A classification system to discriminate epileptic patients using multi-valued coarse-graining Lempel-Ziv complexity

A classification system to discriminate epileptic subjects from healthy subjects is proposed. All the previous research to detect epileptic seizures/patients using Lempel-Ziv complexity measure had used Binary Coarse-Graining (BLZC). In this work, we show that employing multi-valued coarse-graining Lempel-Ziv complexity (MLZC) improves the performance of classification. This finding is confirmed using Receiver Operating Characteristic (ROC) plots. Both the measures yielded excellent results with the MLZC showing improved results with, a sensitivity of 96.6%, specificity of 100%, precision of 100% and an accuracy of 98.3%. The classification system in this paper will be a valuable asset to the clinician in the separation of epileptic patients from the healthy group.