Study on Deep Learning Based Classification Models for Multiple Sclerosis in MRI Datasets

The deep learning-based approach for identifying Multiple Sclerosis in the human brain using MRI datasets has been proposed in this study. Sclerosis affects the functionalities of the spinal cord and central nervous system. The investigations on sclerosis-affected brain cells show improved accuracy though they do not provide accurate treatment and diagnosis planning. This study addresses the challenges encountered in the treatment planning of sclerosis-affected brain disorders by incorporating deep learning techniques. For MRI slice datasets, VGG16, CNN, ResNet50, Inception V3, and for MRI volume, 3D models of Inception V3 and ResNet50 are implemented and tested to identify and extract multiple sclerosis. Categorical loss functions and Adam optimizer are embedded to train the models, enabling to learning of discriminative features for accurate classification of sclerosis-affected cells. Experimental results, axial & sagittal accuracy, f1-score, confusion matrix, and qualitative outcomes, demonstrate promising performance in accurate prediction of multiple sclerosis for the early diagnosis and treatment planning. This study contributes to the effort to develop advanced computational tools for aiding the diagnosis of multiple sclerosis-affected patients.