A feature extraction method for neurosis based on graph theory and machine learning

Abstract

The invention discloses a neurosis feature extraction method based on graph theory and machine learning. Firstly, the brain function network of neurosis's body dimension and emotional dimension is constructed based on EEG data, and 62 channels are used as nodes of the network, and the phase lag index is respectively selected. and the weighted phase lag index as a measure of edges in the network. Extract the network topology attribute corresponding to the two indicators—global efficiency, and use it as a feature vector to identify the body dimension and emotional dimension after fusion, and use the machine learning classification model to classify, and the brain function network corresponding to this feature is mapped to the 3D brain model It can locate the abnormal connection position of the network in two dimensions. The present invention can be further used to provide markers for the recognition of neurosis, and to further study the targets of diseases.

Description

technical field [0001] Aiming at the somatic dimension and emotional dimension under neurosis, the present invention proposes a neurosis feature extraction method combining EEG brain function network construction based on graph theory and machine learning method. Background technique [0002] The morbidity, mortality and morbidity rate of chronic diseases remain high, and tend to be younger. They are extremely hidden, and the early symptoms are not obvious. Various mental disorders such as anxiety, depression, neurasthenia, etc. Collectively referred to as neurosis, it is a type of chronic disease, and seeking effective markers for the disease has become one of the problems to be solved urgently. The human brain itself is an extremely complex system, with intricate connections between neurons, based on graph theory [1] brain functional network [2] Analytical methods provide insight into the structure of the brain, how functional areas work, and how they interact and coordi...

AI Technical Summary

Problems solved by technology

Traditional EEG feature extraction of independent channels or brain regions based on time domain, frequency domain or time-frequency domain, such as information entropy [6] ,power spectrum [7] etc. are still widely studied, but the disease is usually not the functional impairment of independent channels or brain regions, but the abnormal interaction and coordination of different functional brain regions, and these characteristics cannot effectively solve the common source problem

Images