Journal: eLife

Article Title: An electrophysiological marker of arousal level in humans

doi: 10.7554/eLife.55092

Figure Lengend Snippet: ( a ) Single subject example: Upper panel: Hypnogram. Wake periods are highlighted in pink, REM periods in light green. Upper middle panel: Multitapered spectrogram of electrode Fz. Lower middle panel: Number of slow wave (SO) events during 30 s segments of sleep in electrode Fz. Note the decreasing number of SO events during the course of the night. Lower panel: Spectral slope of SO events occurring in N3 (blue), wakefulness (red) and REM sleep (green) in electrode Fz. Background: Time-resolved slope of electrode Fz in light gray. ( b ) Right panel: Average spectral slope changes over the time course of all slow waves in scalp EEG (n = 20) during sleep (blue; mean ± SEM); superimposed in red is the average slow wave of all subjects. Highlighted are the following 0.5 s time windows relative to the slow wave trough: −750 to −250 (center −0.5 s; green), −250 to 250 (center 0 s; pink) and 250 to 750 ms (center 0.5 s; purple). Left panel: Power spectra in log-log space within specified time windows during the slow wave: −750 to −250 (center: −0.5 s; green), −250 to 250 (center: 0 s; pink) and 250 to 750 ms (center: 0.5 s; purple). Note the steep power decrease during the trough of the slow wave (pink). ( c ) Group level (n = 20) average waveforms in electrode Fz during N3 (blue), REM sleep (green) and wakefulness (red; mean ± SEM). ( d ) Left: Slow wave events per minute in wakefulness (red), N3 (blue) and REM (green) in scalp EEG channel FZ (n = 20). In black mean ± SEM. Permutation t-tests : ***p<0.001. Right: Slope of slow wave events on the group level (n = 20; averaged across all 19 EEG electrodes) in wakefulness (red), N3 (blue) and REM sleep (green). Mean ± SEM in black. Permutation t-tests : ***p<0.001.

Article Snippet: To compute a time resolved estimate of the spectral slope, we calculated the best line fit to the 10 (anesthesia) or 30 (sleep) second segments of the Multitapered power spectra (see above) in log-log space using polynomial curve fitting ( polyfit.m, MATLAB and Curve Fitting Toolbox Release R2015a, The MathWorks, Inc, Natick, Massachusetts, United States).

Techniques:

Journal: eLife

Article Title: An electrophysiological marker of arousal level in humans

doi: 10.7554/eLife.55092

Figure Lengend Snippet: ( a ) Single subject example at EEG electrode Fz during induction of propofol anesthesia. Left panels – Original observation: 10 s segments (nine discrete prolate spheroidal sequences (dpss) tapers). Right panels – New observation: 30 s segments (29 dpss tapers). Upper panels - Multitapered (MT) time-frequency decomposition. Middle panels – Time-resolved spectral slope (gray; mean ± SEM). Lower panels – Slow oscillation (<1.25 Hz; gray) and alpha power (8–12 Hz; purple; mean ± SEM). Dotted lines: Induction with propofol. Note the temporal smoothing with 30 s compared to 10 s segments. MT power spectra in log-log space (left panels; mean ± SEM) and 1/f slope (right panels; black - mean ± SEM) in wakefulness (Wake - red) and under general anesthesia with propofol (Ana - blue) in scalp EEG (n = 9; averaged across all channels). ( b ) Original observation: Power spectra and slopes calculated in 10 s segments (blue; nine dpss tapers). ***p<0.001. ( c ) New observation: Power spectra and slopes calculated in 30 s segments (orange; 29 dpss tapers). ***p<0.001. ( d ) At electrode Fz, the 1/f slope values derived from 10 and 30 s segments in anesthesia scalp EEG are strongly correlated (n = 9; r = 1.00, p<0.0001).

Article Snippet: To compute a time resolved estimate of the spectral slope, we calculated the best line fit to the 10 (anesthesia) or 30 (sleep) second segments of the Multitapered power spectra (see above) in log-log space using polynomial curve fitting ( polyfit.m, MATLAB and Curve Fitting Toolbox Release R2015a, The MathWorks, Inc, Natick, Massachusetts, United States).

Techniques: Derivative Assay

Journal: eLife

Article Title: An electrophysiological marker of arousal level in humans

doi: 10.7554/eLife.55092

Figure Lengend Snippet: ( a ) Single subject example at EEG electrode Fz during a full night of sleep. Upper panel – Technician scored hypnogram. Wake periods highlighted in pink, REM in green. Middle Panel - Original observation: Multitapered (MT) time-frequency decomposition with 30 s segments (29 dpss tapers). Lower panel – New observation: MT time-frequency decomposition with 10 s segments (nine dpss tapers). MT Power spectra in log-log space (left panels; mean ± SEM) and 1/f slope (right panels; black - mean ± SEM) during baseline rest (BL - magenta), in NREM sleep stage 3 (N3 - blue) and REM sleep (REM - green) in scalp EEG (n = 14; averaged across all channels). ( b ) Original observation: Power spectra and slopes calculated in 30 s segments (orange; 29 dpss tapers). ***p<0.001. ( c ) New observation: Power spectra and slopes calculated in 10 s segments (blue; nine dpss tapers). ***p<0.001. ( d ) At electrode Fz, the 1/f slopes derived from 30 and 10 s segments in sleep scalp EEG are strongly correlated (n = 14; r = 0.95, p<0.0001).

Article Snippet: To compute a time resolved estimate of the spectral slope, we calculated the best line fit to the 10 (anesthesia) or 30 (sleep) second segments of the Multitapered power spectra (see above) in log-log space using polynomial curve fitting ( polyfit.m, MATLAB and Curve Fitting Toolbox Release R2015a, The MathWorks, Inc, Natick, Massachusetts, United States).

Techniques: Derivative Assay

Journal: eLife

Article Title: An electrophysiological marker of arousal level in humans

doi: 10.7554/eLife.55092

Figure Lengend Snippet: ( a ) Single subject example of time-resolved average at EEG electrode Fz during a full night of sleep. Upper panel: Expert-scored hypnogram (black), wake (pink), REM (light green). Multitapered time-frequency decomposition using: Upper middle panel – the original reference of linked bilateral mastoids (orange). Middle panel - common average reference (CAR, light blue). Middle lower panel - clinical bipolar reference (‘double banana’ (DB), purple). Lower panel - Laplacian reference (LP, green). ( b ) 1/f spectral slope in baseline rest (BL, magenta), NREM sleep stage 3 (N3, blue) and REM sleep (REM, green) using the original (Orig., orange), a CAR (light blue), DB (purple) or LP (green) reference in scalp EEG during sleep (n = 14, averaged across electrodes). Paired t-tests (uncorrected): Orig.-CAR : p=0.065, t 13 = 2.02, d = 0.36; Orig.-DB : p=0.0875, t 13 = −1.85, d = −0.32, Orig.-LP : p=0.005, t 13 = −3.35, d = −0.64. n.s. – not significant. **p<0.01. ( c ) At electrode Fz, 1/f slope values derived from the original (Orig., orange) and the other reference schemes are strongly correlated (n = 14). Upper panel - CAR reference (light blue; r = 0.90, p>0.001), middle panel - DB reference (purple; r = 0.77, p<0.001), lower panel – LP reference (green; r = 0.80, p<0.001). ( d ) Mutual Information (MI) between hypnogram and 1/f slope derived from power spectra obtained by different reference schemes at electrode Fz (n = 14): Orig. (orange): 0.22 ± 0.04 (black - mean ± SEM), CAR (light blue): 0.22 ± 0.03, DB (purple): 0.15 ± 0.03, LP (green): 0.18 ± 0.03. Paired t-tests (uncorrected) : MI Orig.-CAR : p=0.94, t 13 = −0.08, d = −0.01; Orig.-DB : p=0.007, t 13 = 3.23, d = 0.62, Orig.-LP : p=0.109, t 13 = 1.72, d = 0.33. n.s. – not significant. **p<0.01.

Article Snippet: To compute a time resolved estimate of the spectral slope, we calculated the best line fit to the 10 (anesthesia) or 30 (sleep) second segments of the Multitapered power spectra (see above) in log-log space using polynomial curve fitting ( polyfit.m, MATLAB and Curve Fitting Toolbox Release R2015a, The MathWorks, Inc, Natick, Massachusetts, United States).

Techniques: Derivative Assay