 The study investigated the use of various combinations of imaging features and machine learning algorithms to predict Montreal Cognitive Assessment, MoCA, scores in Parkinson's disease patients at year four. The results showed that the combination of clinical features, radiomic features, and machine learning algorithms provided the most accurate predictions of MoCA scores. This suggests that combining multiple sources of data and machine learning algorithms may provide more accurate predictions than using any one source alone. This article was authored by Maudie Hossien-Zeta, Armin Gorgie, Ali Fathy-Jasdani, and others.