Longitudinal Prediction Modeling of Alzheimer Disease using Recurrent Neural Networks

Abstract

This paper proposes an implementation of Recurrent Neural Networks (RNNs) for (a) predicting future Mini-Mental State Examination (MMSE) scores in a longitudinal study and (b) deploying a multiclass multimodal neuroimaging classification process that involves three different known stages of Alzheimer's progression, cognitively normal (CN), Mild Cognitive Impairment (MCI) and Alzheimer's Disease (AD). This multimodal data is fed into two well-studied variations of the RNNs; Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU). The accuracy, F-score, sensitivity, and specificity of the models are reported for the classification task as well as the root mean square error (RMSE) and correlation coefficient for the regression task. The results demonstrate the superiority of the proposed model over state-of-the-art classification and regression techniques of Support Vector Machine (SVM), Support Vector Regression (SVR) and Ridge Regression.