Predicting Longitudinal Cognitive Decline in MS Using Baseline Clinical Routine T2-FLAIR MRI (2577)

Objective: To predict whether people with multiple sclerosis (PwMS) exhibit longitudinal cognitive decline from baseline clinical MRI data using artificial intelligence and state-of-the-art automated processing methods for measuring atrophy and structural network disruption. Background: Longitudinal cognitive decline in PwMS can in part be predicted from baseline research-quality MRI and intensive processing pipelines. However, newer techniques are available to automatically generate robust quantitative and network-level data from routine, clinical T2-FLAIR MRI, with minimal user expertise. Design/Methods: A retrospective analysis of 147 PwMS was conducted to predict cognitive decline over several years (mean=3.9±2.2) from baseline clinical T2-FLAIR MRI. The previously validated NeuroSTREAM and Network Modification tools were applied on baseline T2-FLAIR MRI to automatically measure lateral ventricular volume and patterns of structural network white matter tract disruption, respectively. This baseline data was then applied to predict cognitive decline, defined as a drop in 4-points or greater on the Symbol Digit Modalities Test (SDMT). The data was divided randomly into two discrete subsets (n=117; n=30). The first was used to train a modular deep neural network to predict cognitive decline and the other was used to test/validate the predictive accuracy of the model. Results: 44.8% of study participants (n=50 training; n=16 testing) exhibited cognitive decline over the period of observation. After training the modular deep neural network, we achieved moderate predictive accuracy in the separate test/validation set (accuracy=63%; sensitivity=0.65; specificity=0.86; false positive=2; false negative=9; F1=0.62). These results compare favorably to previously reported R2 values of 0.20–0.35 on research-quality full connectomic investigations. Conclusions: With the application of state-of-the-art automated MRI processing pipelines and artificial intelligence, cognitive decline in PwMS can be predicted from routine, clinical-quality, T2-FLAIR MRI data with minimal user expertise. These automated methods may allow for patient risk-stratification even in clinical environments. Disclosure: Dr. Fuchs has nothing to disclose. Dr. Benedict has received personal compensation for consulting, serving on a scientific advisory board, speaking, or other activities with Ralph H. B. Benedict is on the speakers’ bureau for EMD Serono, and consults for Biogen Idec, Genentech, Roche, Sanofi/Genzyme, Takeda, NeuroCog Trials, and Novartis.. Dr. Benedict has received royalty, license fees, or contractual rights payments from Psychological Assessment Resources. Dr. Benedict has received research support from Ralph H. B. Benedict has received research support from Accorda, Novartis, Genzyme, Biogen Idec, and Mallinkrodt, and is on the speakers’ bureau for EMD Serono.. Dr. Tran has nothing to disclose. Dr. Brior has nothing to disclose. Dr. Bergsland has nothing to disclose. Dr. Jakimovski has nothing to disclose. Dr. Ramasamy has nothing to disclose. Dr. Zivadinov has received personal compensation for consulting, serving on a scientific advisory board, speaking, or other activities with Received personal compensation from EMD Serono, Genzyme-Sanofi, Celgene, and Novartis. Dr. Zivadinov has received research support from Received financial support for research activities from Genzyme-Sanofi, Novartis, Celgene, KeyStone Heart, V-WAVE Medical, Mapi Pharma and Protembis.Dr. Dwyer has received personal compensation for consulting, serving on a scientific advisory board, speaking, or other activities with Serono Novartis and Clare Medical.. Dr. Dwyer has received research support from Michael G. Dwyer received financial support for research activities from Mapi Pharma, Keystone Heart, Protembis and V-WAVE Medical..