Classification of Stroke Disease Assessment based on Body Surface Electrical Signals at Acupuncture Points

Abstract

Stroke is a common brain disease with high sudden onset, high lethality, and a single means of rehabilitation assessment. In this paper, we designed a novel stroke condition rehabilitation classification experiment using a 4-channel body surface electrical signal acquisition circuit independently developed by the Integrated Microsystems Laboratory of Peking University to acquire body surface electrical signals from acupoints on both sides of the body meridians of 8 healthy volunteers and 16 stroke patients with different degrees of rehabilitation. The signals were filtered and wavelet transformed to extract features, which were fed into LR, RR, XGBoost, and SVM models for condition assessment and classification (types: healthy, mild hemiplegia, and severe hemiplegia). The results showed that the experimental scheme had strong classification ability for stroke condition recognition, with F1-score of 0.75 for LR, 0.72 for RR, 0.81 for XGBoost, and 0.79 for SVM. Furthermore, the “anti-interference ability of 50Hz” feature was separately introduced for classification, and a better classification effect was obtained. Among them, the F1 score of stroke evaluation and classification based on the SVM model increased by 2.65%, which is helpful to realize the intelligent prediction of stroke disease in clinical practice Diagnostic function.