Ictal EEG Non-linear And High Order Spectral Analysis Methods In Electroconvulsive Therapy And Its Clinical Utility.

ABSTRACT

Introduction: Analysis of ictal electroencephalogram (EEG) during electroconvulsive therapy (ECT) early in the course provides information to predict clinical outcome. Previous studies used visual, power spectral and non linear methods to analyze EEG. EEG signals are non stationary, non linear, non Gaussian and chaotic in nature. Such signals can be better characterized by non linear and higher order spectrum analysis. However there is scarcity of data assessing such measurers in predicting clinical outcome. We conducted nonlinear and high order spectrum analyses of ictal EEG recorded during ECT and correlated the measures with clinical outcome. Methods: Schizophrenia patients receiving ECT were assessed using the brief psychiatric rating scale (BPRS) before and 2 weeks after the start of ECT. EEG was recorded during seizure from left frontal-pole (FP1) channel. In 26 patients, completely artifact-free EEG was available. Approximate entropy (ApEn), Sample entropy (SamEn), Hurst exponent (H), Bispectrum entropy (HOS.En), correlation dimension (CD) and Largest Lyapunov exponent (LLE) were computed for EEG from the earliest ECT session (2nd or 3rd). Results: HOS.En emerged as a significant measure which predicted outcome at two weeks (HOS En1 r = - 0.434; p = 0.027 & En2 r = -0.414; 0.036) other measures, viz., ApEn (r = -0.001; p = 0.995), SampEn (r = - 0.152; p = 0.458), H (r = 0.123; p = 0.549), CD (r = 0.119; p = 0.563) and LLE (r = -0.293; p = 0.146) did not predict the outcome. Conclusion: In patients with schizophrenia receiving ECT higher bispectrum entropy of ictal EEG early in the ECT course predicts better clinical outcome at the end of two weeks. None of the other non linear measures evaluated in the study predicted clinical outcome.