Intelligence-based medicine最新文献

{"title":"Cardiac Magnetic Resonance-to-Computed Tomography Angiography image conversion using diffusion models for Transcatheter Aortic Valve Implantation planning","authors":"Carmen Guadalupe Colin-Tenorio ,&nbsp;Agnes Mayr ,&nbsp;Christian Kremser ,&nbsp;Markus Haltmeier ,&nbsp;Enrique Almar-Munoz","doi":"10.1016/j.ibmed.2025.100335","DOIUrl":"10.1016/j.ibmed.2025.100335","url":null,"abstract":"<div><h3>Introduction:</h3><div>Transcatheter Aortic Valve Implantation (TAVI) has become the preferred method for treating severe aortic stenosis, especially in patients who are unsuitable for traditional surgery. Typically, preoperative imaging for TAVI involves contrast-enhanced Computed Tomography Angiography (CTA). However, for patients with contraindications to contrast agents, Cardiac Magnetic Resonance imaging (CMR) is a viable alternative, albeit with its limitations in visualizing calcifications.</div></div><div><h3>Methods:</h3><div>This study explores the application of diffusion models to enhance CMR-to-CTA contrast-free image conversion, to avoid the use of contrast agents and ionizing radiation. We developed a pipeline incorporating Denoising Diffusion Probabilistic Models (DDPMs) and Stochastic Differential Equations (SDE) models to synthesize CTA-equivalent images from CMR scans. We evaluated this approach using an in-house dataset consisting of 39 paired CTA and CMR scans. For the training process, coregistration of both modalities was required, which we achieved by performing rigid registration using the segmented aorta masks.</div></div><div><h3>Results:</h3><div>Our results show that the synthesized CTA images maintain high fidelity to the actual scans. This is quantitatively supported by a mean Structural Similarity Index Measure (SSIM) of 0.8 and a Peak Signal-to-Noise Ratio (PSNR) of 22<!--> <!-->dB using conditional Stochastic Differential Equations (SDE) and Prediction-Correction (PC), indicating strong structural preservation and low reconstruction error. However, the model occasionally fails to accurately detect valve calcifications, likely due to limitations in capturing subtle pathological details that are not visually discernible in CMR images.</div></div><div><h3>Conclusion:</h3><div>Diffusion models used to synthesize CTA images from CMR datasets achieve high accuracy, providing a contrast-free alternative for TAVI planning and potential insights into valvular calcification patterns. However, accurate visualization of valve calcification occasionally fails, and larger datasets are desirable for validation.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100335"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926926","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Advancing glaucoma diagnosis: Multi-modal deep learning with vision transformer architectures","authors":"Vaibhav C. Gandhi ,&nbsp;Priyesh P. Gandhi ,&nbsp;Ali k. Abdul Raheem ,&nbsp;Yasser Taha Alzubaidi ,&nbsp;Kabul Khudaybergenov ,&nbsp;Mohammad Khishe","doi":"10.1016/j.ibmed.2026.100355","DOIUrl":"10.1016/j.ibmed.2026.100355","url":null,"abstract":"<div><div>One of the significant causes of irreversible blindness is glaucoma which develops without symptoms and is only revealed in time when the case is very severe. The existing diagnostic models that use single-modes imaging and convolutional neural networks (CNNs) have limitations of local features dependence, less interpretability, and a lack of accuracy. This paper suggests a multi-modal deep learning model combining retinal fundus images and optical coherence tomography (OCT) scans with Vision Transformer (ViT) networks that could improve the detection and progression analysis of glaucoma. In this work, multimodal refers to an architectural and representational fusing through a hybrid Vision Transformer design and not a concurrent multi-sensor data acquisition. The structural and contextual information across modalities provide the framework with the ability to capture subtle pathological changes than CNN baselines. The benchmark experiments prove that the suggested model achieves both an accuracy of 94.5 percent and AUC-ROC of 91.7 percent, being better than VGG16, ResNet-50, and InceptionV3. Such findings highlight the promise of transformer-based multi-modal solutions to enhance early detection of glaucoma and assist with more feasible and interpretable clinical judgment.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100355"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188805","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Advanced AI framework for accurate detection and classification of brain tumours from MRI images","authors":"M. Rajesh ,&nbsp;K. Swaminathan ,&nbsp;K. Vengatesan ,&nbsp;Usha Moorthy ,&nbsp;Sathishkumar Veerappampalayam Easwaramoorthy","doi":"10.1016/j.ibmed.2026.100348","DOIUrl":"10.1016/j.ibmed.2026.100348","url":null,"abstract":"<div><div>Brain tumours adversely affect patient outcomes owing to their intricacy and the difficulties associated with diagnosis. The accuracy and timeliness of diagnosis are hindered by the subjectivity and unpredictability inherent in manual magnetic resonance imaging (MRI) interpretation. We present novel research on artificial intelligence systems capable of detecting, segmenting, and categorising brain cancers utilising MRI data, which may assist in addressing these issues. The system utilises advanced convolutional neural network (CNN) designs and unique explainability methods; it is designed for application in therapeutic and evidential contexts. This approach addresses deficiencies in cancer categorisation, differentiation, and AI interpretability, hence enhancing the accuracy and reliability of diagnosis. The method's efficacy and practical utility were evidenced through validation on extensive MRI datasets encompassing gliomas, meningiomas, pituitary tumours, and healthy controls. An AI-driven diagnostic tool can increase clinical decision-making, decrease diagnostic error rates, expedite therapy initiation, and improve patient outcomes.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100348"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037647","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Quantum computing in medical diagnostics and treatment: A systematic review of trends, challenges and future directions","authors":"Najnin Sultana Shirin ,&nbsp;Md Mehedi Hasan ,&nbsp;Md Kishor Morol ,&nbsp;Nafiz Fahad ,&nbsp;Md Tanzib Hosain ,&nbsp;Md Jakir Hossen ,&nbsp;Dip Nandi","doi":"10.1016/j.ibmed.2026.100356","DOIUrl":"10.1016/j.ibmed.2026.100356","url":null,"abstract":"<div><div>The growing digitalization of medical diagnostics has resulted in massive amounts of complicated data, including medical imaging and genomic sequences, clinical writing and real-time patient monitoring. Classical machine learning (CML) has achieved amazing success in evaluating such data. But its computational restrictions hinder scalability and efficiency when dealing with high-dimensional biomedical problems. Quantum machine learning (QML) combines the principles of quantum computing (QC) with advanced learning algorithms to offer a transformative paradigm for digital healthcare. This paper provides a systematic overview of QML foundations including quantum data encoding (QDC), variational quantum circuits (VQC), kernel methods, and hybrid quantum classical models. This paper also focuses on their applications in medical imaging, genomics, natural language processing (NLP) for electronic health records, drug discovery and healthcare security. We present comparative insights between classical and quantum approaches such as slow processing of high-dimensional data, limited scalability and inefficiency in complex optimization problems. This review also emphasizes the emerging directions towards quantum-based personalized digital healthcare approaches. By combining medical science with quantum QML has the potential to revolutionize the future of precision diagnostics, treatment optimization and healthcare data security. This study also provides a valuable resource for those interested in quantum computing and researchers who want to stay updated on the fast-growing area.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100356"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147396417","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hybrid spatiotemporal feature fusion for robust lesion detection and tracking in breast ultrasound video data","authors":"Radwan Qasrawi ,&nbsp;Suliman Thwib ,&nbsp;Ghada Issa ,&nbsp;Razan AbuGhoush ,&nbsp;Hussein AlMasri ,&nbsp;Marah Qawasmi ,&nbsp;Nael Abu Halaweh","doi":"10.1016/j.ibmed.2025.100330","DOIUrl":"10.1016/j.ibmed.2025.100330","url":null,"abstract":"<div><h3>Background</h3><div>Speckle noise, tissue deformation, low contrast, and frame inconsistencies limit the reliability of traditional breast lesion tracking approaches in ultrasound videos.</div></div><div><h3>Objective</h3><div>This study aims to develop a robust hybrid framework that integrates advanced image enhancement, deep learning-based detection, and spatiotemporal feature fusion for improved lesion detection and tracking in breast ultrasound video sequences.</div></div><div><h3>Methods</h3><div>We propose a two-phase computational framework. The first phase employs Contrast-Limited Adaptive Histogram Equalization (CLAHE) for local contrast enhancement, followed by a hybrid denoising strategy combining anisotropic diffusion and unsharp masking to suppress noise and preserve edge sharpness. In the second phase, lesion detection is performed using a YOLOv11-L model, fine-tuned on a curated dataset of annotated breast ultrasound images. For tracking, we utilize Kernelized Correlation Filtering (KCF) enhanced with a Hybrid Spatiotemporal Context (STC) representation. The system is evaluated on a dataset comprising 11,382 ultrasound images and 40 video sequences, with performance assessed using Intersection over Union (IoU), success rate, failure rate, and processing speed.</div></div><div><h3>Results</h3><div>The proposed framework achieved an IoU of 0.878 for benign lesions and 0.881 for malignant lesions. The integration of STC features and YOLO detection reduced tracking failure rates by over 25 % and improved success rates to 99.0 % for benign and 99.4 % for malignant lesions. The system processed 41–45 frames per second in real time.</div></div><div><h3>Conclusions</h3><div>Our framework provides an effective solution for real-time lesion detection and tracking in breast ultrasound videos. By enhancing both accuracy and reliability, it supports improved clinical decision-making in breast cancer diagnostics.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100330"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145791919","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The second-generation 3D-Printed localization grid for MRI-guided interventional procedures","authors":"Queenie T.K. Shea ,&nbsp;Wing Ki Wong ,&nbsp;Louis Lee","doi":"10.1016/j.ibmed.2026.100343","DOIUrl":"10.1016/j.ibmed.2026.100343","url":null,"abstract":"<div><div>Patient-specific three-dimensional (3D)-printed magnetic resonance imaging localization grids (MR-Grids) have demonstrated feasibility as an assistive tool for image-guided interventional procedures. However, a limitation of conventional designs arises when the target lesion lies in an imaging plane where the discrete grid markers are not visible. To address this challenge, we proposed a patient-specific MR-Grid incorporating a contrast-filled flexible tubing system arranged in a criss-cross pattern, ensuring visibility across all imaging slices.</div><div>The proposed MR-Grid comprises two primary components: (1) a 3D-printed patient-specific scaffold designed to conform to individual anatomical contours, and (2) a contrast-filled flexible tubing system inserted into the grooves of the scaffold.</div><div>The MR-Grid was tested in an interventional procedure using a biopsy phantom containing MRI-visible lesions to validate its utility. The grid facilitated precise needle insertion by identifying the optimal entry point under MR guidance, demonstrating its potential to improve accuracy and efficiency in image-guided interventions.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100343"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926931","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The potential of artificial intelligence in advancing neuroscience: A systematic review of current applications and models","authors":"Fatemeh Afroughi ,&nbsp;SeyedAhmad SeyedAlinaghi ,&nbsp;Pegah Mirzapour ,&nbsp;Shabnam Shirdel ,&nbsp;Zohal Parmoon ,&nbsp;Mohammad Musa Khorshidi ,&nbsp;Somaye Mansouri ,&nbsp;Mahdi Sheykhi ,&nbsp;Yusuf Popoola ,&nbsp;Esmaeil Mehraeen","doi":"10.1016/j.ibmed.2025.100338","DOIUrl":"10.1016/j.ibmed.2025.100338","url":null,"abstract":"<div><h3>Introduction</h3><div>Artificial intelligence (AI) is the simulation of human intelligence, in which machines perform problem-solving like the human brain. AI and neuroscience are interrelated. In this study, a systematic review of current AI models and applications was conducted to consider the potential of AI in advancing neuroscience.</div></div><div><h3>Methods</h3><div>Relevant articles were selected based on a search in three reputable databases, including Web of Science, PubMed, and Scopus. Two independent researchers conducted the selection process in two stages.</div></div><div><h3>Results</h3><div>A total of 99 studies (2019–2024) met PRISMA criteria. Of these, 83 studies focused on specific brain disorders—most notably Alzheimer's disease (n = 26), stroke (n = 14), epilepsy (n = 7), and Parkinson's disease (n = 7)—while 22 addressed broader neuroscience applications. A range of AI methods were applied, including traditional machine learning techniques (e.g., Support Vector Machines (SVM), Random Forest) and deep learning approaches (e.g., Convolutional Neural Networks (CNNs), Generative Adversarial Networks (GANs)), with several studies employing hybrid models. A comparative analysis of study designs revealed a heavy reliance on public datasets (e.g., Alzheimers Disease Neuroimaging Initiative (ADNI)) for Alzheimer's research, while studies on other disorders predominantly utilized private cohorts. Regarding validation, the majority of studies employed internal cross-validation strategies, with fewer utilizing independent external datasets to test generalizability.</div></div><div><h3>Conclusion</h3><div>The transformative potential of AI in advancing neuroscience lies in its ability to increase diagnostic accuracy, predict disease progression, and enhance imaging techniques. Future research should focus on refining AI methods to enhance generalizability and foster collaborations between AI practitioners and neuroscientists.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100338"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926927","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Emerging anomaly detection techniques for electronic health records: A survey","authors":"Soumendra N. Bhanja ,&nbsp;Haoran Niu ,&nbsp;Yang Chen ,&nbsp;Olufemi A. Omitaomu ,&nbsp;Angela Laurio ,&nbsp;Amber Trickey ,&nbsp;Vijayalakshmi Sampath ,&nbsp;Jonathan R. Nebeker","doi":"10.1016/j.ibmed.2026.100349","DOIUrl":"10.1016/j.ibmed.2026.100349","url":null,"abstract":"<div><h3>Background</h3><div>Anomaly detection in electronic health records (EHRs) is a cornerstone of biomedical informatics, with direct implications for patient safety, clinical decision-making, and the prevention of healthcare fraud. Once guided primarily by simple rule-based methods, the field has advanced rapidly, driven by increased computing power, richer and more detailed health data, and the rise of machine learning and deep learning techniques. The objective of this paper is to provide a comprehensive overview of modern approaches to detecting anomalies in EHRs, outlining their strengths, limitations, and relevance to key healthcare challenges. We review traditional statistical methods alongside newer ML- and DL-based strategies and hybrid models, with particular attention to how these techniques support transparency and build clinical trust.</div></div><div><h3>Methods</h3><div>This paper presents a thorough and critical survey through systematic review (PRISMA-based) of the latest anomaly detection strategies in time-sequence data domains within electronic health record systems.</div></div><div><h3>Results</h3><div>We explore a broad spectrum of methodologies, including statistical models, supervised and unsupervised learning approaches, hybrid frameworks, and state-of-the-art ML-based techniques that collectively advance the precision and scalability of detecting anomalies in complex clinical datasets. In addition to mapping current capabilities, we address the enduring challenges that hinder widespread implementation and provide a forward-looking perspective on the future of anomaly detection in the data-rich landscape of modern healthcare.</div></div><div><h3>Summary</h3><div>The advancement in AI-based approaches is reported along with the basic principles of the individual approaches and their applicability. The increased availability of high-quality data, advancements in DL approaches, and enhanced computation power are leading to more frequent adaptation of DL-based approaches. Emerging DL-based approaches that have been adapted in other domains or recently applied in the EHR domain are also discussed in detail. Although DL-based approaches can improve model predictions by incorporating comorbidities, their application is limited in low-frequency data domains (e.g., when the total available data remains in the single digits). Therefore, the user must carefully consider the application based on data availability.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100349"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077477","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Dual-stage deep learning framework for brain tumor classification and localization using multimodal MRI scans","authors":"Deependra Rastogi ,&nbsp;Prashant Johri ,&nbsp;Sumit Singh Dhanda ,&nbsp;Anand Singh ,&nbsp;Suman Avdhesh Yadav ,&nbsp;Arfat Ahmad Khan ,&nbsp;Seifedine Kadry","doi":"10.1016/j.ibmed.2026.100361","DOIUrl":"10.1016/j.ibmed.2026.100361","url":null,"abstract":"&lt;div&gt;&lt;h3&gt;Background&lt;/h3&gt;&lt;div&gt;Brain tumor classification and localization are essential for accurate diagnosis and effective treatment planning. With the increasing use of MRI in clinical workflows, automated tools have become crucial for assisting radiologists in providing fast and reliable analysis. Traditional approaches for tumor classification and segmentation are often time-consuming and subjective, underscoring the need for deep learning-based frameworks that can enhance diagnostic accuracy and efficiency.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Method&lt;/h3&gt;&lt;div&gt;This study used the TCGA-GBM dataset from The Cancer Imaging Archive (TCIA) to assess a dual-stage deep learning framework combining a classification model with a RESUNET for segmentation. TCGA-GBM, part of the broader TCGA project, focuses on Glioblastoma Multiforme (GBM), a highly aggressive brain cancer. The dataset includes 3929 images, with 2556 non-tumor (class 0) and 1373 tumor (class 1) samples. The framework incorporated three convolutional neural network (CNN) architectures—MobileNet, NASNetMobile, and ResNet101—each enhanced with a Transfer Learning Layer for classification, followed by a RESUNET network for tumor localization. Transfer learning enabled the use of pre-trained weights, improving convergence speed and generalization. MobileNet offered a lightweight, efficient solution; NASNetMobile provided a strong balance between accuracy and computational cost; and ResNet101 delivered deeper feature extraction for higher precision. The RESUNET architecture, combining U-Net and residual learning, accurately segmented tumor regions, enabling effective integration of classification and localization within a unified framework.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Results&lt;/h3&gt;&lt;div&gt;In the classification stage, the models achieved average accuracies of 0.9600 for MobileNet + Transfer Learning Layer, 0.9700 for NASNetMobile + Transfer Learning Layer, and 0.9500 for ResNet101 + Transfer Learning Layer, demonstrating their effectiveness in categorizing brain tumors. Performance was further evaluated using precision, recall, and F1 scores for both classes. MobileNet + Transfer Learning Layer achieved 0.95 precision, 0.99 recall, and 0.97 F1 for class 0, and 0.98, 0.92, and 0.95 for class 1. NASNetMobile + Transfer Learning Layer achieved 0.96, 0.99, and 0.98 for class 0, and 0.99, 0.93, and 0.96 for class 1. ResNet101 + Transfer Learning Layer achieved 0.97, 0.96, and 0.96 for class 0, and 0.93, 0.94, and 0.94 for class 1. For tumor localization, the RESUNET segmentation network accurately delineated tumor regions across all classification models.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Conclusions&lt;/h3&gt;&lt;div&gt;The proposed dual-stage deep learning framework effectively automates both classification and localization of brain tumors from MRI scans. The results demonstrate strong performance across all architectures, with NASNetMobile + RESUNET achieving the most balanced combination of precision and efficiency. Compared to baseline CNNs and pr","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100361"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147396425","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"MoE-Net: A deep ensemble framework optimized by genetic algorithm for nasal bone fracture detection on lateral X-ray images","authors":"Mobin Mehrpour ,&nbsp;Seyed Abolghasem Mirroshandel ,&nbsp;Tahereh Mortezaei ,&nbsp;Zahra Dalili Kajan","doi":"10.1016/j.ibmed.2026.100359","DOIUrl":"10.1016/j.ibmed.2026.100359","url":null,"abstract":"<div><div>Reliable identification of nasal bone fractures on lateral radiographs is a persistent clinical challenge, largely due to the subtle radiographic presentation of fractures, variability and noise inherent in routine imaging, and the limited availability of expert-annotated datasets. To overcome these obstacles, we present MoE-Net, a precision-oriented Mixture-of-Experts Network designed to enhance automated nasal bone fracture detection. The proposed approach systematically evaluates a pool of ten pre-trained convolutional neural network (CNN) and transformer-based architectures, from which the three most accurate and error-resilient models, InceptionResNetV2, DenseNet121, and Xception, are selected as specialized experts. Each expert model undergoes targeted fine-tuning, while a genetic algorithm optimizes their contribution weights within the ensemble to maximize predictive performance. MoE-Net demonstrates substantial performance gains over individual models and existing handcrafted-feature-based methods. On the test dataset, the framework achieves an accuracy of 91.70%, sensitivity of 91.70%, specificity of 89.95%, F2-score of 91.69%, Matthews Correlation Coefficient of 83.48%, and an area under the ROC curve of 91.99%. This performance reflects a clinically meaningful balance between minimizing false negatives, critical for preventing missed diagnoses, and controlling false positives to avoid unnecessary interventions. These findings support the clinical applicability of MoE-Net as a robust, high-performance decision-support tool for nasal bone fracture detection. The study highlights the advantages of precision-driven ensemble strategies in medical imaging and underscores their potential to improve diagnostic accuracy and contribute to more efficient patient care.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"13 ","pages":"Article 100359"},"PeriodicalIF":0.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146188803","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}