Analysis Of Abnormal Brain Function Network Based On EEG And Machine Learning

At present,chronic diseases are the most important causes of death in the world,and tend to be younger.How to reduce the incidence rate,mortality rate and disease burden of chronic diseases has become the focus of attention worldwide.Chronic disease itself has the characteristics of long incubation period,early symptoms and signs are not obvious,often lead to a variety of complications,coupled with limited detection methods,treatment drugs and patients’ own stigma,which bring great challenges to its diagnosis and treatment.Therefore,it is of great significance to explore the mechanism of disease and its representative markers.As we all know,the human brain is like a complex network.Due to the complex connection of different neurons,the functions of the brain regions formed by different neurons are also different.EEG is an electrophysiological signal,which can truly reflect the functional state of the cerebral cortex at any time and has rich physiological information.Therefore,based on EEG data,it is an important way to explore the abnormal mechanism and markers of diseases to construct brain functional network and analyze its topological characteristics by using graph theory.Neurosis is one of the chronic diseases.It is difficult to diagnose neurosis based on whether brain tissue structure is damaged or not,and there are few related studies.In this paper,neurosis is subdivided into physical dimension and emotional dimension,and a feature extraction algorithm for neurosis is explored and designed.The purpose is to locate abnormal connections and identify diseases through attribute features of brain functional network based on EEG in different dimensions.The EEG data of 96 subjects were analyzed,and each data was strictly screened,collected and preprocessed.In order to ensure the reliability of the results,a variety of analysis methods are used,that is,the coherence,phase lag index and weighted phase lag index are selected as the synchronization indexes of EEG signals in different channels,and they are also the measurement of edge in the network.The network is constructed by using threshold selection method of network sparsity,and the four topological attributes,namely aggregation coefficient,average shortest path,local efficiency and global efficiency,are extracted to analyze the differences between the two groups.According to the characteristics of significant differences,the corresponding abnormal connections of brain functional networks were explored.It was found that in the physical dimension,the connections between various brain regions and the parietal lobe,temporal and parietal lobe and occipital lobe were abnormally enhanced,while in the emotional dimension,the connections between frontal lobe,prefrontal lobe and central parietal lobe were abnormally enhanced.Using machine learning algorithm for classification,based on the three indicators of brain network attribute features,using different classification models for classification,the classification accuracy can basically reach more than 70%,and the phase lag index and weighted phase lag index feature combination can achieve more than 90% of the classification accuracy,which also objectively proves that the extracted difference features can be effective to distinguish the physical dimension and emotional dimension of neurosis.In this paper,we analyzed the topological abnormalities of EEG brain network in different dimensions of neurosis.The results can be used as indicators to identify the physical dimension disorder and emotional dimension disorder of neurosis,and the corresponding abnormal brain network connection can be used as the target of human intervention,which provides a new direction for its diagnosis and treatment.