Construction Analysis And Application Of Task State Brain Network Based On MEG

The brain is the most complex network system in the real world.Humans use this network system to process and transmit information between different brain regions to complete cognitive activities.When we understand and study the cognitive neural mechanisms of the brain from the perspective of the network,the brain network provides us with a new method and a new perspective.The brain network based on the magnetoencephalogram(MEG)can reveal the mechanism and characteristics of the brain's cognitive function,and has been widely used in the study of brain cognition.At present,there are still some key problems to be solved in the research of brain cognition based on MEG brain network.First of all,although relevant studies have shown that task-state brain networks have advantages in brain cognition research,there are many ways to construct MEG task-state brain networks.Commonly used are imaginary coherent(IC)and amplitude envelope correlation(AEC).Which method is used to construct a task-state brain network that is more stable and can accurately show the brain changes of the subject during the task?Secondly,existing studies all use the connections in the brain network as indicators to analyze cognitive ability,whether new indicators such as the topological properties of the brain network can be used for research;and the method of combining brain networks and machine learning algorithms is used to analyze the cognitive ability.In the study of cognitive ability,feature selection was not performed before the regression model was constructed,and there was a lack of research on the key connections of the brain in the cognitive process.Therefore,this article first uses the test-retest reliability to compare the consistency and stability of the commonly used MEG task-state brain network construction algorithms;secondly,analyzes the significant degree of the topological attribute difference between the resting state and the task-state brain network The brain network based on which algorithm can accurately represent the brain changes of the subjects under the conditions of the task;finally,the most important reading comprehension ability in the brain cognitive function is selected as the research point,and the selected construction algorithm is applied to the MEG-based task state In the construction and analysis of the prediction model of the reading comprehension ability of the brain network,it provides a more objective index for evaluating the reading comprehension ability of human beings.The main innovative work and research results of this article include:(1)Analyze the test-retest reliability of the MEG task-state brain network construction algorithm,and compare and analyze the stability of the algorithm.The imaginary coherence and amplitude envelope correlation algorithm is widely used in the research of MEG task-state brain network.The higher the algorithm test-retest reliability,the more stable it can ensure the repeatability of the research results.This paper uses the resting state and task state MEG data of two repeated measurements to construct a brain network through two algorithms,and calculates the network topology indicators,and compares and analyzes the test-retest reliability between the indicators.The results show that the retest reliability of the IC-based brain network topology index is higher than that of the AEC-based brain network.(2)Analyze the difference between the topological indicators of the MEG task state brain network and the resting state brain network.In this paper,IC and AEC methods are used to construct the task state and resting state brain networks,and the network topology indicators are calculated,and the difference between the task state brain network and the resting state brain network topology indicators is analyzed.The results found that the difference between the resting state and task state brain network topology indicators constructed based on the IC method was more significant than the difference between the AEC brain network,indicating that the task state brain network constructed based on the IC algorithm showed more accurate performance of the brain network.Try the changes in the brain when performing tasks.Subsequently,the node efficiency with differences in 148 brain regions was screened out as an indicator to correlate with reading comprehension scores.It was found that the number of brain regions related to the scale was small,and the model constructed using node efficiency as a predictive feature could not successfully predict reading.Comprehension ability,indicating that node attributes cannot be used as an objective indicator to predict reading comprehension ability.(3)Use improved machine learning algorithms to build a reading comprehension prediction modelIn view of the current problems in the pre-models for reading comprehension,this paper uses partial least squares(PLS)to construct a reading comprehension prediction model based on MEG resting state and task state data,and uses univariate feature selection algorithms for feature selection,and finally establishes reading comprehension predictive model,through the two indicators of model prediction R~2 and model mean square error,analyze and evaluate the predictive ability of the model,and determine the key functional connection in reading comprehension.Experimental results show that the partial least squares regression model based on MEG imaginary coherence function connection can successfully predict reading comprehension scores;the model with univariate feature selection has higher predictive performance and more accurate prediction;and it is found that the magnetoencephalogram related to reading comprehension is used Task state data sets are more suitable for predicting reading comprehension ability than resting state data sets.