Multimodal PET Data Analysis Of Alzheimer’s Disease And Its Application

Alzheimer’s disease(AD),the leading cause of dementia,is a neurodegenerative disease that greatly threatens the health of the elderly.AD has caused heavy economic and social burdens all around the world.Unfortunately,for now there is no medicine to cure AD.The death rate of AD has become the third highest,preceded only by cancer and cardiovascular diseases.However,the pathogenesis of AD is still unclear,which leads to great difficulty in the development of medicine of AD.The main pathological features of AD include neuronal loss(brain atrophy),extracellular Aβ plaques(formed by extracellularly deposited Aβ protein)and intracellular neurofibrillary tangles(due to hyperphosphorylated tau protein).The clinical manifestations of AD mainly include progressive cognitive decline(including memory,intelligence,orientation,judgment,etc.),affective disorders,and decline in daily living ability and behavioral disorders.The development of brain imaging technology provides effective methods for the clinical diagnosis,efficacy evaluation and pathogenesis research of AD.Positron emission tomography(PET),a non-invasive medical imaging technology with high sensitivity and specificity,has been widely used in the clinical diagnosis,efficacy evaluation and mechanism research of AD.PET is an important application of particle physics and nuclear physics in the field of medical image research.The molecular probes of PET consist of positron-emitting nuclide-labeled compounds can specifically bind to a certain tissue,cell or molecule,thus researchers could get the biological information in the human body.Various biological information in the human body could be obtained with different molecular probes of PET.Currently,Aβ-PET,tau-PET,and FDG-PET are the three most prevalently used PET modalities in clinical and mechanism research of AD.Aβ-PET can detect extracellularly deposited Aβ plaques,tau-PET can detect excessively deposited hyperphosphorylated tau protein in neurons,and FDG-PET can detect glucose metabolism in neurons.FDG-PET reflects neuronal activity and functional status because glucose metabolism is the only source of energy for the brain.Combination of the above three PET modalities can provide us with comprehensive imaging information from pathological hallmarks to changes in brain function,which is important for the further understanding of the pathological mechanism of AD.However,how to accurately extract valid biological information from different PET modalities and comprehensively analyze it is a challenge in AD studies.Therefore,this paper focuses on the study of joint analysis of multi-modal PET data and its application in the following four aspects.1.The establishment of Aβ-PET and tau-PET standard human brain template.Standard brain template,which provides unified reference space for the analysis of imaging data,is a prerequisite for the voxel-wise analysis of PET data.To solve the lack of standard human brain templates in Aβ-PET and tau-PET data processing,this paper established the Aβ-PET and tau-PET standard human brain template based on a large number of data obtained from ADNI(Alzheimer’s Disease Neuroimaging Initiative),a globally public database for AD research.Besides,the established Aβ-PET and tau-PET standard human brain template are in the same space as the human brain atlas,which is the basis of the establishment of the subsequent single-modal PET data processing method.2.The establishment of single-modal PET data processing method.Based on the established Aβ-PET and tau-PET standard human brain templates,a single-modal PET data processing method was established and various quantitative indicators were used to measure its performance in PET data analysis.Besides,this method was applied to Aβ-PET and tau-PET images to study the neural changes in AD patients with visuospatial dysfuncitons.Results showed the method established in this paper significantly outperformed the one prevalently used in PET data analysis at present.Furthermore,correlation analysis indicated that the impairment of integration of visuospatial information may be the cause of visuospatial dysfuncitons in AD patients,which is important for revealing the neural mechanism of visuospatial dysfuncitons of AD.3.A comprehensive analysis of the multi-modal PET images was performed on basis of single-modal PET data analysis.After constructing the FDG metabolic network,graph theory analysis was introduced to calculate the various attributes of the network,subsequently correlation analysis was performed to measure the relationship between Aβ,tau deposition and attributes of FDG metablic network.Results revealed significantly positive correlation between Aβ deposition and attributes of FDG metabolic network in mild and moderate AD stage,and significantly negative correlation was observed between tau deposition and attributes of FDG metabolic networks in severe AD stage.These findings revealed the different roles of Aβ and tau in the neurodegeneration of AD,which may be meanfuling for further understanding of the pathological mechanism of AD.4.To solve the problem in the current multi-modal PET data processing,a new method for the joint analysis of multi-modal PET data was proposed.Firstly,given that PET images are comprehensive reflection of multiple physiological processes,only a part of which are related to AD pathology,and that the signals related to physiological processes should not be negative,PET images were decomposed into a few physiologically explainable basic images which were all non-negative via the nonnegative matrix factorization algorithm.Then,the random forest(RF)algorithm was applied to screen out feature images that related to AD pathology.This method provides a powerful analytical tool for the study of the pathological mechanism of AD.In short,this paper addressed a number of methodological improvements and applications on both single-and multi-modal PET data processing in AD studies,which could promote the understanding of pathology of Alzheimer’s disease.