Computer methods and programs in biomedicine最新文献

{"title":"Predicting metabolite-disease associations based on dynamic adaptive feature learning architecture","authors":"Bo Wang ,&nbsp;Shiyu Liu ,&nbsp;Xiaoxin Du ,&nbsp;Jianfei Zang ,&nbsp;Chunyu Zhang ,&nbsp;Xue Yang ,&nbsp;Yang He","doi":"10.1016/j.cmpb.2025.108867","DOIUrl":"10.1016/j.cmpb.2025.108867","url":null,"abstract":"<div><h3>Background and Objective</h3><div>In recent years, the association between metabolites and complex human diseases has increasingly been recognized as a major research focus. Traditional wet-lab experiments are considered time-consuming and labor-intensive, while computational methods have been shown to significantly enhance research efficiency. However, existing methods for predicting metabolite-disease associations primarily depend on predefined similarity metrics and static network structures, often failing to capture the complex interactions among node neighborhoods within metabolite and disease networks. This limitation hinders the capture of deeper dynamic relationships between metabolites and diseases, resulting in information loss and noise that deteriorate prediction performance.</div></div><div><h3>Methods</h3><div>An innovative dynamic adaptive feature learning architecture (DAF-LA) is proposed to predict metabolite-disease associations. This architecture integrates dynamic subgraph construction and an adaptive feature enhancement mechanism, enabling high-precision feature learning and association prediction through progressive optimization from initial to high-order feature representations.</div></div><div><h3>Results</h3><div>The architecture was evaluated through five-fold cross-validation, achieving an AUC of 0.9742 and an AUPR of 0.9734. Additionally, the case study demonstrates that DAF-LA accurately predicts metabolites associated with Alzheimer's disease, Type 2 diabetes mellitus and Parkinson's disease.</div></div><div><h3>Conclusions</h3><div>The results demonstrate that our method effectively uncovers potential associations between metabolites and diseases through dynamic topological modeling and multi-scale collaborative learning. It enables faster identification of likely metabolite-disease relationships, reduces the time and resource costs associated with inefficient large-scale screening in traditional wet-lab experiments, and provides more targeted guidance for subsequent biological validation.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108867"},"PeriodicalIF":4.9,"publicationDate":"2025-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144134792","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 practical strategy for data assimilation of cerebral intra-aneurysmal flows using a variational method with boundary control of velocity","authors":"Tsubasa Ichimura ,&nbsp;Shigeki Yamada ,&nbsp;Yoshiyuki Watanabe ,&nbsp;Hiroto Kawano ,&nbsp;Satoshi Ii","doi":"10.1016/j.cmpb.2025.108861","DOIUrl":"10.1016/j.cmpb.2025.108861","url":null,"abstract":"<div><h3>Background and objective</h3><div>Evaluation of hemodynamics is crucial to predict growth and rupture of cerebral aneurysms. Variational data assimilation (DA) is a powerful tool to characterize patient-specific intra-aneurysmal flows. The DA inversely estimates a boundary condition in fluid equations using personalized flow data; however, its high computational cost in optimization problems makes its use impractical.</div></div><div><h3>Methods</h3><div>This study proposes a practical strategy for the DA to evaluate patient-specific intra-aneurysmal flows. To estimate personalized flows, a variational DA was combined with computational fluid dynamics (CFD) and four-dimensional flow magnetic resonance imaging (4D flow MRI) for intra-aneurysmal velocity data, and an inverse problem was solved to estimate the spatiotemporal velocity profile at a boundary of the aneurysm neck. To circumvent an ill-posed inverse problem, model order reduction based on a Fourier series expansion was used to describe temporal changes in state variables.</div></div><div><h3>Results</h3><div>In numerical validation using synthetic data from the CFD, the present DA achieved excellent agreement with the CFD as ground truth, with velocity mismatch within the 4%-7% range. In flow estimations for three patient-specific datasets, the proposed DA shows the velocity mismatch within the 35%-63% range, which is less than half that of the CFD using main vessel branches, and would mitigate unphysical velocity distributions in the 4D flow MRI.</div></div><div><h3>Conclusions</h3><div>By focusing only on the intra-aneurysmal region, the present strategy based on the DA provides an attractive way to evaluate personalized flows in aneurysms with greater reliability than conventional CFD and better efficiency than existing DA approaches.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108861"},"PeriodicalIF":4.9,"publicationDate":"2025-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144146782","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":"Understanding deep learning models for Length of Stay prediction on critically ill patients through latent space visualization","authors":"Lyse Naomi Wamba Momo ,&nbsp;Vincent Scheltjens ,&nbsp;Wouter Verbeke ,&nbsp;Frank Rademakers ,&nbsp;Bart De Moor","doi":"10.1016/j.cmpb.2025.108832","DOIUrl":"10.1016/j.cmpb.2025.108832","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Continuous, real-time monitoring of Length of Stay (LoS) for critically ill patients in Intensive Care Units (ICUs) is essential for anticipating patient needs, reduce the risk of adverse events, optimize resource allocation, plan incoming patients, and improve overall care. While previous research has focused primarily on predicting LoS, less attention has been given to how these prediction systems can be used by non-machine learning experts in real hospital environments for capacity planning.</div></div><div><h3>Methods:</h3><div>In this work, we predict remaining ICU LoS using data from the Amsterdam University Medical Center dataset, which is the first freely available European critical care dataset and shows higher variability in LoS compared to U.S. datasets. We applied state-of-the-art sequence-to-sequence deep learning models – Long Short-Term Memory Networks, Gated Recurrent Units, Temporal Convolutional Networks with and without attention, and Transformer models – on 271 input features extracted from 20,481 ICU stays. Additionally, the latent spaces of these models were extracted, projected onto a 2D space and explored interactively to intuitively understand how the models learn patterns from the clinical data over time.</div></div><div><h3>Results:</h3><div>The TCN model with attention (TCN-att) returned the best performance, reducing the prediction error by 2.24 days from a MAE of 6.94 days to a MAE of 4.70 days. Latent space analysis revealed that with just 5–6 h of patient data, the models could clearly differentiate between short (<span><math><mrow><mn>0</mn><mo>&lt;</mo><mtext>LoS</mtext><mo>≤</mo><mn>3</mn></mrow></math></span> days) and long (<span><math><mrow><mn>7</mn><mo>≤</mo><mtext>LoS</mtext><mo>&lt;</mo><mo>+</mo><mi>∞</mi></mrow></math></span> days) ICU stays, identify a patient cluster at risk of in-hospital mortality, and more.</div></div><div><h3>Conclusion:</h3><div>Beyond reporting prediction error metrics, this study shows how an interactive dashboard can be used to gain insights into the inner workings of complex algorithms and how these can be integrated into clinical decision support systems. This approach allows for real-time intuitive understanding of how models learn and how patient predictions evolve, facilitating their use in real hospital environments.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108832"},"PeriodicalIF":4.9,"publicationDate":"2025-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144123270","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":"ADEPT: An advanced data exploration and processing tool for clinical data insights","authors":"Lingyu Shao ,&nbsp;Jiarui Wang ,&nbsp;Lin Li ,&nbsp;Yujie Liu ,&nbsp;Zhaoqing Liu ,&nbsp;Boming Song ,&nbsp;Shuyan Li","doi":"10.1016/j.cmpb.2025.108860","DOIUrl":"10.1016/j.cmpb.2025.108860","url":null,"abstract":"<div><h3>Background and objective</h3><div>The rapid growth of clinical data creates challenges in analysis and interpretation for medical professionals. To address these issues, we developed the Advanced Data Exploration and Processing Tool (ADEPT), integrating data preprocessing, modeling, visualization, and statistical reporting to resolve common problems like inaccurate terminology, outliers, and missing values.</div></div><div><h3>Methods</h3><div>ADEPT incorporates advanced preprocessing, including standardizing numerical values, detecting outliers via Isolation Forest and DBSCAN, and filling missing data with KNN and MissForest. Tokenized text features are processed through keyword-based classification and K-means clustering. Five machine learning models—Gradient Boosting Machine, Random Forest, Extreme Gradient Boosting, Logistic Regression, and Support Vector Machine—are combined with a dynamic voting mechanism. Performance was assessed using precision, sensitivity, and specificity.</div></div><div><h3>Results</h3><div>ADEPT demonstrated substantial performance improvements, with the Area Under the Curve (AUC) increasing by over 14 %. Key results include enhanced precision, sensitivity, and specificity, validating the tool's ability to extract valuable insights from complex datasets.</div></div><div><h3>Conclusions</h3><div>ADEPT offers a comprehensive solution for automated clinical data analysis, combining rigorous preprocessing with advanced modeling. Its dynamic voting mechanism and integrated tools enhance accuracy and interpretability, addressing critical challenges in clinical data management and decision-making.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108860"},"PeriodicalIF":4.9,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144099302","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":"MT-RCAF: A Multi-Task Residual Cross Attention Framework for EEG-based emotion recognition and mood disorder detection","authors":"Xinni Kong ,&nbsp;Yaru Guo ,&nbsp;Yu Ouyang ,&nbsp;Wenjie Cheng ,&nbsp;Ming Tao ,&nbsp;Hong Zeng","doi":"10.1016/j.cmpb.2025.108835","DOIUrl":"10.1016/j.cmpb.2025.108835","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Prolonged abnormal emotions can gradually evolve into mood disorders such as anxiety and depression, making it critical to study the relationship between emotions and mood disorders to explore the causes of mood disorders. Existing research on EEG-based emotion recognition and mood disorder detection typically treats these two tasks separately, missing potential synergies between them. The purpose is to reveal the relationship between emotions and mood disorders and propose a Multi-Task Residual Cross Attention Framework (MT-RCAF) to enhance both classification performances.</div></div><div><h3>Methods:</h3><div>In MT-RCAF, the Feature Extraction module extracts specific and shared features for the corresponding tasks. The Residual Multi-head Cross Attention (RMCA) module dynamically adjusts attention weights to explicitly capture both shared and task-specific information, enhancing complementarity and feature sharing. The Gated Multi-embedding (GME) module filters out irrelevant features, improving task-specific performance. Finally, the Task Tower Classification module balances losses across tasks to facilitate both emotion recognition and mood disorder detection.</div></div><div><h3>Results:</h3><div>We conducted experiments on the DEAP dataset Black as well as the self-collected Emotion and Mood Disorder Dataset (EMDD) to validate the effectiveness of MT-RCAF. The results show that the framework gains improvement in strongly correlated task groups, with average accuracy increases of 3.22% for emotion recognition and 3.91% for mood disorder detection, and in generally correlated task groups, with average accuracy increases of 2.87% for valence and 3.34% for arousal. The study also reveals that mood disorders (depression or anxiety) increase sensitivity to negative emotions, and intense emotions enhance mood disorder detection.</div></div><div><h3>Conclusion:</h3><div>The study validates the relationship between emotions and mood disorders from a deep-learning perspective and finds that interconnected tasks result in more accurate and robust results.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108835"},"PeriodicalIF":4.9,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144083808","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":"Dual-modeling computational framework of magnetoelectric core-shell nanoparticles and nanochain for wireless peripheral nerve regeneration","authors":"Marta Bonato ,&nbsp;Valentina Galletta ,&nbsp;Emma Chiaramello ,&nbsp;Serena Fiocchi ,&nbsp;Marta Parazzini","doi":"10.1016/j.cmpb.2025.108862","DOIUrl":"10.1016/j.cmpb.2025.108862","url":null,"abstract":"<div><h3>Objective</h3><div>This study introduces a novel dual-modeling computational framework to comprehensively analyze magnetoelectric CFO-BTO core-shell nanoparticles and nanochains for wireless peripheral nerve regeneration. Our approach covers both single-particle physics and tissue-level effects for wireless peripheral nerve regeneration applications, evaluating the capability to generate therapeutic electric fields (5–140 V/m).</div></div><div><h3>Methods</h3><div>We developed a two-step computational strategy: first, a detailed multi-physics model characterizing individual nanostructure magnetoelectric coupling, incorporating experimental parameters from literature; second, a tissue-level model analyzing electric field distributions in both 2D in-vitro and 3D in-vivo configurations, with varying magnetic stimulations and nanochain concentrations (0.5–5 % w/v).</div></div><div><h3>Results</h3><div>The magnetoelectric analysis revealed better performance in nanochains compared to single nanoparticles, with optimal results achieved using cylindrical shell geometry. The 3D in-vivo configurations demonstrated higher field values compared to 2D in-vitro configuration, with therapeutic levels achieved across significant tissue volumes at higher nanochain concentrations.</div></div><div><h3>Conclusion</h3><div>This computational analysis validates that magnetoelectric nanochains can generate therapeutically relevant electric fields for nerve regeneration through wireless magnetic stimulation.</div></div><div><h3>Significance</h3><div>This study provides the first computational framework that quantitatively predicts electric field distributions from magnetoelectric nanostructures in neural tissue, providing essential design guidelines for optimizing wireless nerve regeneration therapies and accelerating their translation to clinical applications.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108862"},"PeriodicalIF":4.9,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144099305","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":"Dynamic cell-mass movement analyses tool","authors":"Zbyněk Dostál ,&nbsp;Veronika Žáková ,&nbsp;Pavel Veselý","doi":"10.1016/j.cmpb.2025.108829","DOIUrl":"10.1016/j.cmpb.2025.108829","url":null,"abstract":"<div><h3>Background and Objective</h3><div>Digital Holographic Microscopy provides a new kind of quantitative image data about live cells’ <em>in vitro</em> activities. Apart from non-invasive and staining-free imaging, it offers topological weighting of cell mass. This led us to develop a particular tool for assessing cell mass dynamics.</div></div><div><h3>Methods:</h3><div>Programming language Python and a training set of time-lapse images of adherent HT-1080 cells derived from human fibrosarcoma taken with dry objective 40x/0.95 at 30-second intervals were used to create the Analytical Image Differencing (AID) method.</div></div><div><h3>Results:</h3><div>The AID makes the best of these new data by evaluating the difference between the chosen two quantitative phase images from the time-lapse series. The contribution of the method is demonstrated on hiQPI (Holographic Incoherent-light-source Quantitative Phase Imaging) image data taken with a Q-phase microscope. The analysis outputs are graphical and complemented with numerical data.</div><div>To underscore the significance of the Analytical Image Differencing (AID) method, an initial pilot experiment was conducted to show the available analyses of sequential overlapping images capturing the movement of cancer cells. Notably, besides defining changes in areas used by the cell (newly or steadily occupied or better abandoned) it is an introduction to the zero-line concept, which denotes spots of tranquility among continuously moving surroundings.</div></div><div><h3>Conclusions:</h3><div>The measurement of zero-line length has emerged as a novel biomarker for characterizing cell mass transfer. The sensitivity of phase change measurements is demonstrated. The noise quality of input images obtained with incoherent (hiQPI) and coherent (QPI) methods is compared. The resulting effect on the AID method output is also shown.</div><div>The findings of this study introduce a novel approach to evaluating cellular behavior <em>in vitro</em>. The concept emerged as a particularly noteworthy outcome. Collectively, these results highlight the substantial potential of AID in advancing the field of cancer cells biology, particularly.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108829"},"PeriodicalIF":4.9,"publicationDate":"2025-05-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144099304","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":"SIMSAMU - A French medical dispatch dialog open dataset","authors":"Aimé Nun ,&nbsp;Olivier Birot ,&nbsp;Gaël Guibon ,&nbsp;Frédéric Lapostolle ,&nbsp;Ivan Lerner","doi":"10.1016/j.cmpb.2025.108857","DOIUrl":"10.1016/j.cmpb.2025.108857","url":null,"abstract":"<div><h3>Background</h3><div>Dispatch Services (DS) are essential to Emergency Medical Services (EMS). Dispatchers enable patients to access medical assistance in emergencies, anytime and anywhere, within limited time and resources. AI-based decision-support tools hold great promise for dispatchers. Developing these tools requires medical field-specific data. Medical dispatch dialogue is unique: it is a brief phone exchange in an emergency, within a limited time frame, without a physical examination.</div></div><div><h3>Objective</h3><div>Our main objective was to (i) create an open French dataset of medical dispatch dialogues. Our secondary objectives were to (ii) develop a detailed medical dispatch scheme from this dataset using an unsupervised method, and (iii) provide a baseline evaluation of diarization and speech recognition models for this domain in French.</div></div><div><h3>Methods</h3><div>From 2022 to 2023, emergency medicine junior doctors simulated real-life medical dispatch calls. These calls were recorded and transcribed to form the SIMSAMU corpus. We developed a dispatch scheme based on (i) recording analysis, (ii) data-driven utterance typology, and (iii) domain expertise. Utterance typology was derived via hierarchical clustering of representations learned by finetuning BERT embeddings on SIMSAMU. Clusters were mapped to the Roter Interaction Analysis System (RIAS) and included in our dispatch scheme. SIMSAMU was used to train and evaluate state-of-the-art neural network models for diarization and speech recognition. Diarization used the PyaNet model, fine-tuned on the ESLO2 dataset. Speech recognition used a CTC model with pre-trained wav2vec 2.0 embedding, compared to the multilingual Whisper model. The CTC-wav2vec model was further fine-tuned on SIMSAMU and evaluated by leave-one-speaker-out cross-validation.</div></div><div><h3>Results</h3><div>The dataset consists of 61 audio recordings totaling 3 h 14 min. Four clusters were identified for callers and 3 for dispatchers. Two main dialogue phases were identified: interrogation and contractualization. The diarization model achieved a 10.4 % error rate. Speech recognition word error rates were 35.8 % for Whisper, 24.8 % for the CTC-wav2vec model fine-tuned on ESLO2, and 16.1 % after in-domain fine-tuning.</div></div><div><h3>Conclusion</h3><div>We propose a French open medical dispatch dialogue dataset and an expert-validated schema of the medical dispatch dialogue based on unsupervised analysis. Notable gaps in how well speech recognition models generalize underscore the need for targeted, in-domain fine-tuning in this specialized application. SIMSAMU is designed to support this effort by serving as a benchmark for evaluating domain-adapted speech recognition and dialogue modeling strategies.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108857"},"PeriodicalIF":4.9,"publicationDate":"2025-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144107374","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 learning techniques for automated coronary artery segmentation and coronary artery disease detection: A systematic review of the last decade (2013-2024)","authors":"Suleyman Yaman ,&nbsp;Ozkan Aslan ,&nbsp;Hasan Güler ,&nbsp;Abdulkadir Sengur ,&nbsp;Abdul Hafeez-Baig ,&nbsp;Ru-San Tan ,&nbsp;Ravinesh C Deo ,&nbsp;Prabal Datta Barua ,&nbsp;U. Rajendra Acharya","doi":"10.1016/j.cmpb.2025.108858","DOIUrl":"10.1016/j.cmpb.2025.108858","url":null,"abstract":"<div><h3>Background</h3><div>Coronary artery disease (CAD) is the most common cardiovascular disease, exacting high morbidity and mortality worldwide. CAD is detected on coronary artery imaging; coronary artery segmentation (CAS) of the images is essential for coronary lesion characterization. Both CAD detection and CAS require expert input, are labor-intensive, and error-prone.</div></div><div><h3>Objectives</h3><div>Deep learning (DL) techniques have achieved significant success in CAS and CAD detection, with many studies published recently. This study is an up-to-date systematic review of research on automated DL models for CAS and CAD detection in the past decade (2013-2024).</div></div><div><h3>Method</h3><div>Using PRISMA methodology, an initial literature search of 1,589 publications was conducted, from which 97 high-impact Q1 studies were selected based on pre-defined eligibility criteria. These studies were analyzed in terms of DL techniques employed, datasets, modalities, and performance metrics.</div></div><div><h3>Results</h3><div>Of the 97 studies, most of which were published after 2016, 47 focused on CAS, 49 on CAD detection, and one on both tasks. CNN-based models were dominant in both domains. For CAS, CCTA was the most frequently used input modality, and U-Net was employed in 38 out of 48 studies, with recent works incorporating attention mechanisms and graph neural networks. ASOCA was the most widely used benchmark dataset. For CAD detection, ECG was the most common modality, with 45 out of 50 studies utilizing CNNs, and 20 of those relying purely on CNN architectures. Hybrid and multimodal models have become more prominent in recent years.</div></div><div><h3>Conclusion</h3><div>This review identified several challenges, including limited public datasets, variability in performance metrics, and model complexity. Future studies should focus on larger, diverse datasets and lightweight models integrating explainable artificial intelligence and uncertainty quantification to improve clinical applicability.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108858"},"PeriodicalIF":4.9,"publicationDate":"2025-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144115552","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":"The impact of clinical history on the predictive performance of machine learning and deep learning models for renal complications of diabetes","authors":"Davide Dei Cas ,&nbsp;Barbara Di Camillo ,&nbsp;Gian Paolo Fadini ,&nbsp;Giovanni Sparacino ,&nbsp;Enrico Longato","doi":"10.1016/j.cmpb.2025.108812","DOIUrl":"10.1016/j.cmpb.2025.108812","url":null,"abstract":"<div><h3>Background and Objective</h3><div>Diabetes is a chronic disease characterised by a high risk of developing diabetic nephropathy. The early identification of individuals at heightened risk of such complications or their exacerbation can be crucial to set a correct course of treatment. However, there are currently no widely accepted predictive tools for this task and, additionally, most of these models rely only on information at a single baseline visit. Considering this, we investigate the potential predictive role of patients’ clinical history over multiple levels of renal disease severity while, at the same time, developing an effective predictive model.</div></div><div><h3>Methods:</h3><div>From the data collected in the DARWIN–Renal (DApagliflozin Real-World evIdeNce-Renal) study, a nationwide multicentre retrospective real-world study, we develop four different types of machine learning models, namely, logistic regression, random forest, Cox proportional hazards regression, and a deep learning model based on recurrent neural network to predict the crossing of 5 clinically relevant glomerular filtration rate thresholds for patients with type 2 diabetes.</div></div><div><h3>Results:</h3><div>The predictive performance of all models is satisfactory for all outcomes, even without the introduction of information referring to past visits, with AUROC and C-index between 0.69 and 0.98 and average precision well above the random model. The introduction of past information results into a clear improvement in performance for all the models, with percentage increases of up to 12% for both AUROC and C-index and 300% for average precision. The usefulness of past information is further corroborated by a feature importance analysis.</div></div><div><h3>Conclusions:</h3><div>Incorporating data from the patients’ clinical history into the predictive models greatly improves their performance, particularly for recurrent neural network where the full sequence of values for dynamic variables is provided compared to synthetic indicators of past history.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"268 ","pages":"Article 108812"},"PeriodicalIF":4.9,"publicationDate":"2025-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144072264","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}