Quantitative methods for detecting cerebral infarction from multiple channel EEG recordings

EEG has been known to be non-stationary and time varying. Time–frequency representation (TFR) is a proper tool for such non-stationary signals. In the present paper, TFR-based quantitative methods that can translate complicated and subjective waveform-based EEG analysis into objective measures are introduced to characterize EEG recorded from normal subjects and cerebral infarction (CI) patients. Relative frequency band energy (RFBE) is computed from time–frequency plane for the five subbands: delta, theta, alpha, beta and gamma. Moreover, we propose the Shannon entropy (SE) of TFR to detect the difference in EEG for the two kinds of subjects. Finally, the temporal evolutions of these quantitative parameters are presented to trace EEG changes. The experiment results show that CI results in the RFBE changes of the five rhythms; however, the RFBEs of some rhythms have stronger association with CI. Increase in EEG SE of CI patients is obvious. The time evolutions of RFBE and SE as valuable objective measures can be displayed in real time and be used as helpful references in detection and monitoring of CI.