Research On Auxiliary Diagnosis Of Alzheimer's Disease Based On Brain Functional Networ

Tau protein positron emission tomography imaging（tau-PET）is an important tool for studying progression of Alzheimer’s disease（AD）.The most commonly utilized quantitative index in tau-PET is the standard uptake value ratio（SUVR）.Topological information derived from different brain regions has also been linked to tau pathology.However,previous studies on PET-based brain network connectivity were mostly conducted at the group level,which only reflected the average functional changes of the group rather than individual differences.Therefore,this paper has proposed a method for constructing a tau-PET network based on individual level using the idea of constructing an individual covariance network based on magnetic resonance imaging（MRI）.From the perspective of brain networks,it explores the changes in tau deposition and the correlation between brain regions in Alzheimer’s disease at different stages.Experimental data were obtained from ¹⁸F-AV1451 PET scan images of the ADNI database,and the HABS database.Comparative analysis of data processed with partial volume correction（PVC）and without PVC（no PVC）.The main research content of this paper includes the following aspects:（1）To investigate changes in the whole-brain network with disease progression.An individual whole-brain network was constructed for each subject using the region of interest（ROI）SUVR obtained through image preprocessing.Analyze the heterogeneity of the brain network,calculate network properties,and perform group difference analysis.The correlation of individual average brain network connectivity and SUVR with clinical rating scales was analyzed.（2）To investigate changes in functional networks with disease progression and identify potential network features that can help in the early diagnosis of AD.Seven functional clusters were extracted from relevant networks in each subject:medial temporal lobe,cognitive control,executive control,default mode,visual,somatomotor,and language networks.Seven functional subnetworks and a functional network with functional clusters as nodes were respectively constructed.Effect sizes were used to evaluate the connectivity strength of functional subnetworks and inter-group differences of SUVR,to identify functional clusters with significant changes in disease progression,and to evaluate the connectivity strength of functional networks.Find the network features that can assist in AD staging diagnosis.（3）Using machine learning methods to evaluate the feasibility of the proposed network features for early diagnosis of AD.The results show that the features extracted in this paper perform well in classification,which verifies that the network features proposed in this paper are effective and reliable for distinguishing AD stages.Overall,the conclusions of the ADNI and HABS datasets are consistent.The results of network feature analysis were better than those of quantitative analysis of traditional SUVR,and the results of data analysis treated with PVC were better than those without partial volume correction.The method of constructing individual brain networks proposed in this paper can potentially monitor Alzheimer’s disease progression at the subject level,in contrast to group based approaches,which have great potential in monitoring disease progression and response to treatment.