Computers in biology and medicine最新文献

{"title":"Effects of different valve-in-valve positions on the hydrodynamic properties of transcatheter aortic valves","authors":"Hong Fang ,&nbsp;Shi Su ,&nbsp;Liang Zhang ,&nbsp;Shuchun Li ,&nbsp;Kun Zhang ,&nbsp;Kejing Yi ,&nbsp;Mengxiao Shi ,&nbsp;Nan Wang ,&nbsp;Qing Zhou ,&nbsp;Min Jin","doi":"10.1016/j.compbiomed.2025.110106","DOIUrl":"10.1016/j.compbiomed.2025.110106","url":null,"abstract":"<div><h3>Objective</h3><div>With the rise of transcatheter aortic valve-in-valve (ViV) procedures for high-risk patients with degenerated surgical aortic valves, precise positioning of the transcatheter heart valve (THV) within the surgical heart valve (SHV) is crucial for optimal functional outcomes. This study aims to explore the impact of implantation depth on functional outcomes post-ViV in a controlled in vitro setting.</div></div><div><h3>Methods</h3><div>This study focused on the impact of valve positioning on fluid dynamics characteristics and subsequent ViV procedural outcomes. Rigorous in vitro experiments measured the structural parameters of the surgical valves, and based on these, the appropriate Taurus Elite valves were selected for pulse flow testing under simulated conditions of varying heart rates and cardiac outputs. Fluid dynamic evaluations were conducted on Taurus Elite THVs in sizes 21, 23, 26, 29 mm and SHVs from two brands: Hancock II (Medtronic, USA) and BalMedic (Balance Medical, China) across a range of diameters. In-depth analysis was performed at a cardiac output (CO) of 5 L/min and heart rate (HR) of 70 bpm, focusing on key metrics such as transvalvular pressure gradient (TVPG), effective orifice area (EOA), and total regurgitation fraction (TRF) at implantation depths of −2.5, 0, 2.5, and 5 mm to gain insights into the dynamic interaction between THV placement and hemodynamic performance. Anchoring force tests were also conducted for SHV-THV combinations at −2.5 and 0 mm depths to ensure safety of implantation.</div></div><div><h3>Results</h3><div>Significant differences were observed in TVPG, EOA, and TRF across various SHV brands and sizes, emphasizing the importance of THV positioning. Specifically, Taurus Elite23 demonstrated superior TVPG performance at various depths compared to Taurus Elite21, indicating a better match with BalMedic19 and Hancock II21, especially at implantation depths ranging from −2.5 to 0 mm. Taurus Elite29 showed the lowest TVPG across all tested depths, making it the preferred choice for BalMedic25 and Hancock II27. These findings highlight the importance of selecting the appropriate THV model and determining the optimal implantation depth for different SHVs.</div></div><div><h3>Conclusions</h3><div>In different surgical valves, both the model and implantation depth of the interventional valve can affect its hemodynamic performance and valve opening-closing morphology. The recommended implantation of the interventional valve as shallowly as possible in this study has guiding significance in clinical valve-in-valve surgeries.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110106"},"PeriodicalIF":7.0,"publicationDate":"2025-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143799948","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":"Mitosis detection and classification for breast cancer diagnosis: What we know and what is next","authors":"Rafi Ullah Khalil ,&nbsp;Muhammad Sajjad ,&nbsp;Sami Dhahbi ,&nbsp;Sami Bourouis ,&nbsp;Mohammad Hijji ,&nbsp;Khan Muhammad","doi":"10.1016/j.compbiomed.2025.110057","DOIUrl":"10.1016/j.compbiomed.2025.110057","url":null,"abstract":"<div><div>Breast cancer is the second most deadly malignancy in women, behind lung cancer. Despite significant improvements in medical research, breast cancer is still accurately diagnosed with histological analysis. During this procedure, pathologists examine a physical sample for the presence of mitotic cells, or dividing cells. However, the high resolution of histopathology images and the difficulty of manually detecting tiny mitotic nuclei make it particularly challenging to differentiate mitotic cells from other types of cells. Numerous studies have addressed the detection and classification of mitosis, owing to increasing capacity and developments in automated approaches. The combination of machine learning and deep learning techniques has greatly revolutionized the process of identifying mitotic cells by offering automated, precise, and efficient solutions. In the last ten years, several pioneering methods have been presented, advancing towards practical applications in clinical settings. Unlike other forms of cancer, breast cancer and gliomas are categorized according to the number of mitotic divisions. Numerous papers have been published on techniques for identifying mitosis due to easy access to datasets and open competitions. Convolutional neural networks and other deep learning architectures can precisely identify mitotic cells, significantly decreasing the amount of labor that pathologists must perform. This article examines the techniques used over the past decade to identify and classify mitotic cells in histologically stained breast cancer hematoxylin and eosin images. Furthermore, we examine the benefits of current research techniques and predict forthcoming developments in the investigation of breast cancer mitosis, specifically highlighting machine learning and deep learning.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110057"},"PeriodicalIF":7.0,"publicationDate":"2025-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143799863","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 deep neural network approach to detect and monitor cocaine drug abuse","authors":"Aleena Swetapadma ,&nbsp;Divya Kumari","doi":"10.1016/j.compbiomed.2025.110130","DOIUrl":"10.1016/j.compbiomed.2025.110130","url":null,"abstract":"<div><h3>Purpose</h3><div>Cocaine is one of the most commonly used drugs that may lead to physical and mental health problems. It is necessary to identify individuals having cocaine use disorder as early as possible to monitor them properly. The objective of this work is to predict the time of cocaine use in scenarios where clinical testing is not possible. The time of cocaine use is defined as how many days before the individual has used cocaine.</div></div><div><h3>Methodology</h3><div>It is possible to predict the time of cocaine use based on personality traits and demographic information as features. The personality traits (neuroticism, extraversion, openness to experience, agreeableness, and conscientiousness, impulsivity, and sensation seeking) along with demographic information features (education level, age, gender, country of residence, and ethnicity) have been used to predict the time of cocaine use. These features are given as inputs to long short-term memory networks (LSTM) to predict the time of cocaine use.</div></div><div><h3>Findings</h3><div>The highest F-score for the prediction of time of cocaine use for the LSTM method is found to be 0.99. A comparative study has also been carried out using both deep neural networks and artificial neural networks to predict the time of cocaine use to demonstrate the superiority of the LSTM method. The proposed method shows promising results for predicting the time of cocaine use and can be considered for monitoring the cocaine use disorder.</div></div><div><h3>Practical and social implications</h3><div>The proposed method will be an efficient tool to identify the mental health of a person if the person has cocaine use disorder. As a result, proper treatment can be given to the individual in time.</div></div><div><h3>Originality</h3><div>The originality of the work is that it predicts the time of cocaine use with better accuracy. The LSTM method has not been used previously for predicting the time of cocaine use.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110130"},"PeriodicalIF":7.0,"publicationDate":"2025-04-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143799949","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 stacked ensemble approach for symptom-based monkeypox diagnosis","authors":"Shimaa Nagro","doi":"10.1016/j.compbiomed.2025.110140","DOIUrl":"10.1016/j.compbiomed.2025.110140","url":null,"abstract":"<div><div>The recent monkeypox outbreak has raised global health concerns. Caused by a virus, it is characterized by symptoms such as skin lesions. Early detection is critical for treatment and controlling its spread. This study uses advanced machine learning and deep learning techniques, including Tab Transformer, Long Short-Term Memory, XGBoost, LightGBM, and a Stacking Classifier, to predict the presence of the virus based on patient symptoms. The performance of these models is evaluated using accuracy, precision, recall, and F1-score metrics. The experiments reveal that the Stacking Classifier significantly outperforms the other models, achieving an accuracy of 87.29 %, precision of 86.12 %, recall of 87.47 %, and an F1 score of 87.89 %. Additionally, applying Conditional Tabular GAN to generate synthetic data helps address data imbalance issues, further improving model robustness. These results highlight the proposed approach's potential for timely, accurate monkeypox detection, aiding in effective disease management and control.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110140"},"PeriodicalIF":7.0,"publicationDate":"2025-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143792430","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":"Explainable transformer-based deep survival analysis in childhood acute lymphoblastic leukemia","authors":"Yuning Cui ,&nbsp;Weixuan Dong ,&nbsp;Yifu Li ,&nbsp;Amanda E. Janitz ,&nbsp;Hanumantha R. Pokala ,&nbsp;Rui Zhu","doi":"10.1016/j.compbiomed.2025.110118","DOIUrl":"10.1016/j.compbiomed.2025.110118","url":null,"abstract":"<div><h3>Background</h3><div>Acute lymphoblastic leukemia (ALL) is the most common type of leukemia among children and adolescents and can be life-threatening. The incidence of new cases has been increasing in recent years. Developing a predictive model to forecast the risk of death can help improve survival rates by enabling clinicians to provide timely and effective treatments. Traditional statistical survival models are limited by predefined assumptions, while current deep survival models, despite their flexibility, struggle with capturing complex and dynamic feature dependencies. Transformers provide a promising solution by using self-attention and multi-head attention mechanisms to overcome these challenges. Moreover, building on recent work in interpretable medical AI, the combination of Transformers and explainable methods can quantify the contributions of each feature to the survival probability prediction.</div></div><div><h3>Methods</h3><div>This paper proposes an explainable Transformer-based deep survival model to predict patient-specific survival probabilities for ALL. The model combines feedforward networks with Transformer architecture and is trained to minimize a loss function that measures the difference between predicted and actual survival outcomes. In addition, we use Shapley Additive Explanations (SHAP) to interpret the contributions of clinical attributes to the predictions, providing insights from both global and local perspectives.</div></div><div><h3>Results</h3><div>The proposed model demonstrates robustness by consistently providing higher average survival probabilities for censored patients compared to deceased patients. It achieves an average concordance index (C-index) of 0.945, demonstrating strong predictive accuracy. Through SHAP analysis, we identify three key factors affecting survival outcomes, namely prognosis status, diagnosis year, and histology. Experiment results reveal that our model outperforms state-of-the-art deep survival models in terms of C-index when these important variables are included.</div></div><div><h3>Conclusion</h3><div>The proposed explainable Transformer-based deep survival model shows strong potential for providing accurate patient-specific survival predictions for ALL. Moreover, the insights gained from SHAP improve the model's interpretability for clinicians, helping them make better-informed decisions regarding prognosis and treatment.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110118"},"PeriodicalIF":7.0,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143785467","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":"Repeated measures analysis for steady-state evoked potentials","authors":"Amir Norouzpour,&nbsp;Tawna L. Roberts","doi":"10.1016/j.compbiomed.2025.110117","DOIUrl":"10.1016/j.compbiomed.2025.110117","url":null,"abstract":"<div><h3>Introduction</h3><div>Brain response to repetitive stimuli generates steady-state evoked potentials (ssEP) that vary depending on the experimental conditions. To analyze these responses, Fourier measurements extracted from ssEP data require statistical techniques to differentiate neural responses across various experimental conditions within the same participant(s). In this study, we introduce new statistical methods to compare multiple dependent clusters of discrete Fourier measurements corresponding to multiple experimental conditions.</div></div><div><h3>Methods</h3><div>We present two statistics: 1) The first statistic is derived from repeated measures analysis of variance (ANOVA) for complex numbers, used to compare multiple dependent circular clusters of Fourier estimates under the assumption of equal variance across the clusters. 2) The second statistic is employed when either the assumption of circularity within the clusters or the assumption of equal variance across the clusters is violated. In this case, we derive the statistic from the rank-sum Friedman test to compare multiple related clusters of complex numbers.</div></div><div><h3>Results</h3><div>We demonstrated the validity of the statistics using simulated and empirical ssEP data. Our methods offer robust statistical tools that maintain a constant Type-I error of 0.05 in all conditions, including equal or unequal variance-covariance matrix of the real and imaginary components of Fourier estimates across the circular and elliptical clusters, even in the presence of outliers in the dataset. Furthermore, our statistics demonstrate a lower Type-II error compared to repeated measures multivariate analysis of variance (rmMANOVA).</div></div><div><h3>Conclusion</h3><div>The statistical methods enable us to compare multiple dependent clusters of Fourier estimates corresponding to multiple experimental conditions within the same participant(s), whether or not the variance is equal across the circular or elliptical clusters, even with outliers in the dataset.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110117"},"PeriodicalIF":7.0,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143785370","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":"The role of oncostatin M in esophageal cancer and its potential as a prognostic maker","authors":"Guangxiang Ding ,&nbsp;Xianwei Li ,&nbsp;Hongtao Yin ,&nbsp;Jianwei Tang ,&nbsp;Bing Sun ,&nbsp;Jun Xiu Sheng","doi":"10.1016/j.compbiomed.2025.110073","DOIUrl":"10.1016/j.compbiomed.2025.110073","url":null,"abstract":"<div><h3>Introduction</h3><div>Esophageal cancer (ESCA) remains one of the most lethal malignancies worldwide, with poor survival rates. Identifying novel biomarkers like Oncostatin M (OSM) could improve early detection, prognosis, and therapeutic strategies. OSM, a cytokine from the IL-6 family, plays complex roles in cancer biology, but its specific function in ESCA is poorly understood.</div></div><div><h3>Materials and methods</h3><div>Gene expression data from 198 ESCA samples were retrieved from The Cancer Genome Atlas (TCGA) and GEO databases. Bioinformatic analyses, including differential expression and survival analysis, were conducted using R software. Functional experiments involved OSM knockdown in ESCA cell lines, followed by cell proliferation, migration, colony formation assays, and wound healing analysis.</div></div><div><h3>Results</h3><div>OSM expression was significantly elevated in ESCA tissues compared to normal tissues. High OSM expression correlated with poor overall survival and was associated with clinical parameters such as tumor stage and lymph node metastasis. In vitro experiments demonstrated that OSM knockdown reduced cell proliferation, colony formation, and migration in ESCA cell lines.</div></div><div><h3>Conclusions</h3><div>This study found that OSM overexpression in esophageal cancer was linked to advanced stage, lymph node metastasis, and poor prognosis. A prognostic nomogram based on OSM expression and clinical features predicted patient survival. Functional analysis suggested that OSM influences cancer progression via cytokine-receptor signaling, and silencing OSM reduced cell proliferation and migration, potentially improving prognosis.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110073"},"PeriodicalIF":7.0,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143785405","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":"Unsteady analysis of the airflow in human nasal airway – a computational study","authors":"Syeda Reham Shahed ,&nbsp;Jingliang Dong ,&nbsp;Jiyuan Tu ,&nbsp;Lin Tian","doi":"10.1016/j.compbiomed.2025.110136","DOIUrl":"10.1016/j.compbiomed.2025.110136","url":null,"abstract":"<div><h3>Background and objective</h3><div>Nasal cavity is the first line of defence against toxicants and pollutants. A deep understanding of the realistic airflow in the complex geometry is crucial to evaluate pollution impact and initiation of various respiratory diseases. Human respiration by nature is unsteady, having two phases – inhalation and exhalation. While prior studies are predominantly on the steady state considering only the inhalation phase, we are curious to understand and investigate the unsteady nature of human respiration.</div></div><div><h3>Methods</h3><div>A sinusoidal unsteady profile simulating a complete breathing cycle at a flow rate of 10 L/min is considered in this study. We have reconstructed a realistic human nasal cavity model from CT scans and with Ansys Fluent we have analysed the flow field. The time-evolving flow patterns and wall shear stress, in particular during the distinctive accelerating and decelerating breathing phases, are extracted. Critical transient insight to the unsteady breathing which is largely unknown in a steady simulation is revealed. Finally, the breathing air flux in the olfactory region are calculated under the transient breathing condition.</div></div><div><h3>Results</h3><div>Evolving flow pattern at different time instants gives a significant variation of the flow dynamics which would not be understood in a steady study. The main differences in inhalation and exhalation are the regions where the main flow is dominant, the location of the vortices around the olfactory and wall shear stress patterns. In addition, past history of peak flow has noticeable effects on flow fields during the deceleration phase, especially when the breathing rate is low. Finally, flow pathways into the olfactory region varies during inhalation and exhalation phases.</div></div><div><h3>Conclusion</h3><div>The cyclic information provides critical transient insight to the unsteady breathing which is largely unknown in a steady study. In the context of a fast-growing interest in nasal transport to accurately evaluate pollution impact and pathogen deposition, investigation on the flow and particle depositions under unsteady condition is valuable and highly desired.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110136"},"PeriodicalIF":7.0,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143785466","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":"Generative adversarial networks in medical image reconstruction: A systematic literature review","authors":"Jabbar Hussain ,&nbsp;Magnus Båth ,&nbsp;Jonas Ivarsson","doi":"10.1016/j.compbiomed.2025.110094","DOIUrl":"10.1016/j.compbiomed.2025.110094","url":null,"abstract":"<div><h3>Purpose</h3><div>Recent advancements in generative adversarial networks (GANs) have demonstrated substantial potential in medical image processing. Despite this progress, reconstructing images from incomplete data remains a challenge, impacting image quality. This systematic literature review explores the use of GANs in enhancing and reconstructing medical imaging data.</div></div><div><h3>Method</h3><div>A document survey of computing literature was conducted using the ACM Digital Library to identify relevant articles from journals and conference proceedings using keyword combinations, such as “generative adversarial networks or generative adversarial network,” “medical image or medical imaging,” and “image reconstruction.”</div></div><div><h3>Results</h3><div>Across the reviewed articles, there were 122 datasets used in 175 instances, 89 top metrics employed 335 times, 10 different tasks with a total count of 173, 31 distinct organs featured in 119 instances, and 18 modalities utilized in 121 instances, collectively depicting significant utilization of GANs in medical imaging. The adaptability and efficacy of GANs were showcased across diverse medical tasks, organs, and modalities, utilizing top public as well as private/synthetic datasets for disease diagnosis, including the identification of conditions like cancer in different anatomical regions. The study emphasized GAN's increasing integration and adaptability in diverse radiology modalities, showcasing their transformative impact on diagnostic techniques, including cross-modality tasks. The intricate interplay between network size, batch size, and loss function refinement significantly impacts GAN's performance, although challenges in training persist.</div></div><div><h3>Conclusions</h3><div>The study underscores GANs as dynamic tools shaping medical imaging, contributing significantly to image quality, training methodologies, and overall medical advancements, positioning them as substantial components driving medical advancements.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110094"},"PeriodicalIF":7.0,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143785408","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":"Interpretable deep learning method to predict wound healing progress based on collagen fibers in wound tissue","authors":"Juan He ,&nbsp;Xiaoyan Wang ,&nbsp;Zhengshan Wang ,&nbsp;Ruitao Xie ,&nbsp;Zhiming Zhang ,&nbsp;Tzu-Ming Liu ,&nbsp;Yunpeng Cai ,&nbsp;Long Chen","doi":"10.1016/j.compbiomed.2025.110110","DOIUrl":"10.1016/j.compbiomed.2025.110110","url":null,"abstract":"<div><h3>Background and objective</h3><div>The dynamic evolution of collagen fibers during wound healing is crucial for assessing repair progression, guiding clinical treatment, and drug screening. Current quantitative methods analyzing collagen spatial patterns (density, orientation variance) lack established criteria to both stratify distinct healing periods and detect delayed healing conditions, necessitating the establishment of a novel classification method for wound healing status based on collagen fibers.</div></div><div><h3>Methods</h3><div>We propose a deep learning method to classify various time points of wound healing and delayed healing using histological images of skin tissue. We fine-tune a pre-trained VGG16 model and enhance it with an interpretable framework that combines LayerCAM and Guided Backpropagation, leveraging model gradients and features to visually identify the tissue regions driving model predictions.</div></div><div><h3>Results</h3><div>Our model achieved 85 % accuracy in a five-class classification task (normal skin, wound skin at 0, 3, 7, and 10 days) and 78 % in a three-class task (normal skin, wound skin at 0 days, diabetic wound skin at 10 days). Our interpretable framework accurately localizes collagen fibers without pixel-level annotations, demonstrating that our model classifies healing periods and delayed healing based on collagen regions in histological images rather than other less relevant tissue structures.</div></div><div><h3>Conclusions</h3><div>Our deep learning method leverages collagen fiber features to predict various time points of wound healing and delayed healing with high accuracy and visual interpretability, enhancing doctors' trust in model decisions. This could lead to more precise and effective wound treatment practices.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"191 ","pages":"Article 110110"},"PeriodicalIF":7.0,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143785468","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}