Computers in biology and medicine最新文献

{"title":"DNMT1 inhibition by synergistic application of curcumin and 5-AZA-2′-deoxycytidine implicates enhanced therapeutic potential against lung adenocarcinoma","authors":"Niharika ,&nbsp;Mukesh Kumar ,&nbsp;Anil Verma ,&nbsp;Pujarini Dash ,&nbsp;Ankan Roy ,&nbsp;Soumen Manna ,&nbsp;Tirthankar Baral ,&nbsp;Jagdish Mishra ,&nbsp;Subhajit Chakraborty ,&nbsp;Piyasa Nandi ,&nbsp;Prahallad Mishra ,&nbsp;Bhagyashree Pradhan ,&nbsp;Samir Kumar Patra","doi":"10.1016/j.compbiomed.2025.110763","DOIUrl":"10.1016/j.compbiomed.2025.110763","url":null,"abstract":"<div><div>DNA methyltransferase 1 (DNMT1) is a key epigenetic regulator that maintains DNA methylation patterns during cell division, regulating gene silencing and cancer progression. Aberrant DNMT1 activity is frequently observed in many cancers including lung adenocarcinoma, contributing to epigenetic dysregulation, chromatin remodelling, and the maintenance of cancer stem-like properties. This study investigates the potential impact of targeting DNMT1 using both nucleoside and non-nucleoside inhibitors, specifically 5-AZA-2′-deoxycytidine (hereafter, 5-AZA-CdR or AZA) and curcumin. Molecular docking and MD simulations revealed that curcumin has a high binding affinity to DNMT1 (−10.3 kcal/mol), stabilizing key active site residues of DNMT1. Additionally, 5-AZA-CdR showed low but significant interactions with DNMT1. Cell culture experiments using lung adenocarcinoma cell lines exhibited that treatment of cells with both the compounds, individual and/or synergistic, inhibits DNMT1 enzyme, significantly reduces cell viability, upregulates HDACs and DNMT1-specific miRNA; along with downregulation of DNMT1, SOX2, OCT4, and NANOG. Chromatin immunoprecipitation (ChIP) assay revealed increased recruitment of HDAC1 to the DNMT1 gene promoter region, indicating that HDAC1-mediated histone deacetylation may contribute to the transcriptional silencing of the DNMT1 gene. Correlation analyses indicated a strong association between DNMT1 and SOX2 expression, suggesting an interplay between DNA methylation and pluripotency transcription factors in lung cancer progression. These findings support that, both genetic and enzymatic level inhibition of DNMT1 by <span>AZA</span> and curcumin perform as a double-edged sword, (i) activation of tumor suppressors, and (ii) repression of oncogenes including DNMT1, which may be considered as a better therapeutic strategy for the treatment of lung adenocarcinoma.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110763"},"PeriodicalIF":7.0,"publicationDate":"2025-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144633011","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":"HVUNet: A hybrid vision transformer-based UNet for accurate detection and localization in histopathology images","authors":"Anusree Kanadath,&nbsp;Angel Arul Jothi J.,&nbsp;Siddhaling Urolagin","doi":"10.1016/j.compbiomed.2025.110680","DOIUrl":"10.1016/j.compbiomed.2025.110680","url":null,"abstract":"<div><div>Precise identification of object of interest (OoI) in histopathology images plays a vital role in cancer diagnosis and prognosis. Despite advances in digital pathology, detecting specific cellular structures within these images remains a significant challenge due to the inherent complexity and variability in cell morphology. Cellular structures exhibit similar visual characteristics, such as colors, shapes, and textures, making them difficult to distinguish from one another. Certain OoIs are much smaller than surrounding cells, rendering manual detection both challenging and error-prone. This paper introduces a hybrid vision transformer-based UNet (HVUNet) model, a novel approach designed to effectively identify and localize OoIs in histopathology images. To improve the detection in histopathology images, the proposed model incorporates UNet with vision transformers (ViTs) within an advanced encoder–decoder architecture. We evaluate HVUNet using the GZMH dataset, which includes histopathology images annotated for mitosis detection and the Lymphocyte detection (LD) dataset for lymphocyte cell detection. Through comprehensive experiments, we demonstrate that HVUNet notably surpasses several state-of-the-art models, including CNN variants, ViT-based models, and hybrid CNN-ViT architectures. Experimental results show that HVUNet outperforms traditional models such as UNet and recent advancements like UNETR and AttentionUNet, with a precision of 0.94, a recall of 0.60, and a F1-score of 0.72 for the GZMH dataset. Furthermore, HVUNet attained an Intersection over Union (IoU) score of 0.76 and a mean Average Precision (mAP) of 0.81, emphasizing its effectiveness in detecting mitotic cells. The model also achieved a F1-score of 0.76, an IoU of 0.63, and a mAP of 0.75, for the lymphocyte detection dataset demonstrating its effectiveness in detecting lymphocyte cells. To evaluate generalizability, we tested HVUNet on the MIDOG 2021 and PanopTILs datasets, observing competitive performance that demonstrated its robustness and broad applicability across diverse histopathology image analysis tasks.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110680"},"PeriodicalIF":7.0,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144632366","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":"Mobile apps to enhance student learning in medical education: A systematic search in app stores and evaluation using the mobile app rating scale","authors":"Sara Owrangi ,&nbsp;Nahid Zarifsanaiey ,&nbsp;Mohammadsadegh Rezaei","doi":"10.1016/j.compbiomed.2025.110740","DOIUrl":"10.1016/j.compbiomed.2025.110740","url":null,"abstract":"<div><h3>Background</h3><div>Mobile phone applications have emerged as a useful tool for integrating information and communication technologies into medical education. However, selecting a valuable application for medical education has become more complex and challenging.</div></div><div><h3>Objectives</h3><div>The objective of this research is to analyze and assess different medical education mobile apps, with the aim of offering support and enhancing medical courses, while also ensuring the creation of suitable learning experiences in the field of medical education.</div></div><div><h3>Methods</h3><div>The PRISMA method was employed to search for mobile applications on the Google Play website. Applications that met the established criteria were assessed using the Mobile Application Rating Scale (MARS) tool. Four evaluators independently downloaded and evaluated the applications. To ensure consensus among evaluators and measurement reliability, the internal correlation criterion (ICC) (Two-Way Mixed type) was calculated based on the evaluators' assessment results.</div></div><div><h3>Results</h3><div>Out of the initial 409 mobile phone applications identified, only 30 applications were considered suitable for the study. Many applications were excluded because they were not relevant to medical education or lacked content in English. The evaluation results from the experts showed a high level of reliability (ICC3k 0.97, 95 % CI 0.95 to 0.98). Among the assessed criteria of the MARS, the Embryology Quiz program achieved the highest score, with an average total score of 4.41. More than half of the applications (20 out of 30 applications) received an acceptable MARS score (&gt;0.3). The results indicate a weak correlation between the subjective quality (the likelihood of recommending the application to others for educational purposes in medicine) sub-criterion and the MARS score.</div></div><div><h3>Conclusion</h3><div>The medical education apps are generally excelling in terms of functionality, Information, Aesthetics, and Engagement as MARS sub-criteria. However, in the subjective quality criterion, the apps did not achieve a satisfactory score. While the apps have received acceptable scores in all MARS sub-criteria, they have fallen short in the criterion specifically assessing the program's effectiveness in medical education by subjective quality criterion. This suggests that certain key sub-criteria influencing program effectiveness in medical education may not be adequately captured by MARS sub-criteria alone. This research identifies the need for developers to enhance application customization in terms of training and user interaction. Additionally, the study recommends improvements to the MARS tool for evaluating the program's effectiveness in delivering impactful training.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110740"},"PeriodicalIF":7.0,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144632365","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":"Wearable technology for cardiovascular disease management: A global bibliometric analysis with emerging insights into artificial intelligence integration","authors":"Novita Rina Antarsih ,&nbsp;Kemal Nazaruddin Siregar ,&nbsp;Prihatin Oktivasari ,&nbsp;Bambang Budi Siswanto","doi":"10.1016/j.compbiomed.2025.110752","DOIUrl":"10.1016/j.compbiomed.2025.110752","url":null,"abstract":"<div><h3>Background and objective</h3><div>Wearable technology has become increasingly essential in managing cardiovascular disease (CVD), offering innovative solutions for real-time monitoring and personalized care. Artificial intelligence (AI) is playing a growing role in enhancing the capabilities of wearable devices, yet the global research trends and knowledge gaps in this area remain underexplored. This study aims to provide a comprehensive bibliometric analysis of wearable technology research for CVD management, with a specific focus on the integration and impact of AI.</div></div><div><h3>Methods</h3><div>We conducted a bibliometric analysis of literature published between 2014 and 2024, sourced from major academic databases. The analysis employed citation, co-citation, and co-word mapping techniques using tools such as VOSviewer and Bibliometrix to identify key studies, emerging themes, and research gaps in wearable technology and AI for CVD management.</div></div><div><h3>Results</h3><div>AI-powered wearables improve CVD diagnostics and patient outcomes, but challenges remain in clinical integration and data interoperability. These devices also play a crucial role in early atrial fibrillation (AF) detection, enhancing diagnostic accuracy and supporting timely medical interventions. AI-enhanced portable ECG technology further improves real-time decision-making in cardiovascular care, offering a transformative approach to personalized, evidence-based medicine.</div></div><div><h3>Conclusions</h3><div>AI integration in wearable technology is revolutionizing CVD management, offering precise, personalized care. However, challenges such as data security, algorithmic bias, and clinical validation persist. Ensuring privacy requires strong encryption and regulatory compliance. Large-scale trials, standardized data frameworks, and clinician training are essential to accelerate adoption, ensuring AI-powered wearables are effective, equitable, and sustainable in healthcare.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110752"},"PeriodicalIF":7.0,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144632367","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 design of a multi-epitope mRNA vaccine against orthopoxviruses: A path toward comprehensive poxvirus protection","authors":"Nafiseh Maghsoodi ,&nbsp;Navid Nezafat ,&nbsp;Amin Ramezani","doi":"10.1016/j.compbiomed.2025.110764","DOIUrl":"10.1016/j.compbiomed.2025.110764","url":null,"abstract":"<div><div>Along with the recent outbreak of monkeypox (MPOX), there are worries about other outbreaks of the poxvirus genus that threaten global public health. Until now, although there are no specific vaccines for MPOXV, there are no comprehensive vaccines for all dangerous poxviruses. This research utilizes immunoinformatics and structural vaccinology methodologies to develop a multi-epitope mRNA vaccine targeting variola virus (VARV), vaccinia virus (VACV), monkeypox virus (MPOXV), and cowpox virus (CPXV), which are four pathogenic orthopoxviruses (OPV). Accordingly, A29L, M1R, A35R, B6R, and F8L antigens were selected. Then, their sequences were retrieved and aligned to determine the conserved parts of each antigenic region among these four types of viruses. Different immunoinformatic methods were employed for forecasting B-cell, cytotoxic T lymphocytes (CTL), and helper T lymphocytes (HTL) epitopes. The epitopes were analyzed through a filtering process that checked for antigenicity, toxicity, allergenicity, and cytokine inducibility. The goal was to find epitopes that can make both T- and B-cells react. The vaccine was constructed and modeled via the trRosetta server. Molecular docking was employed between the epitopes and adjuvant receptors along with MHC type I and II molecules. Subsequently, the best structure with low energy was selected, and molecular dynamics (MD) simulations were performed. The immunological simulation data indicated that the developed vaccine possesses significant potential to provoke both cellular and humoral immune responses. Finally, with in vitro and in vivo experiments in the future, the designed vaccine could be a promising candidate for vaccination against pathogenic poxviruses.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110764"},"PeriodicalIF":7.0,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144613780","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":"Can your brain signals reveal your romantic emotions?","authors":"Dor Zazon ,&nbsp;Nir Nissim","doi":"10.1016/j.compbiomed.2025.110754","DOIUrl":"10.1016/j.compbiomed.2025.110754","url":null,"abstract":"<div><div>The process of partner selection may result in emotions of romantic attraction when one expresses interest towards a potential partner, and rejection when one receives negative feedback from a potential partner. Previous EEG studies have found distinct neural correlates for both emotions in the context of dating apps. However, to the best of our knowledge, no study has demonstrated the ability to predict the associated intra-subject romantic emotions based on a single-trial analysis of event related potential (ERP). In this study, 61 participants (31 females and 30 males) agreed to use our simulated dating app, and their EEG brain activity was recorded during their engagement with the app. Based on each participant's EEG signals, we induced multiple machine and deep learning models aimed at predicting single-trial romantic attraction and rejection for each participant. Our results show that the best model obtained 71.38 % and 81.31 % average ROC-AUC scores across the participants respectively for romantic attraction and rejection. We also found that our learning models were able to predict romantic emotions more accurately for picky participants than they could for those that were less fussy, which might suggest that picky people have stronger brain activity signals when it comes to romantic preference.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110754"},"PeriodicalIF":7.0,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144631387","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":"ML-based prediction to experimental validation: Development of dihydroquinazoline based multi-potent ligands as anti-Alzheimer's agents","authors":"Kailash Jangid ,&nbsp;Bharti Devi ,&nbsp;Naveen Kumar ,&nbsp;Shubham Upadhayay ,&nbsp;Vinay Kumar ,&nbsp;Suresh Thareja ,&nbsp;Vinod Kumar","doi":"10.1016/j.compbiomed.2025.110762","DOIUrl":"10.1016/j.compbiomed.2025.110762","url":null,"abstract":"<div><div>Alzheimer's disease (AD) is a multifactorial neurological disorder accounting for the cognitive decline in the patients. The disease is linked to numerous pathological factors including hyperactivation of acetylcholinesterase (AChE) and monoamine oxidase B (MAO-B), accumulation of amyloid-beta (A<em>β</em>) plaques and neurofibrillary tangles in the brain etc. The single-target medications already available in the market are found to be ineffective and the research focus is shifting towards the development of multitargeting agents. In order to find a multi-potent inhibitor against AChE, MAO-B and A<em>β</em>_1-42, in the present study we employed a machine learning-based tool, PyRMD, to screen an in-house generated library of dihydroquinazoline derivatives. This screening process identified six promising compounds, KV-271, KV-832, KV-968, KV-1131, KV-1159, KV-1234 with dual inhibition potential against AChE and MAO-B enzymes. In the docking studies, these compounds showed good interactions at the active cavity of the AChE and MAO-B comparable to the standard inhibitors donepezil (AChE) and pargyline (MAO-B). To validate these predictions, the six identified compounds were synthesized and subjected to <em>in vitro</em> enzymatic assays. All the six compounds displayed significant inhibitory activity, with IC₅₀ values below 5 μM for both AChE and MAO-B. Amongst these compounds, KV-1131 and KV-1234 were found to be the most potent inhibitors with IC₅₀ values of 0.93 μM and 0.85 μM against AChE and IC₅₀ values of 1.17 μM and 0.79 μM against MAO-B, respectively. In addition, KV-1131 and KV-1234 exhibited inhibitory activity against A<em>β</em>_1-42 self-aggregation inhibition of 34.79 % and 45.70 %, respectively, after 48 h of incubation. Both KV-1131 and KV-1234 were found to be non-toxic up to 10 μM concentration and showed neuroprotective potential against 6-hydroxydopamine induced neurotoxicity in the SHSY-5Y cells. Thus, KV-1131 and KV-1234 were identified as potent leads that can be developed as drug candidates for the treatment of Alzheimer's disease.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110762"},"PeriodicalIF":7.0,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144631388","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":"Therapeutic targeting of triple-negative breast cancer: A multi-model evaluation of LNA-anti-miR-19b-3p and small molecule inhibitors","authors":"Mohammad Javad Kamali ,&nbsp;Fatemeh Saeedi ,&nbsp;Azin Khoshghiafeh ,&nbsp;Mahsa Aghajani Mir ,&nbsp;Cena Aram ,&nbsp;Mohamadreza Ahmadifard","doi":"10.1016/j.compbiomed.2025.110771","DOIUrl":"10.1016/j.compbiomed.2025.110771","url":null,"abstract":"<div><div>The potential of inhibiting hsa-miR-19b-3p as a therapeutic approach for triple-negative breast cancer (TNBC) was investigated. To explore this, we studied the function of hsa-miR-19b-3p in TNBC cells, specifically the MDA-MB-231 cell line. We transfected these cells with LNA-anti-miR to inhibit the miRNA and then used qRT-PCR to measure the level of inhibition. Interestingly, we observed over 95 % inhibition at 24 h, and even at 48 h, the inhibition remained high at over 75 %. Upon closer examination, we determined the impact of this inhibition on cell viability using MTT assays, which showed a significant decrease in cell numbers following treatment with LNA-anti-miR-19b-3p. Furthermore, our apoptosis analysis, using Annexin V/Propidium iodide staining, revealed an increased apoptosis rate in the transfected cells compared to the controls. Alongside these experimental studies, we employed computational methods to investigate hsa-miR-19b-3p in greater detail, including RNA-Seq analysis of TCGA data, which identified 2585 upregulated and 4251 downregulated genes. Cross-referencing downregulated genes with target genes from miRTarBase, RNAInter, and miRWalk led to the identification of four potential hsa-miR-19b-3p targets: <em>TMTC1</em>, <em>MBNL3</em>, <em>FAT3</em>, and <em>GFOD1</em>, with <em>TMTC1</em> and <em>MBNL3</em> showing statistically significant downregulation. Additionally, we screened for potential small molecule inhibitors, identifying four promising candidates, including Dovitinib, S-Adenosylmethionine, Guanosine-5′,3′-tetraphosphate, and Neomycin, which exhibited favorable drug-like characteristics. In conclusion, our multifaceted approach demonstrates the significant potential of LNA-anti-miR-19b-3p as a therapeutic option for TNBC patients, and the small molecule inhibitors we've uncovered could open new avenues for treating this aggressive form of breast cancer.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110771"},"PeriodicalIF":7.0,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144614913","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":"Deep siamese residual support vector machine with applications to disease prediction","authors":"Xinjia Yang,&nbsp;Pinchao Meng,&nbsp;Zhixia Jiang,&nbsp;Linhua Zhou","doi":"10.1016/j.compbiomed.2025.110693","DOIUrl":"10.1016/j.compbiomed.2025.110693","url":null,"abstract":"<div><div>Support Vector Machines (SVMs) excel in classification and regression tasks involving high-dimensional nonlinear data, boasting high accuracy, strong generalization ability, and robust performance. Particularly noteworthy is their outstanding performance when integrated into end-to-end collaborative frameworks with deep learning models. However, such frameworks often leverage the strengths of deep learning and SVMs separately. To achieve a synergistic effect that surpasses the mere sum of its parts, this paper proposes an end-to-end learning model called the Deep Siamese Residual Support Vector Machine (DSRSVM), which integrates deep neural networks and SVMs. The learning process of the model consists of two phases: deep residual network siamese pre-training and deep residual support vector machine fine-tuning. During the deep siamese pre-training phase, the model leverages the deep residual network to capture similarities and differences in data features. Subsequently, an SVM is embedded within the pre-trained deep residual network to construct the DSRSVM model. The SVM loss is then propagated backward to the deep residual neural network using a gradient descent algorithm, enabling end-to-end learning of the DSRSVM. This paper presents the DSRSVM training algorithm and provides a theoretical proof of its convergence. The algorithm was validated on publicly available medical datasets, demonstrating superior performance in prediction accuracy, recall, and F1 score compared to traditional end-to-end collaborative frameworks of deep learning and SVMs. These results affirm that the DSRSVM model achieves a synergistic improvement, exemplifying the principle of \"Synergy creates greater outcomes\".</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110693"},"PeriodicalIF":7.0,"publicationDate":"2025-07-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144613781","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":"Robustness evaluation against corruptions for Optical Diffraction Tomography-based classifiers","authors":"Hyungjoo Cho ,&nbsp;Jimin Lee ,&nbsp;Dongmin Ryu ,&nbsp;Juhyeong Ki ,&nbsp;Sung-Joon Ye","doi":"10.1016/j.compbiomed.2025.110682","DOIUrl":"10.1016/j.compbiomed.2025.110682","url":null,"abstract":"<div><div>Optical Diffraction Tomography (ODT) is a promising technique for three-dimensional imaging, but practical use demands rigorous robustness testing due to real-world noise factors. Despite the growing importance of machine learning safety, robustness in ODT remains underexplored. We propose the first comprehensive robustness testing protocol for ODT-based classifiers by simulating 16 corruption scenarios to create a corrupted dataset. To enhance robustness and accuracy, we introduce <span>CutPix</span>, a data augmentation strategy that balances shape and texture information through fractal pattern mixing and a cut-and-concatenate approach. Our experiments show that <span>CutPix</span> significantly improves robustness under various corrupted environments compared to existing techniques, particularly against pattern noises. All code and corruption simulation scripts are publicly available at <span><span>https://github.com/NySunShine/odt-robustness-evaluation</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"196 ","pages":"Article 110682"},"PeriodicalIF":7.0,"publicationDate":"2025-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144613863","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}