Assessing stroke rehabilitation degree based on quantitative EEG index and nonlinear parameters

The assessment of motor function is critical to the rehabilitation of stroke patients. However, commonly used evaluation methods are based on behavior scoring, which lacks neurological indicators that directly reflect the motor function of the brain. The objective of this study was to investigate whether resting-state EEG indicators could improve stroke rehabilitation evaluation. We recruited 68 participants and recorded their resting-state EEG data. According to Brunnstrom stage, the participants were divided into three groups: severe, moderate, and mild. Ten quantitative electroencephalographic (QEEG) and five non-linear parameters of resting-state EEG were calculated for further analysis. Statistical tests were performed, and the genetic algorithm-support vector machine was used to select the best feature combination for classification. We found the QEEG parameters show significant differences in Delta, Alpha1, Alpha2, DAR, and DTABR (P < 0.05) among the three groups. Regarding nonlinear parameters, ApEn, SampEn, Lz, and C₀ showed significant differences (P < 0.05). The optimal feature classification combination accuracy rate reached 85.3%. Our research shows that resting-state EEG indicators could be used for stroke rehabilitation evaluation.