Classification of Mild Cognitive Impairment and Normal by pattern recognition of EEG Lemple-Ziv complexity and alpha power

The paper describes Lemple-Ziv complexity(LZC) and alpha power were used to classify Normal and Mild Cognitive Impairment(MCI) using color task related EEG data . Thirty elders, including 9 patients with MCI and 21 normal controls, participated in our experiment. We recorded their EEG when they were judging whether the color of two graphics was matched or not. The 0-1 s,1 s-2 s and 2 s-3 s EEG data after the stimulus onset were extracted. Then features including LZC and alpha power were calculated in different encephalic regions as inputs to a support vector machine(SVM). Result show that medical-anterior has higher classification accuracy than other areas and LZC do better than alpha power. Typically, AFz lead has the best result in medical-anterior(85.7%) with LZC as feature.