Computers in biology and medicine最新文献

{"title":"Optimizing clustering-based analytical methods with trimmed and sparse clustering","authors":"José Antonio Bernabé-Díaz ,&nbsp;Manuel Franco ,&nbsp;Juana-María Vivo ,&nbsp;Jesualdo Tomás Fernández-Breis","doi":"10.1016/j.compbiomed.2025.110436","DOIUrl":"10.1016/j.compbiomed.2025.110436","url":null,"abstract":"<div><div>Clustering is an essential tool in biomedical research, often used to identify patterns and subgroups within complex, high-dimensional datasets, such as gene expression profiles, metabolomics, and patient stratification data. However, searching the optimal number of clusters and other input parameters such as trimmed and sparse represent challenging tasks. Traditional clustering methods may struggle to handle noisy, outliers, redundancy, and high-dimensional data, which are common in biomedical applications, leading to unreliable or biologically uninterpretable results.</div><div>Sparse clustering methods help by emphasizing significant features while suppressing noise, and trimmed clustering can enhance robustness by excluding outliers. Yet, existing approaches often require manual tuning of parameters, such as the trimming proportion, and the sparsity level, which can be time-consuming and based on a trial-and-error approach.</div><div>To address these limitations, this work presents an automated trimmed and sparse clustering method, which automatically determines both the optimal number of clusters and the necessary tuning parameters. Our method has been made available to the biomedical community through the <em>evaluomeR</em> package, which enables researchers to efficiently implement sophisticated clustering without extensive computational background. This advancement not only increases the usability of trimmed and sparse clustering, but also promotes reproducibility and accuracy in data-driven biomedical discoveries.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110436"},"PeriodicalIF":7.0,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144291314","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":"Artificial neural networks computing for heat transfer flow of hybrid nanofluid in rectangular geometry","authors":"Saraj Khan ,&nbsp;Muhammad Imran Asjad ,&nbsp;F. Maiz","doi":"10.1016/j.compbiomed.2025.110475","DOIUrl":"10.1016/j.compbiomed.2025.110475","url":null,"abstract":"&lt;div&gt;&lt;div&gt;This study explores the complex dynamics of heat transfer in hybrid nanofluid flow, focusing on the unsteady squeezing motion of Graphene–Fe&lt;span&gt;&lt;math&gt;&lt;msub&gt;&lt;mrow&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mn&gt;3&lt;/mn&gt;&lt;/mrow&gt;&lt;/msub&gt;&lt;/math&gt;&lt;/span&gt;O&lt;span&gt;&lt;math&gt;&lt;msub&gt;&lt;mrow&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mn&gt;4&lt;/mn&gt;&lt;/mrow&gt;&lt;/msub&gt;&lt;/math&gt;&lt;/span&gt;/water confined between two parallel plates under the influence of a magnetic field. The lower plate is assumed to be stretchable and permeable, allowing for suction/injection, while the upper plate induces a squeezing effect by moving toward the lower plate. A system of highly nonlinear partial differential equations (PDEs) governing the velocity and temperature fields is transformed into a system of nonlinear ordinary differential equations (ODEs) using appropriate similarity transformations. These equations are solved by employing a machine learning-based framework that combines physics-informed neural networks with the atomic orbital search (PINNs-AOS) algorithm. The proposed PINNs-AOS model is validated through 50 independent runs, with convergence and accuracy assessed using statistical metrics such as fitness, absolute error, mean squared error (MSE), and Theil’s inequality coefficient (TIC). These metrics consistently fall within optimal ranges: fitness (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;msup&gt;&lt;mrow&gt;&lt;mn&gt;0&lt;/mn&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mo&gt;−&lt;/mo&gt;&lt;mn&gt;5&lt;/mn&gt;&lt;/mrow&gt;&lt;/msup&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt; to &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;msup&gt;&lt;mrow&gt;&lt;mn&gt;0&lt;/mn&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mo&gt;−&lt;/mo&gt;&lt;mn&gt;7&lt;/mn&gt;&lt;/mrow&gt;&lt;/msup&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;), absolute error (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;msup&gt;&lt;mrow&gt;&lt;mn&gt;0&lt;/mn&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mo&gt;−&lt;/mo&gt;&lt;mn&gt;3&lt;/mn&gt;&lt;/mrow&gt;&lt;/msup&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt; to &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;msup&gt;&lt;mrow&gt;&lt;mn&gt;0&lt;/mn&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mo&gt;−&lt;/mo&gt;&lt;mn&gt;5&lt;/mn&gt;&lt;/mrow&gt;&lt;/msup&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;), MSE (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;msup&gt;&lt;mrow&gt;&lt;mn&gt;0&lt;/mn&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mo&gt;−&lt;/mo&gt;&lt;mn&gt;6&lt;/mn&gt;&lt;/mrow&gt;&lt;/msup&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt; to &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;msup&gt;&lt;mrow&gt;&lt;mn&gt;0&lt;/mn&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mo&gt;−&lt;/mo&gt;&lt;mn&gt;8&lt;/mn&gt;&lt;/mrow&gt;&lt;/msup&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;), and TIC (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;msup&gt;&lt;mrow&gt;&lt;mn&gt;0&lt;/mn&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mo&gt;−&lt;/mo&gt;&lt;mn&gt;6&lt;/mn&gt;&lt;/mrow&gt;&lt;/msup&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt; to &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;1&lt;/mn&gt;&lt;msup&gt;&lt;mrow&gt;&lt;mn&gt;0&lt;/mn&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mo&gt;−&lt;/mo&gt;&lt;mn&gt;8&lt;/mn&gt;&lt;/mrow&gt;&lt;/msup&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;), indicating the model’s reliability and precision. Key findings reveal that enhancing the squeezing parameter (&lt;span&gt;&lt;math&gt;&lt;msub&gt;&lt;mrow&gt;&lt;mi&gt;S&lt;/mi&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mi&gt;q&lt;/mi&gt;&lt;/mrow&gt;&lt;/msub&gt;&lt;/math&gt;&lt;/span&gt;) significantly enhances the velocity profile &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;msup&gt;&lt;mrow&gt;&lt;mi&gt;f&lt;/mi&gt;&lt;/mrow&gt;&lt;mrow&gt;&lt;mo&gt;′&lt;/mo&gt;&lt;/mrow&gt;&lt;/msup&gt;&lt;mrow&gt;&lt;mo&gt;(&lt;/mo&gt;&lt;mi&gt;η&lt;/mi&gt;&lt;mo&gt;)&lt;/mo&gt;&lt;/mrow&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;, while raising the suction parameter (&lt;span&gt;&lt;math&gt;&lt;mi&gt;S&lt;/mi&gt;&lt;/math&gt;&lt;/span&gt;) reduces it. An rise in the stretching parameter &lt;span&gt;&lt;math&gt;&lt;mi&gt;λ&lt;/mi&gt;&lt;/math&gt;&lt;/span&gt; boosts fluid velocity near the wall, but causes a decline farther away due to dominant viscous forces. The temperature profile &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mi&gt;θ&lt;/mi&gt;&lt;mrow&gt;&lt;mo&gt;(&lt;/mo&gt;&lt;mi&gt;η","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110475"},"PeriodicalIF":7.0,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144291316","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":"Flow behavior in idealized & realistic upper airway geometries","authors":"Brenda Vara Almirall ,&nbsp;Hadrien Calmet ,&nbsp;Hua Qian Ang ,&nbsp;Kiao Inthavong","doi":"10.1016/j.compbiomed.2025.110449","DOIUrl":"10.1016/j.compbiomed.2025.110449","url":null,"abstract":"<div><div>This study investigated the flow characteristics in idealized and realistic upper airway models during oral inhalation, focusing on their ability to replicate airflow dynamics and turbulence that impact pharmaceutical aerosol delivery. While idealized airway models, such as the United States Pharmacopeia (USP) model, are widely used in regulatory testing, they lack anatomical fidelity, potentially underestimating critical features, including the laryngeal jet formation, which are essential for accurate particle deposition predictions. Understanding the implications of idealized and realistic models is addressed using Large Eddy Simulations (LES) at inhalation rates of 15 and 30 L/min. Four airway models were analyzed: the USP model, the Virginia Commonwealth University (VCU) model, and two realistic models reconstructed from CT scans of healthy adults. The findings revealed the limitations of the USP model, while the VCU models demonstrated laminar flow behavior with a laryngeal jet with a lower magnitude compared to the realistic models. The realistic models (R01 and R02) exhibited more complex flow features, including an earlier laryngeal jet formation, emphasizing the importance in understanding the limitations of idealized models and the variations between realistic models.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110449"},"PeriodicalIF":7.0,"publicationDate":"2025-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144291312","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":"Ensemble mutation and multi-population-driven differential evolution for numerical optimization and feature selection of breast cancer","authors":"Shubham Gupta ,&nbsp;Balkrishna Dwivedi ,&nbsp;Vinay Kumar","doi":"10.1016/j.compbiomed.2025.110495","DOIUrl":"10.1016/j.compbiomed.2025.110495","url":null,"abstract":"<div><div>Breast cancer stands as a critical health challenge for women worldwide, whose numbers continue to increase because of multiple complicating elements. The distribution of high-dimensional medical datasets creates an essential challenge for breast cancer diagnosis because it leads to reduced predictive model efficiency. This research develops an advanced version of the differential evolution (DE), named ensemble mutation and multi-population-driven differential evolution (EMMDE), for optimal feature selection. EMMDE unites multiple novel mechanisms that include population division and ensemble mutation rules, together with a time-varied geometrically diversified scheme to enhance the diversity in the population. The Gaussian binning approach creates a new guiding vector that enters the mutation rule to strike an equilibrium between exploration and exploitation, which promotes fast convergence and diverse population states. The first stage of the EMMDE testing employs standard and complex benchmark functions from the IEEE CEC2017 benchmark set, which demonstrates its outperform search ability. Later, the transfer function-based binary version is developed and validated over 10 benchmark datasets from the UCI repository and 4 datasets of breast cancer. Based on MCE and other performance metrics, it is experimentally verified that the proposed binary EMMDE algorithm achieves highly accurate and promising results for the feature selection problem. Comparison with 12 well-known and recent metaheuristics has demonstrated the impact of proposed strategies in dealing with diverse categories of numerical optimization and feature selection problems.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110495"},"PeriodicalIF":7.0,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144288934","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":"Empirical mode decomposition in clinical signal analysis: A systematic review","authors":"Shanglin Yang ,&nbsp;Hsientsai Wu ,&nbsp;Xuwei Liao ,&nbsp;Yuyang Lin ,&nbsp;Jianjung Chen","doi":"10.1016/j.compbiomed.2025.110566","DOIUrl":"10.1016/j.compbiomed.2025.110566","url":null,"abstract":"<div><div>This systematic review examines the transformative applications of empirical mode decomposition (EMD) in healthcare, focusing on its ability to analyse diverse physiological signals. By a thorough exploration of key databases and stringent study selection, the effectiveness of EMD has been highlighted across various medical fields. In cardiology, EMD has significantly improved electrocardiographic analysis, surpassing conventional techniques such as Fourier and wavelet transforms, with accuracy rates reaching up to 98 % for detecting subtle cardiac abnormalities. In neurology, EMD has enhanced electroencephalographic analysis, better capturing dynamic brain activity and offering higher sensitivity and specificity for the diagnosis of neurological disorders. In respiratory medicine, EMD has demonstrated superior computational efficiency and accuracy in analysing complex respiratory patterns, thereby reducing false-positive rates by 20 %. Despite these advantages, challenges related to intrinsic mode function (IMF) selection and boundary effects introduce performance variability. This review emphasises the need for standardised guidelines and the development of advanced algorithms to address the limitations. Future research should explore hybrid approaches that combine EMD with machine learning models to improve the robustness and efficiency in computation. Overall, this review showcases the potential of EMD to revolutionise physiological signal analysis and provides valuable recommendations for overcoming current challenges, offering insights for further research and clinical practice.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110566"},"PeriodicalIF":7.0,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144279377","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":"ThreeF-Net: Fine-grained feature fusion network for breast ultrasound image segmentation","authors":"Xuesheng Bian ,&nbsp;Jia Liu ,&nbsp;Sen Xu ,&nbsp;Weiquan Liu ,&nbsp;Leyi Mei ,&nbsp;Chaoshen Xiao ,&nbsp;Fan Yang","doi":"10.1016/j.compbiomed.2025.110527","DOIUrl":"10.1016/j.compbiomed.2025.110527","url":null,"abstract":"<div><div>Convolutional Neural Networks (CNNs) have achieved remarkable success in breast ultrasound image segmentation, but they still face several challenges when dealing with breast lesions. Due to the limitations of CNNs in modeling long-range dependencies, they often perform poorly in handling issues such as similar intensity distributions, irregular lesion shapes, and blurry boundaries, leading to low segmentation accuracy. To address these issues, we propose the ThreeF-Net, a fine-grained feature fusion network. This network combines the advantages of CNNs and Transformers, aiming to simultaneously capture local features and model long-range dependencies, thereby improving the accuracy and stability of segmentation tasks. Specifically, we designed a Transformer-assisted Dual Encoder Architecture (TDE), which integrates convolutional modules and self-attention modules to achieve collaborative learning of local and global features. Additionally, we designed a Global Group Feature Extraction (GGFE) module, which effectively fuses the features learned by CNNs and Transformers, enhancing feature representation ability. To further improve model performance, we also introduced a Dynamic Fine-grained Convolution (DFC) module, which significantly improves lesion boundary segmentation accuracy by dynamically adjusting convolution kernels and capturing multi-scale features. Comparative experiments with state-of-the-art segmentation methods on three public breast ultrasound datasets demonstrate that ThreeF-Net outperforms existing methods across multiple key evaluation metrics.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110527"},"PeriodicalIF":7.0,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144289014","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":"FFLUNet: Feature Fused Lightweight UNet for brain tumor segmentation","authors":"Surajit Kundu ,&nbsp;Sandip Dutta ,&nbsp;Jayanta Mukhopadhyay ,&nbsp;Nishant Chakravorty","doi":"10.1016/j.compbiomed.2025.110460","DOIUrl":"10.1016/j.compbiomed.2025.110460","url":null,"abstract":"<div><div>Brain tumors, particularly glioblastoma multiforme, are considered one of the most threatening types of tumors in neuro-oncology. Segmenting brain tumors is a crucial part of medical imaging. It plays a key role in diagnosing conditions, planning treatments, and keeping track of patients’ progress. This paper presents a novel lightweight deep convolutional neural network (CNN) model specifically designed for accurate and efficient brain tumor segmentation from magnetic resonance imaging (MRI) scans. Our model leverages a streamlined architecture that reduces computational complexity while maintaining high segmentation accuracy. We have introduced several novel approaches, including optimized convolutional layers that capture both local and global features with minimal parameters. A layerwise adaptive weighting feature fusion technique is implemented that enhances comprehensive feature representation. By incorporating shifted windowing, the model achieves better generalization across data variations. Dynamic weighting is introduced in skip connections that allows backpropagation to determine the ideal balance between semantic and positional features.</div><div>To evaluate our approach, we conducted experiments on publicly available MRI datasets and compared our model against state-of-the-art segmentation methods. Our lightweight model has an efficient architecture with 1.45 million parameters — 95% fewer than nnUNet (30.78M), 91% fewer than standard UNet (16.21M), and 85% fewer than a lightweight hybrid CNN-transformer network (Liu et al., 2024) (9.9M). Coupled with a 4.9× faster GPU inference time (0.904 ± 0.002 s vs. nnUNet’s 4.416 ± 0.004 s), the design enables real-time deployment on resource-constrained devices while maintaining competitive segmentation accuracy. Code is available at: <span><span>FFLUNet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110460"},"PeriodicalIF":7.0,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144279379","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 review: Lightweight architecture model in deep learning approach for lung disease identification","authors":"Dhiyaa Ageng Maharani,&nbsp;Fitri Utaminingrum,&nbsp;Devi Nazhifa Nur Husnina,&nbsp;Bening Sukmaningrum,&nbsp;Fauziyah Nur Rahmania,&nbsp;Fika Handani,&nbsp;Hidayati Nur Chasanah,&nbsp;Abdurrizqo Arrahman,&nbsp;Faris Febrianto","doi":"10.1016/j.compbiomed.2025.110425","DOIUrl":"10.1016/j.compbiomed.2025.110425","url":null,"abstract":"<div><div>As one of the leading causes of death worldwide, early detection of lung disease is a very important step to improve the effectiveness of treatment. By using medical image data, such as X-ray or CT-scan, classification of lung disease can be done. Deep learning methods have been widely used to recognize complex patterns in medical images, but this approach has the constraints of requiring large data variations and high computing resources. In overcoming these constraints, the lightweight architecture in deep learning can provide a more efficient solution based on the number of parameters and computing time. This method can be applied to devices with low processor specifications on portable devices such as mobile phones. This article presents a comprehensive review of 23 research studies published between 2020 and 2025, focusing on various lightweight architectures and optimization techniques aimed at improving the accuracy of lung disease detection. The results show that these models are able to significantly reduce parameter sizes, resulting in faster computation times while maintaining competitive accuracy compared to traditional deep learning architectures. From the research that has been done, it can be seen that SqueezeNet applied on public COVID-19 datasets is the best basic architecture with high accuracy, and the number of parameters is 570 thousand, which is very low. On the other hand, UNet requires 31.07 million parameters, and SegNet requires 29.45 million parameters trained on CT scan images from Italian Society of Medical and Interventional Radiology and Radiopedia, so it is less efficient. For the combination method, EfficientNetV2 and Extreme Learning Machine (ELM) are able to achieve the highest accuracy of 98.20 % and can significantly reduce parameters. The worst performance is shown by VGG and UNet with a decrease in accuracy from 91.05 % to 87 % and an increase in the number of parameters. It can be concluded that the lightweight architecture can be applied to medical image classification in the diagnosis of lung disease quickly and efficiently on devices with limited specifications.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110425"},"PeriodicalIF":7.0,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144279380","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 multi-phase approach using supervised algorithms and clinical models to generate high-accuracy signatures for pancreatic cancer","authors":"Akash Bararia ,&nbsp;Agniswar Chakraborty ,&nbsp;Gourav Ghosh ,&nbsp;Debabrata Ghosh Dastidar ,&nbsp;Sumit Mukherjee ,&nbsp;Nilabja Sikdar","doi":"10.1016/j.compbiomed.2025.110559","DOIUrl":"10.1016/j.compbiomed.2025.110559","url":null,"abstract":"&lt;div&gt;&lt;h3&gt;Background&lt;/h3&gt;&lt;div&gt;The in silico analyses provide evidence supporting the potential of methylation-driven differentially expressed genes as therapeutic targets across cancer types. This leads us to identify novel targets and their associated drug compounds for further progress towards pancreatic cancer treatment.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Objective&lt;/h3&gt;&lt;div&gt;To identify targeted drugs based on methylation driven genes identified using bulk multi-omics data and single-cell level data to pinpoint important disease markers.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Methods&lt;/h3&gt;&lt;div&gt;The workflow involves screening using the TCGA and ICGC databases, followed by validation with GEO datasets. The study employs supervised learning algorithms like kNN and random forests, and constructs a prediction model using adaptive LASSO-Cox regression. The process also includes pathway analysis, evaluation of survival status, and immune profile deconvolution, as well as multistage evaluation of the methylation driven genes. We conducted drug targeting and molecular dynamic simulations, taking into account genes of interest.Lastly, molecular docking and dynamics simulations were used to find out if the key MEDEGs could be utilized as drug targets.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Results&lt;/h3&gt;&lt;div&gt;&lt;em&gt;CD36, UGT1A1, TFF1, S100P, MUC13, CALHM3&lt;/em&gt; and &lt;em&gt;ANKRD44&lt;/em&gt; were found to be top 7 methylation driven genes. The mutational profile was also documented along with pathway analysis, which showed concordance with our observation based on their significant enriched terms namely “Maintenance of Gastrointestinal Epithelium”, and “Digestive System Homeostasis”. &lt;em&gt;CD36&lt;/em&gt; had prognostic capabilities and was seen to significant in terms of survival and also showed significant immune dysregulation. Our novel findings suggest &lt;em&gt;TFF1, S100P,&lt;/em&gt; and &lt;em&gt;MUC13&lt;/em&gt; were found to be associated with cell type specific expression as seen in single cell data and &lt;em&gt;UGT1A1&lt;/em&gt; was found to be suitable for probable drug targeting. &lt;em&gt;CD36, UGT1A1, TFF1, S100P,&lt;/em&gt; and &lt;em&gt;MUC13&lt;/em&gt; showed concordance when observed at proteomics level and across other datasets. Apigenin-7-O-glucuronide emerged as the top binder for UDP-glucuronosyltransferase 1A1 (also known as UDP 1A1), forming stable complexes with favourable interactions. Catechin and epicatechin were identified as the best ligands for &lt;em&gt;TFF1&lt;/em&gt; and &lt;em&gt;S100P&lt;/em&gt;, while rutin showed high-affinity binding to &lt;em&gt;MUC13&lt;/em&gt;.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Conclusion&lt;/h3&gt;&lt;div&gt;The study successfully identified and validated a panel of biomarkers specific to pancreatic cancer, with potential applications in early diagnosis and treatment. The findings highlight the importance of multi-omics data integration in cancer research and the potential of personalized medicine in improving patient outcomes. The &lt;em&gt;in-silico&lt;/em&gt; drug targeting analysis provides a foundation for the development of novel drugs for PanCa treatment. Hence &lt;em&gt;TFF1, S100P, MUC13&lt;/em&gt;, and &lt;e","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110559"},"PeriodicalIF":7.0,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144288942","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":"Patient-specific investigation into plaque rupture risk due to catheter tracking during TAVR","authors":"David G. Symes ,&nbsp;Laoise M. McNamara ,&nbsp;Claire Conway","doi":"10.1016/j.compbiomed.2025.110558","DOIUrl":"10.1016/j.compbiomed.2025.110558","url":null,"abstract":"<div><div>Catheter delivery during transcatheter aortic valve replacement (TAVR) may result in plaque damage in patients with high plaque burden levels. Primarily, catheter tracking performance is measured through in-vitro methods, such as trackability testing, but these cannot account for the in vivo conditions that are dictated by anatomical variation and tissue properties. This study aims to apply patient-specific finite element (FE) modelling to investigate the potential stresses and contact pressures experienced in the aortic wall and by plaques during catheter tracking for TAVR delivery. This study utilised two patient-specific anatomies, derived from pre-TAVR CT imaging, to develop solid aorta and plaque models. A parameter study of plaque burden (low, moderate and high) and plaque stiffness (soft, intermediate and stiff) revealed a risk of plaque rupture only for one of the patients when they had a high plaque burden of intermediate stiffness. Highly stiff calcified nodules lead to very high plaque tissue stresses (6–9 MPa), but the rupture stress threshold for these plaque types remains unknown. We also observe no difference in catheter reaction forces between patients regardless of burden level or plaque stiffness, except for Patient 2's high plaque burden with stiff plaque variation.</div></div>","PeriodicalId":10578,"journal":{"name":"Computers in biology and medicine","volume":"194 ","pages":"Article 110558"},"PeriodicalIF":7.0,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144288935","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}