Parkinson depression auxiliary identification method based on fMRI graph neural network

Abstract

The invention provides a Parkinson depression auxiliary recognition method based on an fMRI graph neural network. The Parkinson depression auxiliary recognition method comprises the steps that image information of a subject is read, MRI data and files collected by a magnetic resonance scanner of the subject are extracted, and SPM12 software is automatically called for data and file preprocessing; based on the region of interest, performing correlation analysis on each ROI time sequence by using a Pearson's correlation coefficient index according to a brain map, and establishing an effective feature model for the Parkinson depression symptom based on a graph theory; and establishing a graph neural network PR-GNN suitable for Parkinson's depression, inputting the effective feature model into the graph neural network PR-GNN suitable for Parkinson's depression, outputting a classification result, and finally identifying the subject as a healthy type or a PD non-depressive symptom type or a PD depressive symptom type. The three-classification prediction of unknown subjects is realized by using the PR-GNN, and the method is efficient and practical.

Description

technical field [0001] The invention relates to the technical field of Parkinson's identification, in particular to an fMRI graph neural network-based auxiliary identification method for Parkinson's depression. Background technique [0002] Parkinson's disease (PD) is a neurodegenerative disease. There is no "gold standard" for the clinical diagnosis of depressive symptoms. It is mainly inferred by scales, and the diagnosis is somewhat subjective; Degenerative diseases are more common in the elderly, with an average age of onset of about 60 years old, and young Parkinson's disease with onset under the age of 40 is rare. The prevalence of PD among people over 65 years old in my country is about 1.7%, while the world average is 1%. Most Parkinson's disease patients are sporadic cases, and only less than 10% of patients have family history; fMRI is a Blood oxygenation relies on contrast to reflect neural activity, and the analysis of fMRI signals has neurological significance, ...

AI Technical Summary

Problems solved by technology

[0003] In the prior art, graph neural network PR-GNN has been used for the classification and identification of autism spectrum disorder, but there is no use of PR-GNN for the identification of Parkinson's disease depression, and there is no auxiliary diagnosis using PR-GNN. tool, so using the graph neural network PR-GNN to assist in the identification of Parkinson's disease depression is a technical problem that needs to be realized urgently

Images