Iterative sparse and deep learning for accurate diagnosis of Alzheimer's disease

Deep learning techniques have been increasingly applied to the diagnosis of Alzheimer's disease (AD) and the conversion from mild cognitive impairment (MCI) to AD. Despite their prevalence, existing methods usually suffer from using either irrelevant brain regions or less-accurate landmarks. In this paper, we propose the iterative sparse and deep learning (ISDL) model for joint deep feature extraction and critical cortical region identification to diagnose AD and MCI. We first design a deep feature extraction (DFE) module to capture the local-to-global structural information derived from 62 cortical regions. Then we design a sparse regression module to identify the critical cortical regions and integrate it into the DFE module to exclude irrelevant cortical regions from the diagnosis process. The parameters of the two modules are updated alternatively and iteratively in an end-to-end manner. Our experimental results suggest the ISDL model provides a state-of-the-art solution to both AD-CN classification and MCI-to-AD prediction.