Multiple Sclerosis (MS) is a progressive neurodegenerative disease affecting the Central Nervous System (CNS), leading to demyelination and neurological impairment. Early diagnosis and continuous monitoring of disease progression are crucial for effective treatment. Magnetic Resonance Imaging (MRI) remains the primary tool for detecting MS lesions; however, traditional segmentation methods rely heavily on visual analysis and struggle to detect early-stage lesions. This study reviews the application of Convolutional Neural Networks (CNNs) for automated lesion segmentation in MS. Through an integrative literature review of articles published between 2022 and 2024 from databases such as PubMed, BVS, Nature, Arxiv, and Google Scholar, following PRISMA guidelines, we assessed the effectiveness of AI-based approaches.CNN models such as U-Net and nnU Net demonstrated superior accuracy and sensitivity in segmenting lesions in FLAIR MRI images, outperforming traditional methods. Models like DeepLabV3+ and ResNet also proved effective in differentiating between active and inactive lesions, aiding in distinguishing acute from chronic lesions. Automated segmentation reduced analysis time, minimized false positives, and enhanced reproducibility, mitigating human variability in clinical evaluations. While these advancements offer faster, more accurate diagnoses and better monitoring of disease progression, challenges remain. Chief among them are the need for large-scale labeled datasets and standardization of MRI acquisition protocols. Despite these obstacles, the integration of AI-driven segmentation into clinical practice holds significant promise for improving MS diagnosis, treatment planning, and long-term patient management.
Tavares et al. (Mon,) studied this question.