Machine Learning Techniques for Magnetic Resonance Imaging Analysis in Multiple Sclerosis: A Comprehensive Narrative Review (ORP-04)

Multiple sclerosis (MS) is an autoimmune disease whose main feature is demyelination and neurodegeneration of the central nervous system (CNS). This disease occurs mostly in young adults. The incidence of this disorder is increasing in the world and in recent years it has been reported up to 0.9%. One of the main reasons for the increase in incidence is the improvement of diagnostic methods, including Magnetic Resonance imaging (MRI). Although a set of clinical symptoms enables the final diagnosis of MS, MRI is one of the most popular and essential diagnostic methods for MS diagnosis among doctors and neurologists. In addition to diagnosis, MRI is used to determine the progress of the disease. Despite the importance of MRI in the diagnosis of MS, it is time-consuming, boring, and associated with human errors in diagnosis. Machine learning and deep learning methods have been used in order to improve diagnosis as well as solve problems such as time-consuming. In this method, by using supervised and unsupervised learning algorithms and different reinforcement such as SVM, CNN, and GBM it is possible to predict the progress of the disease, diagnose the disease, and categorize the disease. In the convolutional learning machine, this happens through feature selection, feature extraction, and feature classification. In this study, a comprehensive review of the studies that have been used in recent years on the application of deep learning techniques in MRIMS analysis has been done. The steps involved in computer diagnostic systems that have used deep learning techniques to monitor the progression and diagnosis of MS are reviewed. Most of the articles are in the field of disease diagnosis and determining the subtype of the disease. Future challenges and opportunities for automated MS diagnosis using deep learning and MRI methods are also presented.