FDG-PET Individual Network Construction And Its Characteristics In The Diagnose Of Alzheimer’s Disease

Along with sharply developed of modern brain imaging techniques and big data analysis technology of measure brain processes,researchers are more and more want figure out the structure and function of human brain.Through the researching of human brain and physiological mechanism of behavior,researchers want to investigate the effective of human neural activity,optimize the research and treatment of neuropsychiatric diseases.Alzheimer’s Disease(AD)is the main factor of senile dementia,which has serious influence on middle-aged and elderly.Since people know little about the causes of Alzheimer’s disease,studies on brain networks reported that AD patients with the abnormalities in the brain structure and neuronal activity,while the connectivity alterations and disrupted coordinated organizations appear in the brain network of AD patients.This study aims to find the abnormal topological attributes in the brain network of AD patients to find the new method of early diagnosis.We creatively use 18F-FDG-PET data to build individual brain metabolic network base on cube partition.The significance level of the differences between AD and normal subjects was analyzed,such as connectivity and node degree of metabolic network.And evaluate the possibility of early diagnosis using these attribute,which get from machine learning.The main work is as follows:(1)The differences of mean glucose metabolic rate between the AD patients and the normal control group were analyzed statistically.The results showed that the cubes with significant hypometabolism in AD were localized in middle temporal gyrus,inferior temporal gyrus,right posterior cingulate gyrus,precuneus and angular gyrus.(2)After the individual brain metabolic network based on cube partition was creatively constructed,the connectivity and node degree of individual metabolic network were compared between AD and NC.At same times,we calculated the parameters of their brain metabolic network,such as clustering coefficient,characteristic path length and so on.Studies found that the long-range connections of AD groups were decreased,and the brain metabolic networks of both groups showed small world properties,but the brain metabolic network of AD patients had greater clustering coefficient and the longer absolute path length than normal subjects.(3)Evaluate the possibility of early diagnosis using these attribute,which get from machine learning.Result shows that the performance of extracted features is outstanding,with accuracy of 95.64% and area of 0.9915 under receiver operating characteristic curve.This shows that the imaging features of FDG-PET and the topological characteristics of brain metabolism in the cube partition model have great potential in the early diagnosis of AD.