Research And Development Of Imaging Genetics Modeling Algorithm For Alzheimer'S Disease

Alzheimer's disease(AD)is a neurodegenerative disease that occurs frequently in the elderly.The patients show dementia symptoms in all aspects,which seriously affects their ability to live and even threatens their lives.Mild cognitive impairment(MCI)is the early stage of AD,between normal aging and AD.Because of the latent condition and long course of AD,when patients have obvious symptoms,they often reach the late stage of AD and can not effectively treat the disease.And because the earliest symptoms of AD,such as short-term memory loss and paranoid suspicion,are often mistaken for being related to aging and stress,or confused with symptoms caused by other brain diseases,accurate early clinical diagnosis of AD is still very important.In the related studies of Alzheimer's disease,patients' disease stages are usually divided into three categories: healthy control(NC),mild cognitive impairment(MCI)and Alzheimer's disease(AD).Nuclear magnetic resonance imaging(MRI),a neuroimaging model,has been widely used in the classification,multi-classification and other research of AD.Meanwhile,recent genome-wide association studies(GWAS)data have been used to characterize several potential genetic risk factors for AD.Therefore,from the perspective of imaging genetics,combining the MRI data with GWAS data to more accurately determine the staging of patients is the biggest goal of this paper.The main work of this paper is as follows:First,we use FreeSurfer to pre-process the standard magnetic resonance imaging(MRI)data provided by the Alzheimer's Neuroimaging Initiative(ADNI)database.Then we intercept two-dimensional images from each brain region,extract the gray distribution,anatomical parameters and Triz features of each brain region,and merge them as the features of the MRI part.Then we use the feature selection method which combines t-test and SVM-RFE to select features.Eight common classification models are used to train and evaluate the obtained MRI features.We also carried out gender difference analysis to prove that gender factors have a great impact on AD.At the same time,we separated male and female samples to establish a three-class model,and carried out experiments on gender complementarity.The results show that gender-based modeling has brought about significant performance improvement.Then,we use the single nucleotide polymorphism(SNP)data provided by ADNI database to model it.Then,two different data fusion schemes are proposed for multi-modal data fusion.The fusion model of magnetic resonance image and SNP data is built,and the gender-based model is also built.The experimental results show that our NC-MCI-AD classification performance is improved by fusing SNP data.Compared with other research results,our algorithm has higher accuracy.