Association Analysis And Classification Research Based On High-dimensional Shape Descriptors And Dementia Susceptibility Genes

The brains of patients with neurodegenerative diseases usually exhibit abnormalities in morphology,which can be used as important biomarkers for monitoring disease progression and assisting clinical diagnosis.Structural magnetic resonance imaging(sMRI)has the advantages of stable image quality and high spatial resolution,which is helpful to study the spatial pattern of brain morphology and its evolution with disease progression.In recent years,researchers have conducted a large number of studies on the volume of the brain regions in Alzheimer’s disease(AD)patients,especially the hippocampus.However,as the pathological model of AD is complicated,simply measuring the brain volume cannot fully capture the spatial-temporal pattern of brain deformation.How to better capture the subtle morphological changes of the brain structure and use them as effective biomarkers for pathological diagnosis and classification,therefore,has become an important problem that needs to be resolved.The thesis takes sMRI as the research object and takes the morphometrics of subcortical structure as the breakthrough point,hoping to provide a new idea for early intervention of AD.The thesis mainly includes the following two parts:First,combined with the morphological characteristics of the subcortical structure,we pioneer to study the synergistic effects of the risk gene(APOE-CLU)on the subcortical structure: hippocampus and amygdala,and the dynamic changes of this effect over time.The results of the study indicate that the CLU genotype regulates APOE ε4 allele expression.At the same time,CLU C and APOE ε4 alleles showed a cumulative effect on the bilateral subcortical structures,and the atrophic area expanded from susceptible subregions to the whole over time.Based on the above conclusions,we speculate that the synergistic effect of APOE and CLU may increase the risk of AD by affecting the atrophy trajectory of the hippocampus and amygdala.In addition,these findings indicated that non-dementia people who carry both CLU CC and APOE ε4/ε4 genotypes may require more attention for early preventive intervention.Second,we explored more sensitive morphological features of the subcortical surface to distinguish between mild cognitive impairment(MCI)and normal control(Normal Control VS.NC).In this study,a novel High-dimensional Shape Descriptors(HDSD)was introduced to help determine the complicated spatial brain atrophy pattern of MCI patients.The results showed that the combined features(RDmTBM)based on Radial Distance(RD)and multivariate Tensor-based Morphometry(mTBM)have stronger statistical effects than using RD or mTBM.In further research,we proposed a combined model of high-dimensional morphological features and genetic information for MCI-assisted diagnosis research.In the study,patches are used to reconstruct the grid,the vertex values of the grid are extracted as features,sparse dictionary learning and Max-Pooling are used for feature dimensionality reduction,and integrated classifiers(Ada Boost,Gentle Boost,Logit Boost)are used for classifier training.And the ten-fold cross-validation were applied to evaluate the accuracy and stability of the classification.We found that the RDmTBM can effectively capture the impact of MCI on the subcortical structure,and its accuracy in identifying MCI patients under different integrated classifiers reached 82.8%,82.4% and 79.6%,respectively.In addition,the inclusion of genetic information further improves the efficiency of diagnosis.This study confirmed the effectiveness of using RDmTBM to assess the surface deformation of subcortical structures,and showed that abnormal subcortical structures and risk genes can be used as biomarkers for early diagnosis of AD.