Research On Abnormal Brain Network Of Patients With Schizophrenia Based On Network Representation Learning

Schizophrenia is one of the most common chronic mental diseases in clinical.It is often accompanied by positive symptoms,mainly hallucinations and delusions,and negative symptoms,mainly lack of motivation and cognitive impairment.Although previous studies have conducted a large number of studies on its pathophysiological mechanism from multiple levels of behavior,physiology,and psychology,the specific pathogenesis is still unclear so far.Clinically,there is no reliable biological markers to detect and diagnose patients in time,and there is also a lack of reliable reference standards for the treatment of patients.In recent years,magnetic resonance imaging technology has provided a powerful tool for studying brain activity.Previous analysis show that schizophrenia is a disease that exhibits abnormal structural and functional connections.This article uses structural and functional connections to study the abnormalities of the large-scale brain network of schizophrenia,to explore the pathogenesis of schizophrenia,and to explore whether the abnormalities of the large-scale brain network can be used as biological markers to assist clinical diagnosis disease.The main content of this article includes:In neuroscience,previous explorations of the relationship between large-scale structures and functional brain networks focused on all or part of the statistical correlation,thus ignoring the contextual information of the network,such as the network topology.In this study,we use the network representation learning method to create a high-level representation of the structure or function of the local network to study the abnormality of the functional structure of the brain on the local network in patients with schizophrenia.We found that the relationship between the structure and the function network reconstructed from the network representation learning method is more stable than the partial structure network connection and the function network connection obtained from the traditional correlation method,and they are mainly distributed in high In the firstorder cognitive network.Compared with healthy controls,the application in patients with schizophrenia showed decreased coupling on the executive control network,and increased coupling on the limbric network.In general,network representation learning can more effectively capture the high-level coupling relationship between brain structure and function,and provide us with a good technical means for studying mental illness.Furthermore,we propose a new diagnosis algorithm for patients with schizophrenia based on the abnormal coupling of structure and function of patients with schizophrenia.First,we use graph neural network combined with structural function network for vector representation,and use the representation vector of fusion structure function to construct a demographic network.We innovatively reduce the diagnosis problem to the node classification problem on this network.Next,we used different node representation algorithms to represent the demographic network of subjects containing inter-subject information as feature vectors based on nodes.Based on the representation vectors obtained by different algorithms,we combined different downstream classifiers for classification and comparison.And finally achieved 86% classification accuracy.This article combines the large-scale network analysis of the structure and function of patients with schizophrenia,from a physiological and psychological point of view,reasonably explains the abnormal brain phenomenon of schizophrenia,and provides new ideas for the diagnosis and treatment of neurological and mental diseases.