Automated Detection of Multiple Sclerosis Using Machine Learning

Overview

Multiple Sclerosis (MS) is a complex neurological disorder characterized by nerve damage to the central nervous system, affecting over two million people worldwide. Symptoms include but are not limited to impaired coordination, body weakness, poor balance, etc. Considering MS is a progressive disorder, the impact on an affected individual increases over time if left untreated. Therefore, effective treatment and prevention of the more severe instances of MS depend on early and accurate detection. Conventional diagnostic methods, which depend on human observations of imaging studies such as Magnetic Resonance Imaging (MRI), are prone to errors, especially with less experienced neurologists. Thus, this study presents an automated approach to enhancing the accuracy of MS diagnosis, while also highlighting indicators of MS within the MRI scans. Convolutional neural networks (CNNs) can recognize intricate patterns within medical imaging data, therefore it serves as a powerful tool for this task. Unfortunately, MS remains as an under researched topic in the medicine field, thus, there are not many practical datasets for this task. That being said, with the use of a dataset of roughly 1200 axial MRI scans, a custom CNN model for the purpose of accurately distinguishing between a healthy MRI scan and a MS MRI scan was developed. The network’s architecture included convolutional layers, batch normalization, pooling, fully connected layers, and dropout for regularization. After training and validation had been completed, the model proved to be effective with a peak accuracy over 95% as well as up to standard values for other metrics such as F1 score, precision, and recall. This project could be incorporated in clinical practice, by assisting healthcare officials with expediting the process of making timely and accurate diagnoses. The program’s feature of highlighting certain indicators could potentially assist in future research for indications of MS.

Competition history

• AJAS 2025 Category not listed

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• AJAS None projects
• Every AJAS 2025 project