A genetic algorithm optimized artificial neural network for the segmentation of MR images in frontotemporal dementia

Abstract

Frontotemporal Dementia (FTD) is an early onset dementia with atrophy in frontal and temporal regions. The differential diagnosis of FTD remains challenging because of the overlapping behavioral symptoms in patients, which have considerable overlap with Alzheimer's disease (AD). Neuroimaging analysis especially Magnetic Resonance Image Imaging (MRI) has opened up a new window to identify, and track disease process and progression. In this paper, we introduce a genetic algorithm (GA) tuned Artificial Neural Network (ANN) to measure the structural changes over a period of 1year. GA is a heuristic optimization method based on the Darwin's principle of natural evolution. The longitudinal atrophy patterns obtained from the proposed approach could serve as a predictor of impending behavioral changes in FTD subjects. The performance of our computerized scheme is evaluated and compared with the ground truth information. Using the proposed approach, we have achieved an average classification accuracy of 95.5 %, 96.5% and 98% for GM, WM and CSF respectively. © 2013 Springer International Publishing.