Intelligence-based medicine最新文献

{"title":"Robust dermatological image analysis via multitask learning on heterogeneous datasets","authors":"Ignazio Gallo,&nbsp;Mattia Gatti,&nbsp;Leonardo Sgroi,&nbsp;Silvia Corchs","doi":"10.1016/j.ibmed.2026.100474","DOIUrl":"10.1016/j.ibmed.2026.100474","url":null,"abstract":"<div><div>Deep learning has substantially advanced dermatological image analysis, but most methods remain tied to single datasets, annotation types, or narrowly defined tasks, limiting their ability to generalize across real clinical variability. This work proposes a unified cross-dataset multitask learning (MTL) framework for skin lesion analysis that jointly integrates classification and segmentation tasks from heterogeneous dermatoscopic and clinical image datasets. The model uses a shared Swin Transformer backbone with task-specific heads, treating each dataset-task pair as a distinct but related objective, thereby allowing datasets with different label structures and annotation types to be trained within the same architecture. To stabilize joint optimization, a dynamic loss orchestration (DLO) mechanism is introduced to adaptively regulate task contributions according to training behavior, reducing task-specific overfitting and encouraging more transferable feature representations. Experiments across multiple public dermatology datasets show that multi-dataset training improves performance across the evaluated benchmarks over single-task configurations. The proposed framework achieves 82.03% accuracy on Derm7pt and 96.95% accuracy on HAM10000 for classification, while reaching IoU scores of 94.03% on PH2 and 92.63% on HAM10000 for segmentation. Qualitative analyses using segmentation outputs, Grad-CAM visualizations, and t-SNE embeddings further indicate that the shared representation captures lesion-focused patterns that are consistent with clinical intuition. These results support cross-dataset MTL as an effective strategy for building versatile and interpretable dermatological image analysis models.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100474"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934461","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From checklist to closed-loop: the case for AI-supported perioperative management in esophagectomy","authors":"Pascal Burri,&nbsp;Dimitrios Chatziisaak,&nbsp;Marcel A. Schneider,&nbsp;Thomas Steffen,&nbsp;Sebastian Wolf,&nbsp;Timon Weiss,&nbsp;Lucien Cron,&nbsp;Christian A. Gutschow","doi":"10.1016/j.ibmed.2026.100471","DOIUrl":"10.1016/j.ibmed.2026.100471","url":null,"abstract":"<div><div>Esophagectomy generates thousands of pre-, intra-, and early postoperative data points per patient, yet these remain siloed, and preoperative risk scores plateau at an area under the curve of about 0.62 for anastomotic leak. Enhanced Recovery After Surgery checklists improve outcomes but are static and open-loop, unable to adapt to a patient's evolving risk. We make the case for a shift toward closed-loop monitoring with AI-supported perioperative management: a sequential, milestone-based system that recalibrates risk at defined timepoints (preoperative assessment, incision/OR start, anastomosis, postoperative day 1 endoscopy, postoperative days 2 to 5) and can indicate which additional data would most improve the next prediction, drawing on value-of-information and active-sensing principles. Preoperative, intraoperative, and early postoperative data are fused, the anastomosis serves as a summed-risk marker of technique and physiology, and a continuously updated risk class drives an escalation-capable treatment pathway. To our knowledge, after searching PubMed and Embase (esophagectomy, anastomotic leak, machine learning, intraoperative, time-series; through June 2026), no published model integrates intraoperative time-series such as hemodynamics and indocyanine-green perfusion dynamics into a leak-prediction model; no AI performs automated postoperative endoscopic assessment of the anastomosis; and multimodal fusion with prospective external validation is largely absent. Explainable models already outperform classical regression (area under the curve 0.84 versus 0.74), and intraoperative time-series integration has improved prediction in adjacent surgical fields but not in esophagectomy. Addressing automation bias, the validation deficit, interoperability, and regulation (including Regulation (EU) 2024/1689 on artificial intelligence), we argue for an explainable, externally validated tool, initially esophagectomy-specific and deployed first as read-only decision support within established frameworks (IDEAL, DECIDE-AI).</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100471"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934526","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Lightweight deep learning models for histopathological image analysis in resource-constrained healthcare environments","authors":"Ahmed Adedeji Adeniyi,&nbsp;Steve Adetunji Adeshina,&nbsp;Yusuf Abass Aleshinloye,&nbsp;Roseline Oluwaseun Ogundokun,&nbsp;Pius Adewale Owolawi","doi":"10.1016/j.ibmed.2026.100475","DOIUrl":"10.1016/j.ibmed.2026.100475","url":null,"abstract":"<div><div>Histopathological images analysis is an essential task in diagnosing cancer, but the application of artificial intelligence (AI) in computational pathology is often constrained by the heavy computational demands of convolutional neural networks (CNNs), Vision Transformers (ViTs), and pathology foundation models due to their computational cost, memory requirements, inference time, and infrastructural needs. This review explores the role of lightweight deep learning models in histopathological image analysis in terms of their diagnostic accuracy, computational efficiency, explainability, and deployability in resource-constrained healthcare environments. Recent advancements in lightweight CNNs, compact ViTs, hybrid CNN-Transformer architectures, efficient attention mechanisms, and optimization algorithms were discussed. The paper also addresses the issue of benchmark histopathology datasets, evaluation procedures, and the necessity of reporting not only clinical performance but also computational metrics like parameters, floating-point operations per second (FLOPs), inference time, memory usage, and the platform for deploying the model. Issues such as insufficient amounts of annotated data, staining, and scanner variability, domain shift between institutions, poor external validation, inadequate explainability, and lack of benchmarking are among the key challenges mentioned. In general, the review indicates that there is an effective way towards scalable, interpretable, and clinically useful computational pathology, particularly in resource-constrained healthcare environments where accuracy must be balanced with efficiency and real-world deployability.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100475"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934459","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Resolving accuracy-fairness trade-offs in clinical decision support: A heterogeneous neural energy-based framework for diabetes management","authors":"Mahmood Ghanbari,&nbsp;Abbas Mirzaei,&nbsp;Behnam Barzegar,&nbsp;Nasser Mikaeilvand,&nbsp;Babak Nouri-Moghaddam","doi":"10.1016/j.ibmed.2026.100460","DOIUrl":"10.1016/j.ibmed.2026.100460","url":null,"abstract":"&lt;div&gt;&lt;h3&gt;Background&lt;/h3&gt;&lt;div&gt;Clinical AI models frequently exhibit degraded performance and systemic bias when applied to highly imbalanced healthcare datasets. Navigating the inherent trade-off between predictive fidelity and demographic equity remains a critical challenge for translational medicine. To address this, we propose the Heterogeneous Neural Energy-Based Classifier (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;HG&lt;/mtext&gt;&lt;mo&gt;‐&lt;/mo&gt;&lt;mtext&gt;NEBC&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;), a novel dual-paradigm framework.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Methods&lt;/h3&gt;&lt;div&gt;The proposed architecture integrates a Dynamic Class Matrix (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;DCM&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;) for adaptive under-sampling with Simulated Annealing (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;SA&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;) for robust hyperparameter optimization across non-convex loss landscapes. Additionally, an Adaptive Fairness Regulation (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;AFR&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;) module is incorporated to explicitly penalize demographic parity deviations during gradient-based optimization, ensuring equitable model convergence.&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Results&lt;/h3&gt;&lt;div&gt;Empirical evaluations on the Diabetes 130-US Hospitals dataset demonstrate that the deep gradient-based configuration (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;HG&lt;/mtext&gt;&lt;mo&gt;‐&lt;/mo&gt;&lt;mtext&gt;NEBC&lt;/mtext&gt;&lt;mo&gt;‐&lt;/mo&gt;&lt;mtext&gt;SAD&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;) achieves highly robust predictive performance in medication prescription, yielding an Area Under the Receiver Operating Characteristic (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;AUROC&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;) of &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;0.9270&lt;/mn&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;, an accuracy of &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;85.32&lt;/mn&gt;&lt;mo&gt;%&lt;/mo&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;, and a Matthews Correlation Coefficient (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;MCC&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;) of &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;0.6881&lt;/mn&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;. Crucially, the AFR module curtails the Predicted Positive Rate (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;PPR&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;) disparity to &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;0.012&lt;/mn&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;, effectively mitigating algorithmic bias. Conversely, for 30-day readmission prediction, the gradient-free heuristic configuration (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;HG&lt;/mtext&gt;&lt;mo&gt;‐&lt;/mo&gt;&lt;mtext&gt;NEBC&lt;/mtext&gt;&lt;mo&gt;‐&lt;/mo&gt;&lt;mtext&gt;FL&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;) prioritizes rapid optimization (completing evaluation in &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mn&gt;121&lt;/mn&gt;&lt;mspace&gt;&lt;/mspace&gt;&lt;mi&gt;s&lt;/mi&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;) and strict demographic equity (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;PPR&lt;/mtext&gt;&lt;mspace&gt;&lt;/mspace&gt;&lt;mtext&gt;Disparity&lt;/mtext&gt;&lt;mo&gt;=&lt;/mo&gt;&lt;mn&gt;0.021&lt;/mn&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;), accepting an anticipated trade-off in global discriminative power (&lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;AUROC&lt;/mtext&gt;&lt;mo&gt;=&lt;/mo&gt;&lt;mn&gt;0.5070&lt;/mn&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;, &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;MCC&lt;/mtext&gt;&lt;mo&gt;=&lt;/mo&gt;&lt;mn&gt;0.0132&lt;/mn&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt;).&lt;/div&gt;&lt;/div&gt;&lt;div&gt;&lt;h3&gt;Conclusions&lt;/h3&gt;&lt;div&gt;The &lt;span&gt;&lt;math&gt;&lt;mrow&gt;&lt;mtext&gt;HG&lt;/mtext&gt;&lt;mo&gt;‐&lt;/mo&gt;&lt;mtext&gt;NEBC&lt;/mtext&gt;&lt;/mrow&gt;&lt;/math&gt;&lt;/span&gt; framework offers a mathematically ri","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100460"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934527","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CoronaryFusion-Net++: A calcium-aware transformer fusion network for coronary calcium removal in cardiac CT angiography","authors":"Sathish K,&nbsp;Arun Sahayadhas","doi":"10.1016/j.ibmed.2026.100468","DOIUrl":"10.1016/j.ibmed.2026.100468","url":null,"abstract":"<div><h3>Background</h3><div>Coronary artery calcium (CAC) causes blooming artifacts in coronary computed tomography angiography (CCTA), leading to overestimation of luminal stenosis and reduced diagnostic reliability in heavily calcified coronary arteries. Existing convolutional neural networks have limited receptive fields, whereas transformer-based models often incur high computational complexity and inadequate local feature preservation.</div></div><div><h3>Objective</h3><div>This study proposes CoronaryFusion-Net++, a calcium-aware hybrid deep learning framework for suppressing coronary calcium blooming artifacts while preserving coronary lumen morphology and anatomical continuity.</div></div><div><h3>Methods</h3><div>CoronaryFusion-Net++ integrates a Dense Encoder Module, a Cross-Scale Fusion Module (CSFM), a Calcium-Aware Swin Transformer Bottleneck, and an Attention-Guided Decoder with Adaptive Multi-Scale Upsampling. The network was trained and evaluated on the AIMI-COCA dataset using patient-level five-fold cross-validation. Performance was assessed using PSNR, SSIM, Dice coefficient, IoU, Precision, Recall, and computational complexity. Ablation experiments quantified the contribution of each architectural component.</div></div><div><h3>Results</h3><div>CoronaryFusion-Net++ achieved a Dice coefficient of 98.83%, an IoU of 97.69%, an SSIM of 0.986, and a PSNR of 39.84 dB, outperforming representative convolutional and transformer-based baseline methods. Furthermore, the framework demonstrated superior anatomical reconstruction and blooming artifact reduction, yielding a localized Lumen Boundary Precision (LBP) of 0.21 mm and an Apparent Stenosis Error (ASE) reduction of 18.42%. Ablation studies demonstrated consistent performance improvements following the integration of dense feature extraction, cross-scale feature fusion, transformer-based contextual modelling, and attention-guided reconstruction while maintaining computational efficiency.</div></div><div><h3>Conclusion</h3><div>CoronaryFusion-Net++ provides an effective framework for calcium blooming suppression and coronary lumen preservation in CCTA images. By combining local feature representation with global contextual modelling, the proposed architecture improves image reconstruction quality while maintaining computational feasibility. These findings indicate its potential to support more reliable stenosis assessment in heavily calcified coronary arteries and provide a foundation for future clinical validation and deployment.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100468"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934455","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"LesionGraph-Net: A clinically motivated Graph Neural Network for accurate skin cancer diagnosis using lesion-focused features","authors":"Morium Akter,&nbsp;Md Rahad Islam Bhuiyan,&nbsp;Md. Zahid Hasan","doi":"10.1016/j.ibmed.2026.100464","DOIUrl":"10.1016/j.ibmed.2026.100464","url":null,"abstract":"<div><div>Early diagnosis of skin cancer is crucial for improving treatment outcomes, increasing survival rates, and enhancing overall patient prognosis. An automated method utilizing clinical markers focused on cancer lesions can significantly improve the cancer diagnosis process. This study aims to leverage significant lesion markers identified by dermatologists using a Graph Neural Network (GNN), LesionGraph-Net, to enhance performance and analyze the complex relationships among biomarkers. The region of interest (ROI) from skin cancer images is segmented using K-means clustering, and biomarkers from skin lesions are extracted using computerized features. Among them, seven key features are identified based on feature selection methods. A graph is constructed from these features using the Pearson correlation formula, with the edge weight threshold determined through experiments at 0.2. LesionGraph-Net is then employed to identify the cancer, and its performance is compared to image-based transfer learning models and feature-based approaches, including GCN, 1D CNN, and various machine learning models. Our proposed model exhibits strong performance, achieving an accuracy of 98.62% and an F1 score of 98.43%. Compared to other methods, the proposed model consistently outperforms both image and feature-based approaches, underscoring the effectiveness of graph-based analysis in diagnosing skin cancer. The study highlights the benefits of integrating automated systems that incorporate significant clinical biomarkers. The retrieved clinical indications align with computerized features, aiding early diagnosis and cancer analysis for dermatologists. The study indicates that GNN-based models outperform conventional methods in identifying cancer, highlighting their potential for clinical trials.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100464"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148951240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"DMA-UNet: A dual-branch multi-scale attention U-Net for MRI segmentation and classification of parotid gland tumors","authors":"Yousef Alswaiti,&nbsp;Ahmed Abed Mohammed,&nbsp;Lamyaa sabeeh Ashour,&nbsp;Ahmed Kateb Jumaah Al-Nussairi,&nbsp;Dhafer G. Honi,&nbsp;Mohd Nadhir Ab Wahab,&nbsp;Laszlo Szathmary","doi":"10.1016/j.ibmed.2026.100454","DOIUrl":"10.1016/j.ibmed.2026.100454","url":null,"abstract":"<div><div>The human salivary glands are classified as crucial components of the exocrine system, which are responsible for generating the saliva required for lubrication, digestion, and oral health, digestion, and maintaining oral health. The tumors of the parotid glands (PGTs) represent a significant clinical concern in head and neck oncology, and both benign and malignant parotid gland tumors occur. The traditional methods of diagnosis, including the use of cytology with fine-needle aspiration (FNAC), are limited, and such limitations are the inconsistent sensitivity and the invasive nature of the method. This paper proposes a novel architecture built upon a multi-scale dual-branch U-Net to balance between the segmentation and classification of parotid gland tumors in parotid gland MRI. The model consists of a single encoder branch to obtain multi-scale features, while two distinct decoder branches facilitate simultaneous tumor segmentation and classification into normal, benign, and malignant classes. The attention mechanisms improve the model’s performance regarding emphasis on the most relevant imaging features. Preprocessing of a diverse dataset of 332 patients, which included different tumor types in MRI, was carried out and augmented to address class imbalance. The Dual Branch U-Net with Multi-Scale Attention was evaluated for segmentation and classification tasks using validation and unseen test datasets. The model demonstrates robust performance, with accuracy, precision, recall, and F1-score values steadily above 0.96 for both validation and testing datasets, while segmentation results showed perfect Dice coefficients of 0.9988 but extremely high IoU values of 0.9984, indicating effective region-level segmentation with some boundary limitations. This indicates that the model generalizes well and can classify tumor samples with high reliability. Category-wise, the model performs exceptionally well in identifying malignant tumors, achieving near-perfect precision and recall.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100454"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934460","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Explainable multi-task ensemble learning for comprehensive stroke classification from multimodal neuroimaging","authors":"Seyyed Mohammad Mousavi,&nbsp;Hamidreza Samzadeh Kermani,&nbsp;Mahdi Mohammadi,&nbsp;Mohammad Mehdi Ghaemi","doi":"10.1016/j.ibmed.2026.100478","DOIUrl":"10.1016/j.ibmed.2026.100478","url":null,"abstract":"<div><h3>Background</h3><div>Stroke is the third most common cause of death and the fourth most common cause of disability in the world, with 11.9 million incident events each year. Although deep learning methods, particularly convolutional neural networks and transfer-learning-based models, have been increasingly applied to stroke classification, existing approaches face several challenges, including limited cross-modality comparison, class imbalance, low interpretability, and suboptimal accuracy when applied to heterogeneous imaging protocols. Moreover, no existing framework has systematically combined these techniques into modality-specific stroke classification models evaluated separately on both CT and MRI data.</div></div><div><h3>Objective</h3><div>To create and test a deep learning architecture that can be used to accurately and interpretably classify stroke from computed tomography and magnetic resonance imaging scans across imaging modalities<strong>.</strong></div></div><div><h3>Methods</h3><div>The proposed framework comprises a five-component ensemble architecture based on transfer learning from the pre-trained convolutional networks on ImageNet, such as Xception, EfficientNetB7 and ConvNeXtSmall, ensemble voting between different architectures, explainable artificial intelligence using Grad-CAM visualization technique, systematic hyperparameter optimization and adaptive data augmentation. The framework was trained and tested on 1005 CT images that were taken at Shafa Hospital and 615 MRI images from a public dataset, including three diagnostic classes: hemorrhagic stroke, ischemic stroke and normal.</div></div><div><h3>Results</h3><div>On the test set of CT, the ensemble model gave an accuracy of 95.12% with a precision of 97%, a recall of 95%, and an F1 score of 96%. The ensemble attained the highest accuracy of 94.12%, precision of 94%, recall of 95% and F1-score of 95% on MRI. CT ensemble performance was better than any of the single architectures and previous best-known methods. Clinically relevant attention patterns were confirmed with Grad-CAM visualizations, which aligned with the radiologically relevant lesion regions.</div></div><div><h3>Conclusion</h3><div>The current study is a system to systematically combine transfer learning, ensemble voting, explainable artificial intelligence, and adaptive augmentation for stroke classification. The framework is clinically interpretable and demonstrated to provide superior performance metrics, thus having the potential to be real-world deployed as a decision-support tool in emergency neuro-imaging workflows. The main contribution of this work is the systematic integration of five complementary components into a modality-specific framework, enabling accurate and clinically interpretable classification for both CT and MRI, with each modality modeled and evaluated independently.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100478"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934457","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"MediLoRA: Clinical-risk-guided low-rank adaptation for parameter-efficient medical question answering","authors":"Seda Bayat Toksoz,&nbsp;Gultekin Isik,&nbsp;Ishak Pacal,&nbsp;Ayse Keles","doi":"10.1016/j.ibmed.2026.100469","DOIUrl":"10.1016/j.ibmed.2026.100469","url":null,"abstract":"<div><div>Medical question-answering systems are commonly adapted to maximize aggregate accuracy even though the consequences of incorrect answers differ across clinical domains. MediLoRA is a physician-informed parameter-efficient fine-tuning framework in which domain criticality conditions a coupled adaptation policy comprising effective rank, target-module coverage, loss weighting, curriculum order, and tier-dependent dropout. Eight physicians rated ten specialties for safety impact, reverse-coded error reversibility, and knowledge complexity. The framework was evaluated with LLaMA-3-8B on 24,000 multiple-choice items, with a 2400-item held-out split containing 2000 MedMCQA questions and 400 clinician-validated synthetic stress-test cases. In the cross-model held-out comparison, MediLoRA achieved 84.5% exact-match accuracy (Wilson 95% CI, 83.0–85.9), compared with 77.5% for uniform LoRA, 64.2% for the frozen backbone, and 85.3% for a full fine-tuning reference. Five independent MediLoRA runs yielded 83.8% mean accuracy (SD, 1.2). For the three critical domains with available domain-level results, gains over uniform LoRA were +18.1 percentage points in pharmacology, +14.3 in emergency medicine, and +9.3 in cardiology. No predefined high-risk benchmark errors occurred for MediLoRA in the 400-case audit; the one-sided exact upper 95% bound was 0.75%. The maximal MediLoRA adapter configuration contained approximately 65.0 million trainable parameters (0.80% of the backbone), versus 6.8 million (0.08%) for uniform LoRA. Because the comparison was not parameter matched and multiple risk-conditioned components changed jointly, the results support the feasibility of the coupled policy rather than an isolated causal effect of rank allocation.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100469"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934525","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"EEG-based biomarker extraction for Alzheimer's disease using principal component analysis for feature optimization and dimensionality reduction","authors":"Kulsoom Fathima,&nbsp;P. Sandhya","doi":"10.1016/j.ibmed.2026.100463","DOIUrl":"10.1016/j.ibmed.2026.100463","url":null,"abstract":"<div><div>Alzheimer's disease (ADs) is a neurodegenerative disorder that progressively destroys brain cells and impairs memory and cognition, That's why the importance of early diagnosis for timely treatment. Even though electroencephalography (EEG) has been gaining importance as a non-invasive and diagnostic tool, many studies in this field focus on binary classification tasks and neglect the use of advanced feature representations obtained from large-scale and heterogeneous datasets. To overcome these shortcomings, this work suggests an EEG-based Machine Learning (ML) scheme that combines state-of-the-art signal pre-processing, Principal Component Analysis (PCA)-based feature selection, and classification performance comparison among Support Vector Machine (SVM), Random Forest (RF), Multi-Layer Perceptron (MLP), and Convolutional Neural Network (CNN) models. The experiments were carried out on the publicly available LEAD EEG-AD dataset that includes data from 813 participants as well as a little over 500,000 pre-processed EEG epochs of AD, Mild Cognitive Impairment (MCI), and Healthy Controls (HC). To reduce feature space, PCA was chosen not only to keep the characteristic neurophysiological features but also to enhance processing speed and the reliability of the classifier. Tests showed that CNN gives the most favourable results with an accuracy rate of 91.8%, a precision rate of 91.0%, a recall rate of 92.5%, an F1-score of 91.7%, and a ROC-AUC of 95.6%, surpassing conventional ML methods. Statistical tests showed a significant difference between the classifiers tested (p &lt; 0.05). These results suggest that well-designed EEG features can be used in detecting cognitive decline and with the help of EEG, one can not only have a scalable, non-invasive tool but also a setup that is clinically relevant for early diagnosis of AD. Subsequent research will revolve around explainable AI, multimodal biomarker fusion, and deployment of the system in clinical practice.</div></div>","PeriodicalId":73399,"journal":{"name":"Intelligence-based medicine","volume":"15 ","pages":"Article 100463"},"PeriodicalIF":0.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934528","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}