A Deep Learning Framework for Classification of Multiple SclerosisBrain Scans: Achievements and Challenges

This study introduces a deep learning-based framework for the classification of MultipleSclerosis (MS) brain scan images, utilizing a hybrid architecture that integratesconvolutional neural networks (CNNs) with advanced feature extraction techniques. Thedataset underwent comprehensive preprocessing, including resizing, normalization, andaugmentation, to enhance model performance and generalizability. The proposed modelachieved an overall accuracy of ۸۸%, with weighted precision, recall, and F۱-scoresexceeding ۸۸%. The results indicate the model's robustness in identifying healthy scansand mild MS cases, though challenges remain in detecting moderate MS due to reducedrecall (۰.۷۲), attributed to class imbalance and subtle feature overlaps. To address theselimitations, future research will focus on expanding datasets, adopting weighted lossfunctions, and integrating multi-modal data for improved sensitivity and interpretability.This framework demonstrates significant potential for clinical applications in MSdiagnosis and highlights avenues for further innovation in medical image analysis.

کلیدواژه ها:

Multiple Sclerosis ، BrainScan Classification ، DeepLearning ، ConvolutionalNeural Networks.