Longitudinal analysis for Alzheimer's disease diagnosis using RNN

Abstract

Alzheimer's disease (AD) is a kind of neurodegenerative disorder with progressive impairment of memory and cognitive functions. Structural magnetic resonance images (MRI) is widely used as an important imaging modality for AD diagnosis. Most of existing methods for MRI classification are based on image data of single time point. However, the longitudinal analysis of sequential MR images is also important to model and measure the progression of the disease along the time axis. In this paper, we present a classification framework based on combination of Multi-Layer Perceptron (MLP) neural network and Recurrent Neural Networks (RNN) for longitudinal analysis of MR images for AD diagnosis. First, MLP is constructed to learn the spatial features of MR images with the task of AD classification. After that, RNN with cascaded two bidirectional gated recurrent units (BGRU) layers is trained on the MLP outputs for extracting the longitudinal features from the imaging data of multiple time points, providing a final classification predicting score. The proposed method can automatically learn the spatial and longitudinal features from the imaging data of multiple time points with variable length for classification. Our method is evaluated using T1-weighted structural MR images on 428 subjects including 198 AD patients and 229 normal controls (NC) from Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Experimental results show the proposed method achieves an accuracy of 89.7% for AD classification, demonstrating the promising classification performance.