Alzheimer's disease auxiliary diagnosis system based on fNIRS and graph neural network

Abstract

The invention discloses an Alzheimer's disease auxiliary diagnosis system based on fNIRS and a graph neural network. The auxiliary diagnosis system is used by the following steps: an electroencephalogram detection module is used to perform multi-channel detection on oxyhemoglobin, deoxidized hemoglobin and total hemoglobin when brains of an AD patient and a normal person are in an active state byusing a functional near-infrared spectral imaging device, and an fNIRS data set is constructed; an adjacency matrix construction module is used to constructs an adjacency matrix for the fNIRS data setby using correlation between channels; a node feature matrix construction module is used to perform feature extraction on the fNIRS data set to construct a node feature matrix; a graph structure construction module is used to respectively construct an AD patient graph structure and a normal person graph structure according to the adjacent matrix and the node feature matrix to form a training dataset; and an Alzheimer's disease recognition module is used to train a constructed graph neural network by using the training data set, recognize the test data set of the to-be-tested person by usingthe trained graph neural network, and output an auxiliary diagnosis result. The graph neural network is applied to classification of the fNIRS data.

Description

technical field [0001] The present disclosure relates to the technical field of deep learning, in particular to an auxiliary diagnosis system for Alzheimer's disease based on fNIRS and a graph neural network. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] Graph Convolutional Neural Network (GCN) directly acts on the graph structure through a feature extractor, which is used to model the dependencies between graph nodes. Currently, most graph convolutional neural networks have a similar structure due to the use of convolution operators that can share weights across the graph. Compared with cumbersome fully-connected models, this neural network structure has at least three advantages in graph data: 1) it avoids the parameter explosion associated with fully-connected layers; 2) it allows parameters to be shared across the network and avoid...

AI Technical Summary

Problems solved by technology

The inventors found that patients with cognitive impairment have differences in physiological signals such as EEG signals and normal people, and the parameters of EEG signals in the time domain and frequency domain are different from those of normal people; at present, there is

Images