Table 3.

Spectral power density in characteristic EEG frequency bands computed with the standard approach and the automated pipeline.

	Band limits (Hz)	Power density (Standard) (log₁₀(uV²/Hz)) mean (std)	Power density (Automated) (log₁₀(uV²/Hz)) mean (std)	Absolute Difference (log₁₀(uV²/Hz)) mean (std)	%-Difference mean (std)	Spearman r
NREM
Slow Osc.	0.25–1	2.34 (0.18)	2.36 (0.19)	0.02 (0.02)	0.65 (0.63)	0.99
Delta	1.25–4	1.35 (0.17)	1.35 (0.17)	0.01 (0.01)	0.46 (0.65)	1.00
Theta	4.25–8	0.87 (0.21)	0.87 (0.21)	0.00 (0.01)	0.47 (0.71)	1.00
Alpha	8.25–12	0.55 (0.22)	0.56 (0.22)	0.01 (0.01)	1.74 (3.41)	1.00
Sigma	12.25–15	0.16 (0.21)	0.17 (0.21)	0.01 (0.01)	6.73 (12.16)	1.00
Beta	15.25–20	−0.27 (0.19)	−0.26 (0.19)	0.01 (0.02)	10.10 (25.04)	1.00
REM
Slow Osc.	0.25–1	1.99 (0.23)	2.03 (0.25)	0.04 (0.05)	1.85 (1.69)	0.99
Delta	1.25–4	0.95 (0.17)	0.96 (0.17)	0.02 (0.04)	2.09 (3.10)	0.99
Theta	4.25–8	0.61 (0.22)	0.62 (0.22)	0.01 (0.02)	2.03 (5.83)	0.99
Alpha	8.25–12	0.41 (0.26)	0.41 (0.26)	0.01 (0.01)	4.12 (12.44)	1.00
Sigma	12.25–15	0.07 (0.25)	0.07 (0.26)	0.01 (0.01)	7.97 (18.24)	1.00
Beta	15.25–20	−0.19 (0.24)	−0.19 (0.24)	0.01 (0.01)	7.58 (18.25)	1.00

Differences between average spectral power density values computed with the two approaches in REM and NREM are all not significant (p > 0.05). All correlations are statistically significant (p <0.01). In the standard approach, spectra were calculated for consecutive 30-s epochs after removing 4-s sub-epochs with visually identified artifacts and then averaging the remaining artifact free 4-s spectra from up to ten overlapping 4-s sub-epochs. In the automated approach, epochs with automatically scored artifact were excluded from the analysis. It must be noted that activity in the slow oscillation band (0.25–1 Hz) is often due to sweating artifact.