Multi-channel Sparse Graph Transformer Network for Early Alzheimer’s Disease Identification

Abstract

With the aging of the global population and increase in life expectancy, the prevalence, incidence and mortality of Alzheimer's disease (AD) have increased rapidly. Clinical intervention via early diagnosis can delay the AD progression and improve its prognosis. In this paper, we design a novel multi-channel sparse graph transformer network of automatic early AD identification. The proposed method fuses each subject's non-image information and image information from the functional magnetic resonance imaging and diffusion tensor imaging. The fused information via local weighted clustering coefficients can be used as the input of the multichannel sparse graph transformation network for early AD identification. Our proposed method achieves promising identification performance on the public Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.