Multiresolution analysis on nonlinear complexity measurement of EEG signal for epileptic discharge monitoring

Various epileptic discharge detection studies have been conducted however, not many clinically significant outcomes have been achieved in developing reliable algorithm using nonlinear measurement techniques. Study has reported that some of the linear measurement techniques performances better than nonlinear technique in detecting the epileptic discharge in terms of accuracy, sensitivity and specificity. The reliability issue has been addressed in nonlinear techniques, by introducing Multiresolution analysis (MRA). MRA with approximate entropy are the most common combination used to detect epileptic discharge, leaving other nonlinear complexity measures are yet to be explored with MRA. Previously, we have study the performance of Approximate Entropy (ApEn) and Lempel Ziv (LZ) using MRA. In this paper, we have expanded the scope by studying performance of MRA with other complexity measurement including Hurst exponent (HE), Kolmogorov complexity (KC), Shannon Entropy (SE) and Sample Entropy (SampEn). Groups of normal with interictal (Set A and B), normal with ictal (Set A and C) and interictal with ictal (Set B and C) were used to evaluate the performance. For the result, MRA managed to enhance the accuracy of ApEn (AB: 74% to 89%, AC: 98% to 100%, BC:88% to 94%) and SE (AB: 69% to 98%, AC: 100% to 100%, BC:96% to 97%) in detecting epileptic discharge the best while deteriorating LZ (AB: 49% to 83%, AC: 91% to 86%, BC:89% to 88%) and HE (AB: 65% to 70%, AC: 89% to 78%, BC:80% to 54%) performance. Computation time tends to increase with the implementation of MRA