Computer methods and programs in biomedicine最新文献

{"title":"A deep learning framework leveraging spatiotemporal feature fusion for electrophysiological source imaging","authors":"Wuxiang Shi ,&nbsp;Yurong Li ,&nbsp;Nan Zheng ,&nbsp;Wenyao Hong ,&nbsp;Zhenhua Zhao ,&nbsp;Wensheng Chen ,&nbsp;Xiaojing Xue ,&nbsp;Ting Chen","doi":"10.1016/j.cmpb.2025.108767","DOIUrl":"10.1016/j.cmpb.2025.108767","url":null,"abstract":"<div><h3>Background and Objectives</h3><div>Electrophysiological source imaging (ESI) is a challenging technique for noninvasively measuring brain activity, which involves solving a highly ill-posed inverse problem. Traditional methods attempt to address this challenge by imposing various priors, but considering the complexity and dynamic nature of the brain activity, these priors may not accurately reflect the true attributes of brain sources. In this study, we propose a novel deep learning-based framework, spatiotemporal source imaging network (SSINet), designed to provide accurate spatiotemporal estimates of brain activity using electroencephalography (EEG).</div></div><div><h3>Methods</h3><div>SSINet integrates a residual network (ResBlock) for spatial feature extraction and a bidirectional LSTM for capturing temporal dynamics, fused through a Transformer module to capture global dependencies. A channel attention mechanism is employed to prioritize active brain regions, improving both the accuracy of the model and its interpretability. Additionally, a weighted loss function is introduced to address the spatial sparsity of the brain activity.</div></div><div><h3>Results</h3><div>We evaluated the performance of SSINet through numerical simulations and found that it outperformed several state-of-the-art ESI methods across various conditions, such as varying numbers of sources, source range, and signal-to-noise ratio levels. Furthermore, SSINet demonstrated robust performance even with electrode position offsets and changes in conductivity. We also validated the model on three real EEG datasets: visual, auditory, and somatosensory stimuli. The results show that the source activity reconstructed by SSINet aligns closely with the established physiological basis of brain function.</div></div><div><h3>Conclusions</h3><div>SSINet provides accurate and stable source imaging results.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"266 ","pages":"Article 108767"},"PeriodicalIF":4.9,"publicationDate":"2025-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834687","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":"Reinforcement learning using neural networks in estimating an optimal dynamic treatment regime in patients with sepsis","authors":"Weijie Liang ,&nbsp;Jinzhu Jia","doi":"10.1016/j.cmpb.2025.108754","DOIUrl":"10.1016/j.cmpb.2025.108754","url":null,"abstract":"<div><h3>Objective:</h3><div>Early fluid resuscitation is crucial in the treatment of sepsis, yet the optimal dosage remains debated. This study aims to determine the optimal multi-stage fluid resuscitation dosage for sepsis patients.</div></div><div><h3>Methods:</h3><div>We propose a reinforcement learning algorithm with neural networks (RL-NN), utilizing the flexibility of deep learning architectures to mitigate model misspecification. We use cross-validation and random search for hyperparameter tuning to further enhance model robustness and generalization.</div></div><div><h3>Results:</h3><div>Simulation results demonstrate that our method outperforms existing methods in terms of both the percentage of correctly classified optimal treatments and the predicted counterfactual mean outcome. Applying this method to the sepsis cohort from the Medical Information Mart for Intensive Care III (MIMIC-III), we recommend that all sepsis patients receive adequate fluid resuscitation (<span><math><mo>≥</mo></math></span> 30 mL/kg) within the first 3 h of admission to the MICU. Our approach is expected to significantly reduce the mean SOFA score by 23.71%, enhancing patient outcomes.</div></div><div><h3>Conclusion:</h3><div>Our RL-NN method offers an accurate, real-time approach to optimizing sepsis treatment and aligns with the ’Surviving Sepsis Campaign’ guidelines. It also has the potential to be integrated with existing electronic health record (EHR) systems, guiding clinical decision-making and thereby improving patient prognosis.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"266 ","pages":"Article 108754"},"PeriodicalIF":4.9,"publicationDate":"2025-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143825341","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":"Impact of the stent footprint on endothelial wall shear stress in patient-specific coronary arteries: A computational analysis from the SHEAR-STENT trial","authors":"Imran Shah ,&nbsp;David Molony ,&nbsp;Adrien Lefieux ,&nbsp;Kaylyn Crawford ,&nbsp;Marina Piccinelli ,&nbsp;Hanyao Sun ,&nbsp;Don Giddens ,&nbsp;Habib Samady ,&nbsp;Alessandro Veneziani","doi":"10.1016/j.cmpb.2025.108762","DOIUrl":"10.1016/j.cmpb.2025.108762","url":null,"abstract":"&lt;div&gt;&lt;h3&gt;Background and Objective:&lt;/h3&gt;&lt;div&gt;Wall shear stress (WSS) has been known to play a critical role in the development of several complications following coronary artery stenting, including in-stent restenosis and thrombosis. Computational fluid dynamics is often used to quantify the post-stenting WSS, which may potentially be used as a predictive metric. However, large-scale studies for WSS-based risk stratification often neglect the footprint of the stent due to reconstruction challenges. The primary objective of this study is to statistically evaluate the impact of the stent footprints (Xience and Resolute stents) on the computed endothelial WSS and quantitatively identify the relationship between these local hemodynamic alterations and the global properties of the vessel, such as curvature, on WSS. The ultimate goal is to evaluate whether and when it is worth including the footprint of the stent in an &lt;em&gt;in-silico&lt;/em&gt; study to compute the WSS reliably.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Methods:&lt;/h3&gt;&lt;div&gt;A previously developed semi-automated reconstruction approach for patient-specific coronaries was employed as a part of the SHEAR-STENT trial. A subset of patients was analyzed (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mi&gt;N&lt;/mi&gt;&lt;mo&gt;=&lt;/mo&gt;&lt;mn&gt;30&lt;/mn&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;), and CFD simulations were performed with and without the stent to evaluate the impact of the stent footprint on WSS. Due to the computationally expensive nature of transient analyses, a sub-cohort of ten patients were used to assess the reliability of WSS obtained from steady computations as a surrogate for the time-averaged results. Global and local vessel curvature data were extracted for all cases and evaluated against stent-induced alterations in the WSS. The differences between the Xience and Resolute stent platforms were also examined to quantify each stent’s unique WSS footprint.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Results:&lt;/h3&gt;&lt;div&gt;Results from the surrogate analysis indicate that steady WSS serves as an excellent approximation of the time-averaged computations. The presence of either stent footprint causes a statistically significant decrease in the space-averaged WSS, and a significant increase in the endothelial regions exposed to very low WSS as well (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mo&gt;&lt;&lt;/mo&gt;&lt;mn&gt;0&lt;/mn&gt;&lt;mo&gt;.&lt;/mo&gt;&lt;mn&gt;5&lt;/mn&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt; Pa). Negative correlations were observed between vessel curvature and WSS differences, indicating that macroscopic vessel characteristics play a more prominent role in determining endothelial WSS at higher curvature values. In our pool of cases, comparison of Xience and Resolute stents revealed that the Resolute platform seems to lead to lower space-averaged WSS and an increase in areas of very low WSS.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Conclusion:&lt;/h3&gt;&lt;div&gt;These results outline (1) the necessity of including the stent footprint for accurate &lt;em&gt;in-silico&lt;/em&gt; WSS analysis; (2) the global features of stented arteries serving as the dominant determinant of WSS past a certain curvature threshol","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"266 ","pages":"Article 108762"},"PeriodicalIF":4.9,"publicationDate":"2025-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834688","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":"Repeatability of radiomic features from brain T1-W MRI after image intensity normalization: Implications for longitudinal studies on structural neurodegeneration","authors":"Noemi Pisani ,&nbsp;Michela Destito ,&nbsp;Carlo Ricciardi ,&nbsp;Maria Teresa Pellecchia ,&nbsp;Mario Cesarelli ,&nbsp;Fabrizio Esposito ,&nbsp;Maria Francesca Spadea ,&nbsp;Francesco Amato","doi":"10.1016/j.cmpb.2025.108738","DOIUrl":"10.1016/j.cmpb.2025.108738","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Radiomics extracts quantitative features from magnetic resonance images (MRI) and is especially useful in the presence of subtle pathological changes within human soft tissues. This scenario, however, may not cover the effects of intrinsic, e.g., aging-related, (physiological) neurodegeneration of normal brain tissue. The aim of the work was to study the repeatability of radiomic features extracted from an apparently normal area in longitudinally acquired T1-weighted MR brain images using three different intensity normalization approaches typically used in radiomics: Z-score, WhiteStripe and Nyul.</div></div><div><h3>Methods:</h3><div>Fifty-nine images of hearing impaired, yet cognitively intact, patients were repeatedly acquired in two different time points within six months. Ninety-one radiomic features were obtained from an area within the pons region, considered to be a healthy brain tissue according to previous analyses and reports. The Intraclass Correlation Coefficient (ICC) and the Concordance Correlation Coefficient (CCC) in the repeatability study were used as metrics.</div></div><div><h3>Results:</h3><div>Features extracted from the MRI normalized with Z-score showed results comparable in both ICC (0.90 (0.82–0.98)) and CCC (0.82 (0.69–0.95)) distribution values, in terms of median and quartiles, with those extracted from the images normalized with WhiteStripe (0.89 (0.84–0.92)) and (0.80 (0.73–0.84)), respectively.</div></div><div><h3>Conclusion:</h3><div>Our findings underline the importance of, providing useful guidelines for, the intensity normalization of brain MRI prior to a longitudinal radiomic analysis.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"265 ","pages":"Article 108738"},"PeriodicalIF":4.9,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143792541","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"How to predict the future face? A 3D methodology to forecast the aspect of patients after orthognathic surgeries","authors":"Elena Carlotta Olivetti ,&nbsp;Federica Marcolin ,&nbsp;Sandro Moos ,&nbsp;Enrico Vezzetti ,&nbsp;Claudia Borbon ,&nbsp;Emanuele Zavattero ,&nbsp;Guglielmo Ramieri","doi":"10.1016/j.cmpb.2025.108757","DOIUrl":"10.1016/j.cmpb.2025.108757","url":null,"abstract":"<div><h3>Background and objective</h3><div>Despite the availability of several commercial solutions for predicting the soft tissue outcomes of maxillofacial surgeries, none have proven sufficiently reliable for routine clinical use. This study proposes a 3D methodology for predicting soft tissue displacement following maxillofacial surgery without relying on mechanical modeling, unlike most existing approaches.</div></div><div><h3>Methods</h3><div>Pre- and post-operative Cone Beam Computed Tomography scans of patients with class III malocclusion were collected. Tailored image processing and volume reconstruction techniques were applied to semi-automatically generate 3D soft tissue models. Cephalometric landmarks were identified to perform a geometrical similarity analysis among patients with the same malocclusion class undergoing the same surgical procedure. Vectorial displacement maps were generated to capture the soft tissue changes from pre- to post-operative and were then applied to the pre-operative of test patients to predict soft tissue outcomes. Euclidean distances were calculated between predicted and real post-operative positions, and the Wilcoxon signed-rank test was conducted to assess statistical differences between predicted and real landmark coordinates.</div></div><div><h3>Results</h3><div>Error maps indicated that approximately 70 % of predicted facial points had errors below 2.5 mm, while around 10 % ranged between 2.5 mm and 3 mm. Statistically significant differences (<em>p</em> &lt; 0.05) were observed only for the gonion and cheilion.</div></div><div><h3>Conclusion</h3><div>. The findings support the validity of the geometrical similarity analysis and the vectorial displacement map approach. The simplicity and promising accuracy of the proposed method encourage further investigations across different surgical procedures. Additionally, integrating this methodology into surgical planning could offer a viable alternative to commercial solutions. This low-cost, computationally efficient prediction method is designed to improve as more patient data become available. The proposed method is patent pending.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"265 ","pages":"Article 108757"},"PeriodicalIF":4.9,"publicationDate":"2025-04-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143824704","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":"Automated determination of hip arthrosis on the Kellgren–Lawrence scale in pelvic digital radiographs scans using machine learning","authors":"Karolina Nurzynska ,&nbsp;Marek Wodzinski ,&nbsp;Adam Piórkowski ,&nbsp;Michał Strzelecki ,&nbsp;Rafał Obuchowicz ,&nbsp;Paweł Kamiński","doi":"10.1016/j.cmpb.2025.108742","DOIUrl":"10.1016/j.cmpb.2025.108742","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Automated analysis of digital radiographs of the pelvis to determine the hip arthrosis state in concordance with the Kellgren–Lawrence scale could facilitate and standardize radiogram descriptions.</div></div><div><h3>Methods:</h3><div>This research evaluates and compares the applicability of the traditional machine-learning approach based on the textural features fed to the classifier and the deep-learning networks of various architectures.</div></div><div><h3>Results:</h3><div>The investigation performed for the binary problem, where the healthy and the most severe state of hip arthrosis were considered, proved that the distinction of these classes is possible for both considered approaches. However, the outcomes recorded for deep-learning methods overcome significantly other approaches, resulting in a correct classification ratio equal to 0.98. When all five classes are considered, the results drop, primarily due to the underrepresentation of such cases.</div></div><div><h3>Conclusions:</h3><div>The influence of data pre-processing was investigated, showing that it is insignificant for deep-learning models and that the statistical dominance analysis approach dominates for traditional models. The evaluation also indicates that the deep-learning models must be trained on the selected region of interest. Otherwise, they lack precision and have problems determining the significant changes depicting the arthrosis.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"266 ","pages":"Article 108742"},"PeriodicalIF":4.9,"publicationDate":"2025-04-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143834689","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":"Corrigendum to “A computational workflow for modeling complex patient-specific coronary stenting cases” [Computer Methods and Programs in Biomedicine Volume 259, February 2025, 108527]","authors":"Luca Antonini ,&nbsp;Gianluca Poletti ,&nbsp;Georgia S. Karanasiou ,&nbsp;Antonis Sakellarios ,&nbsp;Dimitrios I. Fotiadis ,&nbsp;Lorenza Petrini ,&nbsp;Giancarlo Pennati ,&nbsp;Francesca Berti","doi":"10.1016/j.cmpb.2025.108741","DOIUrl":"10.1016/j.cmpb.2025.108741","url":null,"abstract":"","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"265 ","pages":"Article 108741"},"PeriodicalIF":4.9,"publicationDate":"2025-04-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143768584","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A novel soft tissue-integrated kinematic solver for skeletal motion: Validation and applications","authors":"K. Duquesne ,&nbsp;A. Van Oevelen ,&nbsp;J. Sijbers ,&nbsp;W. Van Paepegem ,&nbsp;E. Audenaert","doi":"10.1016/j.cmpb.2025.108766","DOIUrl":"10.1016/j.cmpb.2025.108766","url":null,"abstract":"<div><h3>Background and Objective</h3><div>Kinematic solvers for human motion analysis, relying on oversimplified joint definitions, face inherent limitations in capturing the true spectrum of skeletal motion. Recent advancements incorporated soft tissue constraints to derive more realistic joint kinematics. However, these methods require marker data input and are computationally expensive, limiting their application to specific joints. This paper proposes a novel kinematic solver that addresses this gap by explicitly accounting for soft tissues, while allowing for accurate and computational efficient modeling across diverse movements and joints.</div></div><div><h3>Methods</h3><div>The proposed soft tissue-integrated kinematic solver determines the kinematics by relying on the principle of force balance. In a cascaded iterative way, the position and orientation of each individual segment is updated by minimizing the force residual acting on the segment The latter is solved through a unique way by defining and aligning two point clouds. Accuracy was assessed with three datasets: in-vivo MRI squats (<em>N</em> = 9), in-vitro cadaver CT squat (<em>N</em> = 1), and in-vitro cadaver arm flexion/extension/pro-supination (<em>N</em> = 1). The accuracy was assessed by computing the absolute error on the joint angles and translations and benchmarked against traditional inverse kinematics with a revolute joint as well as two computer vision techniques (OSSO and SKEL).</div></div><div><h3>Results</h3><div>All experiments showed that with sufficient input data (over 5 rigid bone markers, or skin zones), the primary motion error was almost without exception under 1.5° This outperformed the inverse kinematics with revolute joint (7.29° flexion-extension), OSSO (9.59° flexion-extension) and SKEL (3.19° flexion-extension) methods. The median error on the secondary kinematics for the humeroulnar and ulnoradial joints were below 3.78° and 2.50 mm when driving the motion with skin zones. For the tibiofemoral joints, errors were under 5.39° and 3.5 mm. Computation time was below 30 s per frame.</div></div><div><h3>Conclusions</h3><div>The kinematic solver enables exploring all degrees of freedom accurately without compromising computational efficiency. Unlike biomechanical methods which are limited to marker data, the kinematic solver can analyze both marker and skin data.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"265 ","pages":"Article 108766"},"PeriodicalIF":4.9,"publicationDate":"2025-04-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143816016","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":"MSigSeg: An R package for multiple signals segmentation","authors":"Xuanyu Liu ,&nbsp;Junbo Duan ,&nbsp;Dian Gong","doi":"10.1016/j.cmpb.2025.108744","DOIUrl":"10.1016/j.cmpb.2025.108744","url":null,"abstract":"<div><h3>Background and Objective:</h3><div>Identifying breakpoints in signals is crucial for uncovering important features in scientific data. In the biomedical field, the heterogeneity of signals leads to increased complexity in identifying breakpoints. While existing methods and software packages most focus on detecting breakpoints in individual signals, a significant challenge in this field is to detect common breakpoints of multiple signals. To address this challenge, a fast and optimal method has been developed and implemented in the R package <strong>MSigSeg</strong> as a practical tool.</div></div><div><h3>Methods:</h3><div>The proposed method utilizes an optimization approach with <span><math><mi>ℓ</mi></math></span>-0 norm penalty to efficiently and accurately detect the locations of common breakpoints in multiple signals. This article provides a detailed description of the mathematical problem, the fast optimization algorithm which is implemented in the package, and the usage of core functions along with example datasets.</div></div><div><h3>Results:</h3><div>To evaluate the performance of the proposed method, a simulation study is conducted, comparing it with other segmentation approaches. Real-world problems such as are also processed to demonstrate the practical value of the package. Substantial efficiency gain can be observed by our results.</div></div><div><h3>Conclusions:</h3><div>Our R package <strong>MSigSeg</strong> implements an efficient and sensitive method for detecting common breakpoints across multiple signals, serving as a valuable resource for the analysis of intricate biomedical signals. The proposed package is available on the Comprehensive R Archive Network (CRAN) repository <span><span>https://CRAN.R-project.org/package=MSigSeg</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"265 ","pages":"Article 108744"},"PeriodicalIF":4.9,"publicationDate":"2025-04-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143785829","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":"Enhancing radiomics features via a large language model for classifying benign and malignant breast tumors in mammography","authors":"Sinyoung Ra ,&nbsp;Jonghun Kim ,&nbsp;Inye Na ,&nbsp;Eun Sook Ko ,&nbsp;Hyunjin Park","doi":"10.1016/j.cmpb.2025.108765","DOIUrl":"10.1016/j.cmpb.2025.108765","url":null,"abstract":"<div><h3>Background and Objectives</h3><div>Radiomics is widely used to assist in clinical decision-making, disease diagnosis, and treatment planning for various target organs, including the breast. Recent advances in large language models (LLMs) have helped enhance radiomics analysis.</div></div><div><h3>Materials and Methods</h3><div>Herein, we sought to improve radiomics analysis by incorporating LLM-learned clinical knowledge, to classify benign and malignant tumors in breast mammography. We extracted radiomics features from the mammograms based on the region of interest and retained the features related to the target task. Using prompt engineering, we devised an input sequence that reflected the selected features and the target task. The input sequence was fed to the chosen LLM (LLaMA variant), which was fine-tuned using low-rank adaptation to enhance radiomics features. This was then evaluated on two mammogram datasets (VinDr-Mammo and INbreast) against conventional baselines.</div></div><div><h3>Results</h3><div>The enhanced radiomics-based method performed better than baselines using conventional radiomics features tested on two mammogram datasets, achieving accuracies of 0.671 for the VinDr-Mammo dataset and 0.839 for the INbreast dataset. Conventional radiomics models require retraining from scratch for an unseen dataset using a new set of features. In contrast, the model developed in this study effectively reused the common features between the training and unseen datasets by explicitly linking feature names with feature values, leading to extensible learning across datasets. Our method performed better than the baseline method in this retraining setting using an unseen dataset.</div></div><div><h3>Conclusions</h3><div>Our method, one of the first to incorporate LLM into radiomics, has the potential to improve radiomics analysis.</div></div>","PeriodicalId":10624,"journal":{"name":"Computer methods and programs in biomedicine","volume":"265 ","pages":"Article 108765"},"PeriodicalIF":4.9,"publicationDate":"2025-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143792603","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}