AI and Neurological Rare & Neuroimmune Diseases ── Multiple sclerosis ── 2026-10-11
Multiple sclerosis: treatment, trials and AI
- Abstract
Multiple sclerosis (MS) is a chronic neurological disorder characterized by demyelinating lesions in the central nervous system that disrupt neural impulse transmission and alter motor, sensory, and visual functions. Early diagnosis and accurate segmentation of MS lesions are critical for effective treatment and preventing disease progression. This study introduces MedSegNet-AXU, a novel framework for MS lesion segmentation. By integrating a U-Net backbone with the Convolutional Block Attention Module (CBAM) and Extended Atrous Spatial Pyramid Pooling (ASPP), the proposed model effectively uses both channel-wise and spatial attention mechanisms to enhance segmentation accuracy. The model was evaluated in the Brain Magnetic Resonance Dataset of Multiple Sclerosis (BMDMS), achieving a Dice score of 98.58% and a specificity of 98.71%. We further evaluated MedSegNet-AXU in the Brain Tumor Segmentation (BRATS) 2019, 2020, and 2021 datasets, achieving mean Dice scores of 92–96% and Jaccard indices of 85–91% in modalities and tumor regions. We extracted 21 radiomic features for the sensory, motor, and visual systems from the segmented MS lesions and integrated them with clinical metadata. Using Chi-square feature selection, the top 20 features were classified with traditional machine learning and graph-based models, where GraphSAGE significantly outperformed others, achieving test accuracies of 86.21%, 96.55%, and 75.86% for sensory, motor, and visual systems. We validated the reliability of the GraphSAGE model using Threshold Analysis with Anomaly Edges and Correlation Range Clustering. By combining MRI imaging and clinical data, our approach offers a comprehensive tool for MS lesion analysis, aiding precise diagnosis and personalized treatment strategies.
Journal IF-equivalent: 5.5 (OpenAlex 2-year mean citedness, value as of 2026-10-10, retrieved 2026-10-11; not the official Clarivate IF)Reference: Karim W, Zaman SB, Sutradhar D, Debnath RK, Azam S, Yeo KC, et al. MedSegNet-AXU: advanced segmentation and graph-based classification of multiple sclerosis lesions from 3D magnetic resonance imaging via radiomics and clinical data fusion. Complex Intell Syst. 2026 Oct 8 [Epub ahead of print]. doi:10.1007/s40747-026-02535-6.Checked: Abstract only