The Application Research Of Genetically Evolved Cluster Algorithm In Imaging Genetics

The brain is the most special organ of the human body,and its complex operation mechanism is inseparable from the joint effect of the nervous system,genes and other substances.Imaging genetics is a new interdisciplinary field that combines neuroimaging and genetics,allowing researchers to explore how the brain works at two levels of the macro and micro.Additionally,machine learning has been widely applied to the complex problem of brain.Therefore,based on the data of imaging genetics,this paper used the genetically evolved cluster algorithm to conduct in-depth research on patients with early mild cognitive impairment or late mild cognitive impairment.The specific research contents are as follows:(1)We propose a genetically evolved cluster algorithm--a new feature extraction and classification method.This paper introduces the ideas of genetic evolution and ensemble learning.The samples and features are randomly selected to build clusters,and the irrelevant or redundant features gradually eliminated in the process of genetic evolution.Through automatic optimization within the global scope,this paper improves the performance of the model and achieves better generalization ability.(2)We use genetically evolved random forest to study early mild cognitive impairment.Firstly,the resting state functional magnetic resonance imaging data and gene data of 37 patients with early mild cognitive impairment and 36 normal subjects are fused to construct fusion features.Secondly,the genetically evolved random forest modal is constructed to distinguish patients from normal subjects and extract the optimal features.Finally,the optimal features are further analyzed to find out the abnormal brain regions and abnormal genes.The results show that compared with other methods,the model has better classification performance and feature extraction ability.We also find out some diseaserelated pathogenic brain regions and abnormal genes.This model is beneficial to the discovery of multifactorial pathogenesis,and provides a new idea for the clinical diagnosis and treatment of early mild cognitive impairment.(3)We use genetically evolved random neural network cluster to study late mild cognitive impairment.Firstly,the imaging genetic data of 26 patients with late mild cognitive impairment and 36 normal subjects are fused.Secondly,a genetically evolved random neural network cluster modal is constructed to classify samples and extract optimal fe atures.Finally,abnormal brain regions and abnormal genes in late mild cognitive impairment are identified.Compared with other methods,the classification and feature extraction framework based on the genetically evolved random neural network cluster achieves better results.The results show that this model is conducive to discovering the multi-factor pathogenesis of patients with late mild cognitive impairment,providing strong support for the clinical diagnosis and treatment of advanced mild cognitive impairment,and promoting the application of machine learning algorithms in precision medicine.