Spectral slope and Lempel-Ziv complexity as robust markers of brain states during sleep and wakefulness

Abstract

Spectral slope and neural complexity are affected in many neurophysiological disorders such as ADHD, autism or epilepsy and are modulated by sleep, anesthesia, and aging. Yet, the relationship between these two parameters remains unclear. Here, we evaluated the effects of sleep stage and task demands on spectral slope and neural complexity within a narrow- (30 - 45Hz) and broadband (3 - 45Hz) frequency range in 28 healthy male adults (21.54 ± 1.90 years) over three consecutive recordings with a set of different tasks (resting, attention and memory). We show that the slope steepens, and complexity decreases from wakefulness to N3 sleep. Importantly, slope and complexity are not only modulated by sleep but also differ between tasks, with flatter slopes and higher complexity being associated with more demanding tasks. While the slope and complexity are strongly correlated within 3 - 45Hz, we observe a functional dissociation in the 30 - 45Hz range. Critically, only the narrowband slope is steepest during REM sleep and associated with better task performance, especially in a Go/Nogo task. Our results demonstrate that both markers are powerful indices of sleep depth, task demand and cognitive performance. However, depending on the frequency range, they provide distinct information about the underlying brain state.

Neural complexity and the spectral slope characterise auditory processing in wakefulness and sleep

“…Additionally, as auditory stimulation was presented throughout overnight sleep, we cannot exclude the possibility that it may have induced changes in sleep architecture. However, our results in pre-auditory stimulation activity very closely mirror previous reports of complexity and spectral slope in sleep, in terms of the direction of effects and range of reported values (Andrillon et al, 2016;Höhn et al, 2022;Lendner et al, 2020;Miskovic et al, 2019;Schartner et al, 2017), suggesting that auditory stimulation did not change the fundamental properties of the sleep EEG signals that were studied here. To assess EEG responses to sounds, we applied a 40 Hz low-pass filter, which could bias the spectral slope towards overall steeper values by suppressing frequency content at the upper edge of the power spectra.…”

“…Similar to neural complexity, the spectral slope of the EEG power spectrum is also mostly evaluated in spontaneous activity and is found to be sensitive to changes in consciousness levels, such as those arising from sleep (Höhn et al, 2022;Kozhemiako et al, 2022;Lendner et al, 2020;Miskovic et al, 2019) and anaesthesia (Colombo et al, 2019(Colombo et al, , 2023Lendner et al, 2020). The spectral slope is computed in either low (generally below 20 Hz), or high frequencies (generally between 20 and 45 Hz) (Alnes et al, 2021;Colombo et al, 2019;Gyurkovics et al, 2022;Höhn et al, 2022;Kozhemiako et al, 2022;Lendner et al, 2020). During sleep, the spectral slope tends to be steeper than during wake irrespective of frequency range (Höhn et al, 2022;Lendner et al, 2020).…”

“…The spectral slope is computed in either low (generally below 20 Hz), or high frequencies (generally between 20 and 45 Hz) (Alnes et al, 2021;Colombo et al, 2019;Gyurkovics et al, 2022;Höhn et al, 2022;Kozhemiako et al, 2022;Lendner et al, 2020). During sleep, the spectral slope tends to be steeper than during wake irrespective of frequency range (Höhn et al, 2022;Lendner et al, 2020). In the higher frequency range, the decay of the spectral slope is steeper during REM compared to NREM sleep (Höhn et al, 2022;Kozhemiako et al, 2022;Lendner et al, 2020), whereas in lower frequencies (generally below 20 Hz), it is flatter in REM relative to NREM (Höhn et al, 2022;Miskovic et al, 2019).…”

Complexity and 1/f slope jointly reflect brain states

“…It changes depending on the behavioral state (60), task demands (60, 61), performance (28) and depending on circadian rhythms (62), which suggests that this property is under fine dynamic control. Recent work have shown brain states and neural complexity can be regulated by ascending arousal activity (63, 64), external stimulation (65) and task demand (66, 67). Future research could address this topic with a multiscale approach to the underlying states of neuromodulation-related psychiatric disorders (68).…”

Considerations towards a neurobiologically-informed EEG measurement of sleepiness

“…Experimental support has been found for a role of all canonical frequency bands of brain activity as modulated by, or markers of, sleepiness [32][33][34][35][36][37][38] . There have also been findings in support of other elements of the EEG in relation to sleepiness, such as event-related potential (ERP) component amplitude changes [39] , microstates and phase locking values between various regions/frequencies [40] , informational complexity markers [41,42] and aperiodic indices [43,44] . That is, sleepiness may be represented in a diffuse manner in the brain, at least in terms of objective measurement via the EEG.…”