Stroke Severity Classification based on EEG Statistical Features

Abstract

Stroke is one of the leading causes of death and disability in the world. Therefore, it is necessary to diagnose stroke at an early stage and provide an accurate prognostic assessment. This study attempts to classify the severity of stroke based on EEG signals by using statistical parameters of time domain features. The results of this study are expected to diagnose the severity of stroke from the parameters used in the time domain and make decisions about the next treatment steps. In this study, the EEG data was obtained from measurement to stroke patients in public hospital in the city of Kediri. From the EEG, 3 statistical features such as Mean Absolute Value (MA V), Standard Deviation (STD) and Variance were calculated. Stroke severity classes were defined as severe, moderate, and mild. The analyzed EEG frequency sub-bands were Alpha Low (8–9 Hz), Alpha High (9–13 Hz), Beta Low (13–17), and Beta High (17–30 Hz). The label for stroke severity classification as a ground truth uses the NIHSS scale which is assessed by doctors based on visual observations. The results showed that stroke severity classification can be identified by using statistical feature such as MA V, STD and Variance, with EEG sub-band frequency are Alpha Low and Alpha High for grasp motion, Beta Low and Beta High for Elbow motion, and Alpha High and Beta High for shoulder motion. This result showed the potential of using this information as a basic for determining the patient-specific rehabilitation program in the future.