The Alzheimer’s Disease Prediction Model Of Regression Analysis Based On Resting State FMRI

Alzheimer’s disease(AD)is an irreversible and common neurodegenerative disease.Clinical features such as memory decline,cognitive disorder,intellectual deterioration,irritability,loss of activity,etc.,and eventually die with atrophy and necrosis of brain neurons.As people getting older,especially for the agedness,AD accounts for a very high proportion,which could be around 50%-60% of dementia population.However,the early development of this disease is very slow and cannot be easily detected,so the early prediction and diagnosis of AD patients are particularly important.With the rapid development of neuroimaging technology,currently resting state functional magnetic resonance imaging(rs-fMRI)has been widely used in the research of AD.This article is mainly based on rs-fMRI,which explores gray matter and functional differences through three different methods: Functional Connectivity(FC),Regional Homogeneity(ReHo),and fractional Amplitude of LowFrequency Fluctuations(fALFF).Explore the role of information reflected by different methods in the diagnosis of AD development,and establish a regression analysis prediction model to provide reference significance for early detection and diagnosis of AD.Firstly,161 subjects were screened from the ADNI database and preprocessed to obtain better rs-fMRI data.FC mainly reflects the synchronization of functional activities between non-adjacent brain regions,ReHo reflects the synchronization of local neuron activities in brain regions,and fALFF reflects the level of spontaneous activity of each voxel at rest from an energy perspective.In this article,the subjects were divided into four groups: the normal control group(NC),the early mild cognitive impairment group(EMCI),the Late MCI group(LMCI),and the AD group.Three methods were used to analyze and calculate the gray matter area of the brain of the subjects,for which statistical analysis was also performed.The analysis results showed that in the pathological changes from NC to EMCI,there were many areas with significant differences in FC and ReHo,so the changes of FC and ReHo might be more suitable for early prediction and diagnosis of AD patients.During the pathological changes from EMCI to LMCI,only FC and ReHo had fewer areas with significant differences;while LMCI to AD,ReHo and fALFF had more areas with significant differences,so these two methods might be more sensitive to advanced AD lesions.Secondly,this paper uses multiple linear regression analysis and ridge regression analysis methods to perform regression analysis on the regions with significant differences in the above three research methods of FC,ReHo and fALFF,and establish prediction models respectively.The results of regression analysis showed that AD had a more obvious impact on FC and ReHo,but had a weaker effect on fALFF.The modeling results of FC and ReHo are better than those of the three methods alone.In the model prediction of AD disease development,by comparing the multiple linear regression analysis model with the ridge regression model,it was found that the multiple linear regression analysis model is better than the ridge regression model.In all the prediction models,FC and ReHo combined the modeling method of multiple linear regression analysis to obtain a better fitting effect.To sum up,in the study of AD lesions,FC and ReHo may be more suitable for early detection and diagnosis,while ReHo and fALFF may be more suitable for advanced research.Compared with fALFF,changes in AD’s condition may have a greater impact on FC and ReHo.And the multiple linear regression analysis prediction model is better than the ridge regression analysis prediction model.The multiple regression model combining FC and ReHo may have greater reference value for predicting the development and diagnosis of AD disease.