A Deep Learning Approach to Predicting Disease Progression in Multiple Sclerosis Using Magnetic Resonance Imaging

Objectives: Magnetic resonance imaging (MRI) is an important tool for diagnosis and monitoring of disease course in multiple sclerosis (MS). However, its prognostic value for predicting disease worsening is still being debated. The aim of this study was to propose a deep learning algorithm to predict disease worsening at 2 years of follow-up on a multicenter cohort of MS patients collected from the Italian Neuroimaging Network Initiative using baseline MRI, and compare it with 2 expert physicians. Materials and methods: For 373 MS patients, baseline T2-weighted and T1-weighted brain MRI scans, as well as baseline and 2-year clinical and cognitive assessments, were collected from the Italian Neuroimaging Network Initiative repository. A deep learning architecture based on convolutional neural networks was implemented to predict: (1) clinical worsening (Expanded Disability Status Scale [EDSS]-based model), (2) cognitive deterioration (Symbol Digit Modalities Test [SDMT]-based model), or (3) both (EDSS + SDMT-based model). The method was tested on an independent data set and compared with the performance of 2 expert physicians. Results: For the test set, the convolutional neural network model showed high predictive accuracy for clinical (83.3%) and cognitive (67.7%) worsening, although the highest accuracy was reached when training the algorithm using both EDSS and SDMT information (85.7%). Artificial intelligence classification performance exceeded that of 2 expert physicians (70% of accuracy for the human raters). Conclusions: We developed a robust and accurate model for predicting clinical and cognitive worsening of MS patients after 2 years, based on conventional T2-weighted and T1-weighted brain MRI scans obtained at baseline. This algorithm may be valuable for supporting physicians in their clinical practice for the earlier identification of MS patients at risk of disease worsening.