Analysis of Electroencephalogram Microstate Characteristics in Acute Ischemic Stroke Patients with Wind-phlegm and Obstructing the Collaterals Syndrome Based on Machine Learning
ZHANG Jiacheng
SUN Jing
HUANG Xing
WANG Peng
GUO Yu
LI Ying
ZHANG Zihan
BAO Weiwei
LI Xiangyu
CHANG Jingling
Abstract：Objective:To explore the traditional Chinese medicine syndrome differentiation model of wind-phlegm and obstructing the collaterals syndrome in acute ischemic stroke(AIS)by machine learning algorithms,to provide early detection and auxiliary diagnosis for the traditional Chinese medicine syndrome differentiation and treatment of AIS.Methods:A total of 18 patients with wind-phlegm and obstructing the collaterals syndrome in AIS and 24 patients without wind-phlegm and obstructing the collaterals syndrome were included.The K-means clustering algorithm was used to analyze the 64-lead electroencephalogram(EEG)data of the two groups in the resting state after preprocessing,and the time parameters(duration,occurrence frequency,coverage rate)of the four types of EEG microstates(A,B,C,D)were obtained and analyzed statistically.On this basis,the back propagation(BP)neural network,K-nearest neighbor(KNN),support vector machine(SVM),and random forest(RF)were adopted to conduct machine learning modeling for patients with wind-phlegm and obstructing the collaterals syndrome of AIS.The model was trained using the one-retention method,and the classification performance of the machine learning was evaluated respectively by accuracy rate,precision rate,recall rate,harmonic mean,receiver operating characteristic(ROC)curve and area under the Roc curve AUC.Results:The occurrence frequency and coverage rate of microstate B in patients with wind-phlegm and obstructing the collaterals syndrome of AIS significantly decreased than those in patients without wind-phlegm and obstructing the collaterals syndrome.All relevant classifiers could achieve the construction of supervised binary classification machine learning models for wind-phlegm obstruction networks and non-wind-phlegm obstruction networks.The BP neural network performed exceptionally well in terms of accuracy rate,precision rate,recall rate and harmonic mean,which were 92.9%,89.5%,94.4%,and 91.9%respectively.The KNN classifier played better in sample recognition and sensitivity,while the AUC was 0.919.Conclusion:The construction of a syndrome differentiation model for wind-phlegm and obstructing the collaterals syndrome of AIS based on machine learning was feasible,which was conducive to the clinical differentiation and classification of AIS and provided support for clinicians to make targeted intervention decisions.
Keywords：acute ischemic strokewind-phlegm and obstructing the collaterals syndromemicrostatemachine learningtraditional Chinese medicine syndrome differentiation model
Publication Date：2025-11-25
Online Publishing Date：2025-12-05(First online date of this platform, not the publication date of the document)
Pages：8( 3397-3404 )