Multiple sclerosis classification and segmentation in neuroimaging MRI using different machine and deep learning techniques: a review

Abstract

Multiple Sclerosis (MS) is a type of brain disease that affects both the brain and spinal cord. In order to diagnose MS, many modalities are used, including Magnetic Resonance Imaging (MRI). MRI modalities are noninvasive medical tests that provide physicians with detailed images containing essential details on the anatomy and physiology of the brain. Diagnosing with MRI of the brain is crucial since it is laborious, time-consuming, and, above all, prone to manual or human error. In the past ten years, Artificial Intelligence (AI)- based Computer Aided Diagnostic (CAD) tools have been increasingly popular due to their low manual error rate and ability to produce accurate and dependable findings for diagnosing Multiple Sclerosis (MS) using MRI neuroimaging modalities. Automated MS diagnosis in AI is performed using both traditional and contemporary AI. Machine Learning (ML) approaches, which rely on selecting and extracting features through trial and error, were utilized in classical AI. Meanwhile, modern AI uses Deep Learning (DL) techniques, which extract and select suitable features automatically, save time and are more efficient than standard classical ML approaches. In this work, we give a summary of recent automated MS diagnosis approaches that combine MRI neuroimaging modalities with ML and DL algorithms. Segmentation and classification are two of the main categories into which AI techniques for MS medical diagnosis can be roughly divided, where each one of them is divided based on whether a supervised learning, un-supervised learning or both are used together to get better analysis for MS diagnosis. We briefly discuss the related work in each category, and finally, we present the important drawbacks and challenges for each work as well as some proposed ideas that can solve the drawbacks and challenges faced.