AD/MCI classification using DMN connectivity networks-based HGCN with attention

Neurodegenerative diseases are a diverse group of progressive disorders characterized by the gradual deterioration and eventual death of neurons, leading to cognitive, motor, and functional impairments. Alzheimer’s disease (AD) is one of the latent progressive neurodegenerative disease. Mild cognitive impairment (MCI) is a transitional state between normal aging and dementia. Patients have decline in memory, language and other cognitive functions which others can not perceive. Clinical examination will also find a certain degree of damage, but it is not enough to affect the function of daily life, and can not be explained by known medical or neuropsychiatric diseases. However, the traditional connected network is usually constructed based on the pairwise association between brain regions, so the high-order relationship between them is ignored. This information loss is unexpected, because the brain itself is a complex network, and high-order interaction may contain useful classification information. We propose a new brain connectivity network method for AD/MCI/CN fMRI classification, which is based on hypergraph convolutional networks that can make use of the high-order relationship and get comparable performance with nowadays methods.