Sample entropy tracks changes in electroencephalogram power spectrum with sleep state and aging.

The regularity of EEG signals was compared between middle-aged (47.2 ± 2.0 yrs) and elderly (78.4 ± 3.8 yrs) female subjects in Wake (W), NREM stages 2 and 3 (S2, S3), and REM. Signals from C3A2 leads of healthy normal subjects, acquired from polysomnograms obtained from the Sleep Heart Health Study, were analyzed using both Sample Entropy (SaEn) and power spectral analysis (delta, theta, alpha, and beta frequency band powers). SaEn changed systematically and significantly (p REM > S2 > S3. SaEn was found to be negatively correlated with delta power and positively correlated with beta power. Small changes in SaEn appear to reflect changes in spectral content rather than changes in regularity of the signal. A better predictor of SaEn than the frequency band powers was the logarithm of the power ratio (alpha+beta)/(delta+theta). Thus, SaEn appears to reflect the balance between sleep-promoting and alertness-promoting mechanisms. SaEn of the elderly was larger than that of middle-aged subjects in S2 (p=0.029) and REM (p=0.001), suggesting that cortical state is shifted towards alertness in elderly subjects in these sleep states compared to middle-aged.