Research On The Characteristics Of Brain Function Network And Brain Muscle Coherence Of Hand Extremity Dyskinesia

Hand limb dyskinesia is a common complication in stroke patients.Since hand function occupies a large proportion in daily life,effective restoration of hand function is of great significance to the rehabilitation of stroke patients.Current studies have shown that the mechanism of functional recovery and remodeling of central nervous system injury after stroke has not been fully clarified.But more and more results show that different brain regions are functionally separate and integrated.Brain function network is an intuitive description of the interactive integration of dynamic activities between different brain regions in the brain structure.Since brain function network can integrate information from different brain regions,it can be used to characterize changes in the cerebral cortex of hand-limb dyskinesia as a whole.Analysis of brain function network is helpful for early prevention and rehabilitation of patients with limb movement disorders.At the same time,brain function network can also provide some new perspectives for the pathological mechanism of patients and expand the theoretical study of complex brain network.Therefore,we take the brain function network as the carrier to study the hand limb movement disorder.The main contents and innovations of this paper are as follows:(1)A brain network synthesis feature of hand limb movement is proposed.Brain function network analysis of different movements of the hands and limbs in healthy and patient subjects showed that the main activation of limb movements was the contralateral region of the cerebral cortex.For specific movements(clenching,wrist flexion,and elbow flexion),the activation each subregion of the opposite region varies.According to this phenomenon,this brain network is divided into left and right brain network,intuitive show the difference between left and right hand movements.We use the network characteristics of left and right brain networks: average node degree,average path length and average clustering coefficient to characterize the difference between left and right hand movements.Due to the relatively independent analysis method of left and right brain network ignores the complexity of electroencephalogram(EEG)signals,we combine the characteristics of brain network with the sample entropy characteristics of EEG signals at C3 and C4 leads to construct the feature vector with both distribution and directivity.This vector not only has the global consideration,but also can highlight the key points,and can more comprehensively reflect the physiological and electrical activity characteristics of the cerebral cortex activated by left and right hand movement.(2)We integrate the coherence feature of brain muscle into the construction of brain function network,and propose a construction method of brain function two-layer network based on minimum spanning tree and regional network.There must be a coupling relationship between cerebral cortex and muscle function when subjects perform hand movements.Therefore,we use the coherence function to calculate the coherence coefficient of EEG,and define the intensity index of corticomuscular coupling(CMC)to quantify the coupling strength between EEG and electromyogram(EMG).We ranked the coherence between the 32-lead EEG signal and the target motor limb-related muscle EMG signal by weighted coupling strength,and gave the core lead combination of healthy and patient subjects under each action.The minimum spanning tree of the brain network connects 32 leads to ensure the index structure and profile of the original network and network connectivity.Since the minimum spanning tree does not have the characteristics of small-world networks,we propose to construct a regional network based on the combination of core leads.Regional network focuses on the specific performance of brain muscle coherence characteristics in the network,highlighting the personalized network characteristics of each hand and limb movement mode.We extracted the diameter and average eccentricity of the minimum spanning tree,the average node degree,average clustering coefficient and average path length of the regional network as the features of the two-layer network to characterize the classification information of multi-class hand movements.(3)In our research,kernel canonical correlation analysis(KCCA)fusion algorithm is introduced to fuse brain function network characteristics with sample entropy,minimum spanning tree and regional network characteristics.Then,we input the fused characteristics into the support vector machine(SVM)for classification.The results reveal that the fusion of different feature vectors has improved the average accuracy of single feature recognition,which proves the rationality of the features proposed in our research.The KCCA fusion algorithm has the most obvious effect.At the same time,it shows that the feature extraction idea based on the combination of global and local features is correct.More importantly,for hand and limb movements,the experimental results prove that the two-layer network constructed by fusing the coherent features of brain and muscle can more effectively characterize the functional coupling relationship of cortex and muscle.This feature reflects the intrinsic physiological characteristics of the nerve-muscle and provides a new idea for the rehabilitation of patients with limb movement disorders.