Correlation Between Multiple Sclerosis Lesion Areas in Brain Magnetic Resonance Imaging and Patient’s Disability

Abstract

Magnetic Resonance Imaging (MRI) plays a very important rule to evaluate Multiple Sclerosis (MS) disease at drug treatment, treatment phase and patient’s follow-up. Identification and characterization of MRI features that are related to MS patient’s disability could be beneficial for efficient treatments, better patient follow-up and avoiding long procedures of physical examination to score MS patient’s disability using Expanded Disability Status Scale (EDSS). This study aims to investigate the correlation between segmented MS-lesion areas in brain MRI and patient’s disability score. Features from manually MS-lesion segmentation done by expert and features from automated MS-lesion segmentation were investigated. Brain extraction, smoothing, threshold-based classifier and SVM have been used for automated MS-lesion segmentation. The automated segmentation produced a Dices Similarity Coefficient (DSC) of 0.5 and high false-positive rate, which indicates the seen and unseen lesions are exist. From the observation, features from automated MS-lesion segmentation have been successfully classified MS patients into two groups at 3.5 EDSS with 100% accuracy using threshold-based classifiers while features from manual lesion segmentation were failed to split MS patients into any group. In conclusion, segmented MS-lesion areas that were obtained by an automated method that only produced DSC of 0.5 with seen lesion correlate with MS patient’s disability, while seen lesion area segmented by manual was not correlated.