Multiple Sclerosis (MS) Classification and Detection

Abstract

Multiple Sclerosis (MS) is a disease that is commonly seen between young people, and it is a neurological disease. Detection of this disease is important to help control the disease. To do such a thing, the use of Magnetic resonance imaging (MRI) plays a significant role as the white lesions matter in the images helps in detection of MS. In this study, the detection of MS using MRI was done using multiple approaches like proposed ResNet-50, CNN, CNN with ResNet-50 and with different classifiers. The dataset that was used contained a total number of images 5242, 2621 of them were MRI images of brain with white matter lesion and same number for non-MS brain images. Training the data using the mentioned models has given accuracy results ranging from 83% to 99%. Showing that CNN was the best used model with 99%, while ResNet50 accuracy was 90%±0.5. Moreover, CNN with ResNet50 gave an accuracy of 95%±0.5, finally the last model that was PCA with SVM gave an accuracy of 83%