective treatment and enhancing the chances of recovery. Recently, skin cancer prediction has been made utilising machine and deep learning algorithms for early detection. This methodology presents a novel hybrid feature extraction and is fused with a deep learning model for dermoscopic image analysis. Skin lesion images from sources like ISIC were pre-processed. Features were extracted using the Grey-Level Co-Occurrence Matrix (GLCM), Redundant Discrete Wavelet Transform (RDWT) and a various pre-trained model. After evaluating all the combinations, the proposed feature fusion model performed well rather than all other models. This proposed feature fusion model includes GLCM, RDWT, and DenseNet121 features, which were estimated with the various classifiers, among which an impressive accuracy of 93.46 % was obtained with the XGBoost classifier and 94.25 % with the ensemble classifier. This study underscores the efficacy of integrating diverse feature extraction techniques to increase the reliability and effectiveness of skin cancer diagnosis.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":"Article 110104"},"PeriodicalIF":7.0,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143747508","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 newly developed 2 mm needle arthroscope with high-definition for orthopedic outpatient knee joint examination","authors":"Mingda Liu ,&nbsp;Ming Yan ,&nbsp;Jianwen Xu ,&nbsp;Qinghui Zeng ,&nbsp;Kai Zhu ,&nbsp;Shuai Zhao ,&nbsp;Wenbo Diao ,&nbsp;Yueying Wang ,&nbsp;Xiangyang Leng","doi":"10.1016/j.compbiomed.2025.110112","DOIUrl":"10.1016/j.compbiomed.2025.110112","url":null,"abstract":"<div><div>Currently, the diagnosis of orthopedic joint diseases relies mainly on magnetic resonance imaging (MRI), which is valuable for detecting bone injuries, as well as lesions in intra-articular cartilage, ligaments, bone marrow, synovium, and other areas. However, MRI has significant drawbacks. Firstly, MRI cannot provide detailed pathological assessments of joint interiors comparable to arthroscopic examination results. Secondly, MRI is costly, imposing substantial financial burdens on patients with each additional site examined clinically. Lastly, MRI resources are scarce in medical institutions due to widespread applications beyond joint examinations, requiring appointments days in advance. Each examination is limited to one site, taking approximately 20–30 min per session; multiple sites necessitate sequential examinations, entailing considerable time costs and potentially delaying diagnoses. The development of needle arthroscopy (NA) for outpatient knee joint examinations offers more accurate and detailed visualization of knee joint pathologies compared to traditional MRI imaging, with lower economic and time costs. Moreover, NA provides a valuable diagnostic alternative for patients unsuitable for MRI examinations. In-office needle arthroscopy (IONA) also enables orthopedic surgeons to deliver more comprehensive medical services beyond diagnostic examinations. Based on this theory, we have designed and manufactured a 2 mm-diameter lens arthroscope useable under local anesthesia in outpatient settings to examine tissue injuries within knee joints. The needle arthroscope objective 2 mm has an 85°field of view, a 30°viewing angle, and an imaging resolution of 10 lp/mm. This device provides clear imaging of cartilage surfaces, ligaments, synovium, and other soft tissues inside cadaveric knee joints, enabling more detailed and accurate evaluations of knee joint tissue pathologies. This method facilitates convenient, efficient joint examinations with lower economic and time costs, serving as a viable alternative to knee joint MRI examinations.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":"Article 110112"},"PeriodicalIF":7.0,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143747438","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":"Efficient multi-task learning with instance selection for biomedical NLP","authors":"Agnese Bonfigli ,&nbsp;Luca Bacco ,&nbsp;Leandro Pecchia ,&nbsp;Mario Merone ,&nbsp;Felice Dell’Orletta","doi":"10.1016/j.compbiomed.2025.110050","DOIUrl":"10.1016/j.compbiomed.2025.110050","url":null,"abstract":"<div><h3>Background:</h3><div>Biomedical natural language processing (NLP) increasingly relies on large language models and extensive datasets, presenting significant computational challenges.</div></div><div><h3>Methods:</h3><div>We propose Blue5, a multi-task model based on SciFive that incorporates instance selection (IS) to enable efficient, multi-task learning (MTL) on biomedical data. We adapt the E2SC-IS framework for the biomedical domain, integrating a calibrated SVM classifier to reduce computational costs.</div></div><div><h3>Results:</h3><div>Our approach achieves an average data reduction of 26.6% across the several tasks of the BLUE (Biomedical Language Understanding Evaluation) Benchmark, while maintaining performance comparable with state-of-the-art models. The multi-task SVM configuration emerges as the most effective, demonstrating the power of combining IS with MTL for biomedical NLP. As a result of the unified framework, Blue5 effectively selects the most informative instances across tasks, ensuring model generalization while efficiently handling multiple NLP tasks.</div></div><div><h3>Conclusion:</h3><div>Our work offers a practical solution to address growing computational demands, enabling more scalable and accessible applications of advanced NLP techniques in biomedical research and healthcare.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":"Article 110050"},"PeriodicalIF":7.0,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143747461","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A lightweight approach to gait abnormality detection for At Home health monitoring","authors":"Chris Lochhead,&nbsp;Robert B. Fisher","doi":"10.1016/j.compbiomed.2025.110076","DOIUrl":"10.1016/j.compbiomed.2025.110076","url":null,"abstract":"<div><div>Gait abnormality detection is a growing application in machine learning based health assessment due to its potential in domains from clinical health reviews to at home health monitoring. This latter application is of particular use for older adults, who are more likely to experience health issues that can be indicated by changes in gait, namely through fall-related injuries or age-related degenerative diseases like Parkinson's disease. While there exists a great deal of research concerning machine learning models for detecting everything from freezing-of-gait to falls, much of this work relies on clinical assessment settings and large models with extensive data, making many developments unusable in at-home applications where such technology could be used to great benefit in maintaining the independence and health of older adults. To address this gap in the literature, we introduce a new 15-person synthetic gait abnormality dataset named WeightGait and a lightweight ST-GCN model to demonstrate the feasibility of smaller models with lower computational costs in detecting gait abnormalities in an environment more analogous to the conditions found in an at-home setting. For the task of identifying gait abnormalities in the WeightGait dataset, this method achieves 94.4 % accuracy, an improvement of between 4.9 % and 15.41 % on comparable gait assessment methods.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":"Article 110076"},"PeriodicalIF":7.0,"publicationDate":"2025-03-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143747460","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"LSDVvac: An immunoinformatics database for vaccine design against lumpy skin disease virus","authors":"Sumit Sharma ,&nbsp;Ritika Bishnoi ,&nbsp;Riya Jain,&nbsp;Deepak Singla","doi":"10.1016/j.compbiomed.2025.110077","DOIUrl":"10.1016/j.compbiomed.2025.110077","url":null,"abstract":"<div><div>Development of an effective vaccine against Lumpy Skin Disease Virus (LSDV) is crucial for protecting livestock. The current study outlines a web-based platform developed to aid the scientific community in designing effective peptide-based vaccines against LSDV. First, we generated all possible overlapping (K-mer value 9 and 15) peptides from the proteins of 73 LSDV strains. Second, after removing redundancy, the obtained peptides were utilized for predicting B-cell and T-cell epitopes. Third, the predicted B-cell and T-cell epitopes were screened for immunogenicity, allergenicity, and toxicity. Finally, the LSDV candidate vaccine database was developed utilizing 3913 unique B-cell (Linear 3344 and conformational 569) and 6473 unique T-cell (MHC-I 3200 and MHC-II 3273) epitopes. The three-dimensional structure of 156 LSDV proteins from reference (AF325528.1) LSDV genome was predicted using I-TASSER software and implemented in the database. Additionally, tools for genome analysis like DotPlot, Gblocks, BLAST, and gRNA designing were incorporated into the database. In summary, LSDVvac has been developed, which integrates information about predicted potential vaccine candidates along with useful computational tools. LSDVvac database is available at <span><span>http://45.248.163.59/bic/lsdb/</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":"Article 110077"},"PeriodicalIF":7.0,"publicationDate":"2025-03-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143734607","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":"Natural language processing for identifying major bleeding risk in hospitalised medical patients","authors":"Anne Bryde Alnor ,&nbsp;Rasmus Bank Lynggaard ,&nbsp;Martin Sundahl Laursen ,&nbsp;Pernille Just Vinholt","doi":"10.1016/j.compbiomed.2025.110093","DOIUrl":"10.1016/j.compbiomed.2025.110093","url":null,"abstract":"<div><h3>Background</h3><div>Major bleeding is a severe complication in critically ill medical patients, resulting in significant morbidity, mortality, and healthcare costs. This study aims to assess the incidence and risk factors for major bleeding in hospitalised medical patients using a Natural Language Processing (NLP) model.</div></div><div><h3>Methods</h3><div>We conducted a retrospective, cross-sectional observational study using electronic health records of adult patients admitted through the Emergency Department at Odense University Hospital from January 2017 to December 2022. Major bleeding during admission was identified and validated using a natural language model, with events classified according to current guidelines. Risk factors, including demographics, comorbidities, and biochemical values at admission, were evaluated. Two risk assessment models (RAMs) were developed using Cox proportional hazards regression. Validation included, bootstrapping, K-fold cross validation, and cluster analyses.</div></div><div><h3>Results</h3><div>Of the 46,439 eligible patients, 1246 (2.7 %) experienced major bleeding. Risk factors for major bleeding included older age, male sex, alcohol consumption, higher systolic blood pressure, lower haemoglobin, and higher creatinine. RAM 1, which included biochemical data and comorbidities, demonstrated robust predictive performance (Harrell's C-statistic = 0.726). RAM 2, a simplified model without comorbidities, maintained similar predictive accuracy (C-statistic = 0.721), indicating its potential utility in clinical settings with limited resources for detailed patient histories. Results were consistent throughout validation.</div></div><div><h3>Conclusion</h3><div>This study highlights the incidence and risk factors of major bleeding in medical patients, emphasizing the predictive value of routinely measured biochemical markers. Furthermore, it shows the applicability of NLP models in identifying bleeding episodes in EHR text.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"190 ","pages":"Article 110093"},"PeriodicalIF":7.0,"publicationDate":"2025-03-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143734608","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}