Multifractal Analysis of Sleep EEG Dynamics in Humans

Abstract

The aim of this study is to investigate the possibility that human sleep EEGs can be characterized by a multifractal spectrum using wavelet transform modulus maxima (WTMM). We used sleep EEGs taken from healthy subjects during the four stages of sleep and REM sleep. Our findings showed that the dynamics in human sleep EEGs could be adequately described by a set of scales and characterized by multifractals. We performed multivariate discriminate analysis to evaluate the use of multifractal features for classification. The multivariate discriminate analysis using within-groups covariance matrices for all sleep stages yielded a total error rate of 41.8%. In conclusion, multifractal formalism, based on the WTMM, appears to be a good tool for characterizing dynamics in sleep EEGs