The Evolution Modeling And Function Assessment Of Alzheimer's Disease Based On Brain Network Analysis

Alzheimer's disease(AD)has a great impact on the lives of people,especially the elderly.At present,we still know little about the cause of AD and related pathological mechanisms,and there is a lack of effective diagnosis and treatment.Functional magnetic resonance imaging(fMRI)technology has promoted people's understanding of the brain,the brain network technology based on whom provides new ideas for brain science research.In this thesis,the longitudinal analysis of AD neuropathy was carried out at the level of brain network.The brain functional network evolution model of AD based on computational experiments and a method of neural function assessment based on brain network level were proposed.Firstly,most studies on the brain network level of AD were the comparison between AD patients and the healthy,and were often limited to the analysis of global network attributes.Aiming at the problem,this thesis analyzed the brain network changing patterns in each stage of AD based on the resting state fMRI data of each group,and explored the trend of nervous system reorganization from the global,subnet and functional connections.The experimental results showed that the subnet attributes had more statistical differences between patient groups.In addition,we have found for the first time that there were two different changing patterns in the subnet attributes of AD.They were “inversion” and “monotonic” evolution pattern,rather than what was commonly considered “monotonic degradation”.It is suggested that subnet reorganization may play an important role at AD pathologies and subnets with the same changing pattern may have similar neural recombination mechanisms.Secondly,brain lesions are a dynamic process,and a single static network model does not adequately describe the evolution of brain networks in disease.For the problem,we proposed an evolution model of AD brain function network based on computational experiments,simulating the non-monotonic dynamic evolution progression of healthy people via EMCI and LMCI to AD at the brain network level.The model successfully simulated the continuous dynamic change of the brain network topology in each stage of AD.The simulation results and the comparison of random model showed that the model was reasonable and effective.Therefore,this work provided a new perspective for the research on neurological diseases.Finally,AD was often accompanied by nerve function decline.The existing methods for assessing nerve function are mostly based on questionnaire tests,which are influenced by environmental and human factors.Therefore,this thesis proposed a comprehensive brain function assessment method based on brain function network.The model used the Lasso algorithm to sparse features of network attributes,and then obtained the evaluation values of neural functions based on the ordinal regression model.The experimental results showed that the method can well distinguish the population of disease,with good discrimination and certain objectivity as well as potential clinical application significance.