Prediction of Conversion from CIS to Clinically Definite Multiple Sclerosis Using Convolutional Neural Networks

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Imaging

Background:

• Multiple sclerosis (MS) is a chronic central nervous system disease.
• Early diagnosis and treatment are crucial for preventing MS-related disability.
• Predicting conversion from clinically isolated syndrome (CIS) to clinically definite MS (CDMS) is vital for timely intervention.

Purpose of the Study:

• To develop and validate a deep learning algorithm for predicting CIS to CDMS conversion.
• To leverage convolutional neural network (CNN) models for analyzing MRI scan features.
• To assess the algorithm's performance using multi-scanner MRI data.

Main Methods:

• A fully automated CNN algorithm, based on VGG16 architecture, was developed.
• The algorithm was trained and tested on volumetric MRI scans from 49 patients (7360 images total) acquired at two time points.
• Preprocessing steps and pretraining on the ADNI dataset were employed to enhance efficiency.

Main Results:

• The algorithm achieved a prediction accuracy of 88.8% for CIS to CDMS conversion.
• The area under the curve (AUC) for the prediction model was 91%.
• The developed CNN approach demonstrated high reliability in predicting clinical outcomes.

Conclusions:

• A highly accurate deep learning algorithm can reliably predict multiple sclerosis conversion using MRI data.
• The automated CNN model shows promise for early identification of patients progressing to CDMS.
• This AI-driven approach facilitates timely treatment initiation and improved patient management.