Detecting MCI: A New Approach Using Brain and Genetic Data

Mild cognitive impairment (MCI) is a crucial early warning sign for Alzheimer's disease. Yet, many current methods only focus on brain images, ignoring other valuable data like genetics and clinical info. This can lead to a lot of noise in the data, making it hard to spot the real patterns. To fix this, researchers created a new system called Multimodal Multiview Bilinear Graph Convolution (MMBGCN). First, they pulled out important features from brain MRI scans, like grey matter, white matter, and cerebrospinal fluid. These features were then combined with non-imaging data to build a shared adjacency matrix. This matrix helps create a multiview network that considers how non-imaging data might affect the disease. After cleaning up the MRI data with weights, the team used bilinear convolution to restore the brain's spatial patterns. Finally, they mixed these patterns with genetic info for a more accurate disease prediction. Testing on the ADNI dataset showed this method worked better than others, hitting an average accuracy of 89. 6% in binary classification tasks. This study paves the way for using multimodal data in MCI diagnosis.