Computers in biology and medicine最新文献

{"title":"Meta-analysis of AI-based pulmonary embolism detection: How reliable are deep learning models?","authors":"Ezio Lanza ,&nbsp;Angela Ammirabile ,&nbsp;Marco Francone","doi":"10.1016/j.compbiomed.2025.110402","DOIUrl":"10.1016/j.compbiomed.2025.110402","url":null,"abstract":"<div><h3>Rationale and objectives</h3><div>Deep learning (DL)–based methods show promise in detecting pulmonary embolism (PE) on CT pulmonary angiography (CTPA), potentially improving diagnostic accuracy and workflow efficiency. This meta-analysis aimed to (1) determine pooled performance estimates of DL algorithms for PE detection; and (2) compare the diagnostic efficacy of convolutional neural network (CNN)– versus U-Net–based architectures.</div></div><div><h3>Materials and methods</h3><div>Following PRISMA guidelines, we searched PubMed and EMBASE through April 15, 2025 for English‐language studies (2010–2025) reporting DL models for PE detection with extractable 2 × 2 data or performance metrics. True/false positives and negatives were reconstructed when necessary under an assumed 50 % PE prevalence (with 0.5 continuity correction). We approximated AUROC as the mean of sensitivity and specificity if not directly reported. Sensitivity, specificity, accuracy, PPV and NPV were pooled using a DerSimonian–Laird random-effects model with Freeman-Tukey transformation; AUROC values were combined via a fixed-effect inverse-variance approach. Heterogeneity was assessed by Cochran's Q and I². Subgroup analyses contrasted CNN versus U-Net models.</div></div><div><h3>Results</h3><div>Twenty-four studies (n = 22,984 patients) met inclusion criteria. Pooled estimates were: AUROC 0.895 (95 % CI: 0.874–0.917), sensitivity 0.894 (0.856–0.923), specificity 0.871 (0.831–0.903), accuracy 0.857 (0.833–0.882), PPV 0.832 (0.794–0.869) and NPV 0.902 (0.874–0.929). Between-study heterogeneity was high (I² ≈ 97 % for sensitivity/specificity). U-Net models exhibited higher sensitivity (0.899 vs 0.893) and CNN models higher specificity (0.926 vs 0.900); subgroup Q‐tests confirmed significant differences for both sensitivity (p = 0.0002) and specificity (p &lt; 0.001).</div></div><div><h3>Conclusions</h3><div>DL algorithms demonstrate high diagnostic accuracy for PE detection on CTPA, with complementary strengths: U-Net architectures excel in true-positive identification, whereas CNNs yield fewer false positives. However, marked heterogeneity underscores the need for standardized, prospective validation before routine clinical implementation.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110402"},"PeriodicalIF":7.0,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144116249","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multimodal brain-body analysis of prefrontal cortex activity and postural sway with sensory manipulation","authors":"Yasaman Baradaran ,&nbsp;Raul Fernandez Rojas ,&nbsp;Roland Goecke ,&nbsp;Maryam Ghahramani","doi":"10.1016/j.compbiomed.2025.110338","DOIUrl":"10.1016/j.compbiomed.2025.110338","url":null,"abstract":"<div><div>The involvement of different brain areas in postural balance control under various sensory inputs remains unknown. This study aims to investigate the effects of different sensory manipulations on prefrontal cortex (PFC) activity and postural sway as well as the associations between PFC activity and postural sway in different sensory conditions. To this end, younger participants underwent eight standing tests with single (visual, vestibular, and somatosensory) and double sensory manipulations. Functional near-infrared spectroscopy was employed to capture their cortical activation in the dorsolateral (DL), ventrolateral (VL), and dorsomedial (DM) PFC. An inertial measurement unit was also used to assess their postural sway. Results showed greater DMPFC and DLPFC activation with vestibular, somatosensory, and double sensory manipulations, and increased DMPFC activation with visual manipulation. These results indicate that the DMPFC and DLPFC are more involved in sensory integration during standing. Postural sway measures increased in response to double and some single sensory manipulations, implying greater regulatory activity and reduced postural stability and sway smoothness. Some meaningful correlations were also found between PFC activity and postural sway measures. Increased DMPFC activation during somatosensory and vestibular manipulation was correlated with better postural stability, highlighting its role in these sensory conditions. A negative correlation was found between DMPFC activity and postural sway in the anteroposterior direction, but not the mediolateral, indicating that DMPFC activity modulation depends on sway direction. The findings of this study lead to a better understanding of the functional architecture of the PFC and are useful for designing better clinical assessments, rehabilitation programs, and assistive devices.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110338"},"PeriodicalIF":7.0,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144124317","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":"Lightweight triangular mesh deformable reconstruction for low quality 3D organ models: Thickness noise and uneven topology","authors":"Yanjie Xu ,&nbsp;Tianmu Wang ,&nbsp;Zheng Xu ,&nbsp;Boxing Su ,&nbsp;Jianxing Li ,&nbsp;Zhenguo Nie","doi":"10.1016/j.compbiomed.2025.110328","DOIUrl":"10.1016/j.compbiomed.2025.110328","url":null,"abstract":"<div><div>Lightweight triangular mesh models have great potential for real-time 3D visualization of lesions during minimally invasive surgery (MIS). However, the blurred tissue boundaries, high imaging noise, and unoriented points in medical images seriously affect the accuracy and topological quality of surface reconstruction, which can lead to inaccurate lesion localization. In this paper, we present a robust and high-topology-quality triangular mesh reconstruction method that aims to provide a deformable expression model for real-time 3D visualization during surgery. Our approach begins by approximating the model prototype under the guidance of an unsigned distance field by simulating inflation. Then, we introduce a variance-controlled cylindrical domain projection search (VC-CDPS) method to achieve the final surface fitting. Additionally, we incorporate topology optimization into the iterative reconstruction process to ensure smoothness and good topology of the reconstruction model. To validate our method, we conduct experiments on a geometric model with high noise and a human organ model manually segmented by novice doctors. The results demonstrate that our reconstructed model exhibits better surface quality and noise immunity. Furthermore, we conduct a comparison experiment of model deformation and propose a metric to measure the topological quality of the model. Through in vitro tissue experiments, we explored the relationship between topological quality and deformation accuracy. The results reveal a positive correlation between deformation accuracy and topological quality.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110328"},"PeriodicalIF":7.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144106736","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":"Augmenting Common Spatial Patterns to deep learning networks for improved alcoholism detection using EEG signals","authors":"Neeraj ,&nbsp;Vatsal Singhal ,&nbsp;Jimson Mathew ,&nbsp;Ranjan Kumar Behera","doi":"10.1016/j.compbiomed.2025.110359","DOIUrl":"10.1016/j.compbiomed.2025.110359","url":null,"abstract":"<div><div>One of the main risk factors for numerous health problems is excessive drinking. Alcoholism is a severe disorder that can affect a person’s thinking and cognitive abilities. Early detection of alcoholism can help the subject regain control over their drinking habits and help them recover faster. The diagnosis of alcoholism is challenging, primarily when standard diagnostic tests rely on blood tests and questionnaires that are subjective to the patient and the examiner. This poses the need for fast, reliable, automatic, and preferably non-invasive ways to detect alcoholism. A non-invasive method to capture the electrical activity of the brain is the electroencephalogram (EEG). It can help detect alcohol use disorders in a subject. To our knowledge, no previous work in the literature proposes a mechanism to detect alcoholism using an EEG signal using a deep learning model that considers the spatio-temporal nature of EEG signals. Two works in the literature using a deep learning architecture are Fayyaz et al. (2019) and Farsi et al. (2021). However, both use a simple deep learning architecture and ignore the spatio-temporal nature of multichannel EEG signals. This paper suggests a new hybrid architecture called CSP-CNN-LSTM-ATTN. To obtain nonlinear and nonstationary spatio-temporal features from EEG signals that can accurately classify them as either alcoholic or control (not alcoholic). Our method integrates Common spatial patterns (CSP) for feature extraction, convolutional neural networks (CNN) for spatial representation, long-short-term memory (LSTM) for temporal learning, and attention networks (ATTN) for feature weighting, enhancing classification performance. Experiments on the publicly available UCI EEG dataset demonstrate that our model achieves state-of-the-art results, outperforming existing methods with an accuracy of 98.60%, F1-score 98.59, recall 98.24, precision 98.95, MCC 97.20, and AUC 99.64. The source code for this study is available at Github repository <span><span>https://github.com/n28neeraj/Alcoholism-Detectio</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110359"},"PeriodicalIF":7.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144116250","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":"Trends and advances in image-based mosquito identification and classification using machine learning models: A systematic review","authors":"Alice Bagyiereyele Lakyiere ,&nbsp;Rose-Mary Owusuaa Gyening Mensah ,&nbsp;Nutifafa Yao Agbenor-Efunam ,&nbsp;Edmund Yamba ,&nbsp;Kingsley Badu","doi":"10.1016/j.compbiomed.2025.110373","DOIUrl":"10.1016/j.compbiomed.2025.110373","url":null,"abstract":"<div><div>Mosquito-borne diseases, such as Yellow fever, Dengue, and Zika, pose a significant global health threat, causing millions of deaths annually. Traditional mosquito identification methods, reliant on expert analysis, are time-consuming and resource-intensive. Machine Learning (ML) has emerged as a transformative solution, enabling rapid and accurate species identification and classification. Recent studies leverage morphological features, such as wings and body structures, to determine species, sex, and age. These innovations aim to revolutionize vector control strategies, making them faster, more accurate, and widely accessible. This systematic review evaluates ML-based mosquito identification research, highlighting its strengths, limitations, and geographic disparities in contributions. Data was collected from Google Scholar, PubHub, IEEE Xplore, and ScienceDirect (2000–2024), with 52 studies meeting the inclusion criteria out of an initial pool of 1,050 papers. A key highlight of this review is the role of feature extraction techniques in achieving high classification accuracy by capturing fine-grained morphological traits. The findings also reveal critical limitations that hinder real-world applicability. These include limited dataset diversity, inconsistent preprocessing practices across devices, all of which reduce the generalizability of models in varied environments. Furthermore, high computational requirements and morphological similarities between certain species challenge the scalability and robustness of machine learning models. To address these gaps, measures such as expanding annotated and diverse datasets, investing in low-resource model deployment strategies, and supporting African-led research initiatives can be utilized to ensure more inclusive and context-relevant mosquito surveillance systems.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110373"},"PeriodicalIF":7.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144116761","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 multimodal deep learning framework for enzyme turnover prediction with missing modality","authors":"Xin Sun ,&nbsp;Yu Guang Wang ,&nbsp;Yiqing Shen","doi":"10.1016/j.compbiomed.2025.110348","DOIUrl":"10.1016/j.compbiomed.2025.110348","url":null,"abstract":"<div><div>Accurate prediction of the turnover number (<span><math><msub><mrow><mi>k</mi></mrow><mrow><mi>cat</mi></mrow></msub></math></span>), which quantifies the maximum rate of substrate conversion at an enzyme’s active site, is essential for assessing catalytic efficiency and understanding biochemical reaction mechanisms. Traditional wet-lab measurements of <span><math><msub><mrow><mi>k</mi></mrow><mrow><mi>cat</mi></mrow></msub></math></span> are time-consuming and resource-intensive, making deep learning (DL) methods an appealing alternative. However, existing DL models often overlook the impact of reaction products on <span><math><msub><mrow><mi>k</mi></mrow><mrow><mi>cat</mi></mrow></msub></math></span> due to feedback inhibition, resulting in suboptimal performance. The multimodal nature of this <span><math><msub><mrow><mi>k</mi></mrow><mrow><mi>cat</mi></mrow></msub></math></span> prediction task, involving enzymes, substrates, and products as inputs, presents additional challenges when certain modalities are unavailable during inference due to incomplete data or experimental constraints, leading to the inapplicability of existing DL models. To address these limitations, we introduce <strong>MMKcat</strong>, a novel framework employing a prior-knowledge-guided missing modality training mechanism, which treats substrates and enzyme sequences as essential inputs while considering other modalities as maskable terms. Moreover, an innovative auxiliary regularizer is incorporated to encourage the learning of informative features from various modal combinations, enabling robust predictions even with incomplete multimodal inputs. We demonstrate the superior performance of MMKcat compared to state-of-the-art methods, including DLKcat, TurNup, UniKP, EITLEM-Kinetic, DLTKcat and GELKcat, using BRENDA and SABIO-RK. Our results show significant improvements under both complete and missing modality scenarios in RMSE, <span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span>, and SRCC metrics, with average improvements of 6.41%, 22.18%, and 8.15%, respectively. Codes are available at <span><span>https://github.com/ProEcho1/MMKcat</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110348"},"PeriodicalIF":7.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144107113","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 hybrid frequency-spatial domain unsupervised denoising model for Gaussian-Poisson mixed noise in medical imaging","authors":"Cheng Zhang,&nbsp;Kin Sam Yen","doi":"10.1016/j.compbiomed.2025.110374","DOIUrl":"10.1016/j.compbiomed.2025.110374","url":null,"abstract":"<div><div>This paper proposes an unsupervised denoising model designed to address Gaussian-Poisson hybrid noise in CT, MRI, and X-ray images. Traditional deep image prior (DIP) methods suffer from slow convergence, spectral bias, and overfitting, limiting their clinical applicability. In this paper, by applying the Fourier transform, we incorporate frequency-domain priors extracted from the observed noisy image at the input stage. Instead of using both amplitude and phase, we rely solely on the amplitude spectrum, which captures the energy distribution of various frequency components while avoiding the instability associated with phase information. Meanwhile, we retain the spatial domain information to preserve the image's structural integrity, ensuring the effective capture of both low-frequency details and high-frequency anatomical features. This dual-domain strategy allows fine details to be captured early in training, thereby mitigating spectral bias, accelerating convergence, and improving the preservation of high-frequency anatomical structures. To further enhance diagnostic fidelity, we replace the conventional mean squared error (MSE) loss with an edge-aware L1 loss function that better preserves critical anatomical textures. Additionally, an entropy-based criterion tracks variations in image uncertainty over iterations to determine the optimal stopping point, effectively preventing overfitting without the need for external validation data. Experimental results demonstrate that our model achieves an average improvement of 10.7 % in PSNR and 17.9 % in SSIM compared to DIP, reaching peak performance in just 60 iterations, faster than the 1360 and 2990 iterations required by DIP and DIP-AITV, respectively. These findings highlight the efficiency and effectiveness of our method for medical image denoising.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110374"},"PeriodicalIF":7.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144116246","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 different material types on ergonomics in lower extremity exoskeleton construction","authors":"İsmail Çalıkuşu ,&nbsp;Ugur Fidan","doi":"10.1016/j.compbiomed.2025.110403","DOIUrl":"10.1016/j.compbiomed.2025.110403","url":null,"abstract":"<div><div>This study examines the effects of materials such as A Glass Fiber, Aluminum Alloy, Stainless Steel, S Glass Fiber, C Graphite, Hexcel, and Thornel on biomechanical performance in the design of lower extremity exoskeletons. Exoskeleton models created using Computer-Aided Modeling software were integrated into the AnyBody Modeling System and combined with a full-body human model to conduct walking simulations. In these simulations, femur and tibia segments were also incorporated into the model to analyze the impacts of the exoskeleton on human movement dynamics in detail. The results reveal that material selection significantly influences joint reaction forces and moments, ground reaction forces, and contact forces. Flexible materials were found to provide greater comfort to the user but demonstrated lower durability performance. Conversely, more durable materials improved overall efficiency by reducing load transfer. These findings emphasize that material selection in exoskeleton design plays a critical role not only in durability and performance but also in meeting ergonomic requirements. This research offers a valuable foundation for developing exoskeleton designs tailored to different user groups and highlights the need to customize material selection to optimize biomechanical performance. The study serves as a guide for enhancing the compatibility of exoskeletons with human movement dynamics and improving user comfort.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110403"},"PeriodicalIF":7.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144116247","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":"CancerNet: A comprehensive deep learning framework for precise and intelligible cancer identification","authors":"S.M. Nuruzzaman Nobel ,&nbsp;Shirin Sultana ,&nbsp;Md All Moon Tasir ,&nbsp;M.F. Mridha ,&nbsp;Zeyar Aung","doi":"10.1016/j.compbiomed.2025.110339","DOIUrl":"10.1016/j.compbiomed.2025.110339","url":null,"abstract":"<div><div>The medical community continually seeks innovative solutions to address healthcare challenges, particularly in cancer detection. A promising approach involves the use of Artificial Intelligence (AI) techniques, specifically Deep Learning (DL) models. This research introduces CancerNet, incorporating convolutional, involutional, and transformer components to extract hierarchical features and capture long-range dependencies from medical imaging data across the channel and spatial domains. CancerNet was trained and evaluated on an extensive dataset of histopathological images (HI) of tumor tissues and validated on the DeepHisto dataset, which comprises whole slide images (WSI) of various subtypes of glioma. CancerNet surpasses other comparative models and, achieves a higher accuracy on both datasets. CancerNet exhibits robustness across various imaging conditions, thereby ensuring reliable performance in various clinical scenarios. By integrating Explainable AI (XAI) techniques, CancerNet enhances transparency in its decision-making process, improves understanding and fosters trust in clinical adoption. CancerNet achieved an accuracy of 98.77% on the Histopathological Image dataset and 97.83% on the DeepHisto validation dataset, proving to be more effective than previous. Furthermore, transparency in AI models is crucial as it enhances healthcare professionals ability to understand and trust the model’s decision-making process, facilitating their adoption in clinical settings.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110339"},"PeriodicalIF":7.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144106737","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 rule-based clinical decision support system for detection of acute kidney injury after pediatric cardiac surgery","authors":"Janice Wachenbrunner ,&nbsp;Marcel Mast ,&nbsp;Julia Böhnke ,&nbsp;Nicole Rübsamen ,&nbsp;Louisa Bode ,&nbsp;André Karch ,&nbsp;Henning Rathert ,&nbsp;Alexander Horke ,&nbsp;Philipp Beerbaum ,&nbsp;Michael Marschollek ,&nbsp;Thomas Jack ,&nbsp;Martin Böhne","doi":"10.1016/j.compbiomed.2025.110382","DOIUrl":"10.1016/j.compbiomed.2025.110382","url":null,"abstract":"<div><h3>Background</h3><div>Acute kidney injury (AKI) is common in children with congenital heart disease following open-heart surgery with cardiopulmonary bypass (CPB). Early AKI detection in critically ill children requires clinician expertise to compile various data from different sources within a stressful and time-sensitive environment. However, as electronic health records provide data in a machine-readable format, this process could be supported by computerized systems. Therefore, we developed a time-aware, rule-based clinical decision support system (CDSS) to detect, stage, and track temporal AKI progression in children.</div></div><div><h3>Methods</h3><div>We integrated retrospective clinical routine data from n = 290 randomly selected cases (n = 263 patients, aged 0–17 years) who underwent cardiac surgery with CPB into a dataset. We adapted <em>Kidney Disease: Improving Global Outcome</em> (KDIGO) criteria, including serum creatinine, urine output, and estimated glomerular filtration rate, and translated them into computable rules for the CDSS. As a reference standard, patients were manually assessed by blinded clinical experts.</div></div><div><h3>Results</h3><div>The AKI incidence, according to the reference standard, was n = 146 cases for stage 1, n = 58 for stage 2, and n = 20 for stage 3. The CDSS achieved sensitivities of 92.2 % (95 % CI: 86.8–95.5 %) for AKI stage 1, 88.1 % (95 % CI: 77.2–94.2 %) for stage 2, and 95 % (95 % CI: 70.5–99.3 %) for stage 3. The specificities were 97.0 % (95 % CI: 94.4–98.4 %), 98.5 % (95 % CI: 96.5–99.4 %), and 99.3 % (95 % CI: 97.3–99.8 %), respectively.</div></div><div><h3>Conclusions</h3><div>We demonstrated that a CDSS is able to perform a complex AKI detection and staging process, including 11 criteria across three stages. For accurate automated AKI detection, standardized machine-readable data of high data quality are required. CDSS with high diagnostic accuracy, like presented, can support clinical management and be used for surveillance and quality management. The prototypical use for surveillance and further studies, such as the development of prediction models, should demonstrate the system's benefits in the future.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"193 ","pages":"Article 110382"},"PeriodicalIF":7.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144107111","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}