Feature Extraction through MS Lesion and Cerebral Lobe Segmentation for Enhanced EDSS Prediction

In this paper, we present a hybrid approach for segmenting multiple sclerosis (MS) lesions and predicting disability progression. Using Concat-UNet for accurate MS lesion segmentation and a multi-class UNet for brain lobe segmentation, we enable regional analysis of lesion impact. By fusing MRI, clinical, and demographic data, we enhanced EDSS prediction, with LightGBM achieving a high R2 score (0.96 ± 0.07)