Improving EEG-based Alzheimer’s Disease Identification with Generative Adversarial Learning

Alzheimer’s disease (AD), as the most common type of dementia, has induced great concentration due to its serious situation. Effectively identifying AD from clinical data plays an important role in exploring AD-related indicators for the risk evaluation of AD at an early stage. Benefiting from the low-cost, non-invasive, and portable advantages, electroencephalograph (EEG) signals have been considered as a promising way of quantifying abnormal brain patterns related to AD at the cortical level. This paper proposes a data augmentation strategy for improving AD identification based on generative adversarial learning. The physiological reasonability and the effectiveness of our generated data have been proved through the experiments implemented based on the recorded EEG data reconstructed in the brain functional network feature space.