Novel Electrophysiological Signatures of Learning and Forgetting in Human Rapid Eye Movement Sleep.

Despite the known behavioral benefits of rapid eye movement (REM) sleep, discrete neural oscillatory events in human scalp electroencephalography (EEG) linked with behavior have not been discovered. This knowledge gap hinders mechanistic understanding of the function of sleep, as well as the development of biophysical models and REM-based causal interventions. We designed a detection algorithm to identify bursts of activity in high-density, scalp EEG within theta (4-8 Hz) and alpha (8-13 Hz) bands during REM sleep. Across 38 nights of sleep, we characterized the burst events (i.e., count, duration, density, peak frequency, amplitude) in healthy, young male and female human participants (38; 21F) and investigated burst activity in relation to sleep-dependent memory tasks: hippocampal-dependent episodic verbal memory and nonhippocampal visual perceptual learning. We found greater burst count during the more REM-intensive second half of the night (p < 0.05), longer burst duration during the first half of the night (p < 0.05), but no differences across the night in density or power (p > 0.05). Moreover, increased alpha burst power was associated with increased overnight forgetting for episodic memory (p < 0.05). Furthermore, we show that increased REM theta burst activity in retinotopically specific regions was associated with better visual perceptual performance. Our work provides a critical bridge between discrete REM sleep events in human scalp EEG that support cognitive processes and the identification of similar activity patterns in animal models that allow for further mechanistic characterization.

A Multi-Level Interpretable Sleep Stage Scoring System by Infusing Experts' Knowledge Into a Deep Network Architecture.

In recent years, deep learning has shown potential and efficiency in a wide area including computer vision, image and signal processing. Yet, translational challenges remain for user applications due to a lack of interpretability of algorithmic decisions and results. This black box problem is particularly problematic for high-risk applications such as medical-related decision-making. The current study goal was to design an interpretable deep learning system for time series classification of electroencephalogram (EEG) for sleep stage scoring as a step toward designing a transparent system. We have developed an interpretable deep neural network that includes a kernel-based layer guided by a set of principles used for sleep scoring by human experts in the visual analysis of polysomnographic records. A kernel-based convolutional layer was defined and used as the first layer of the system and made available for user interpretation. The trained system and its results were interpreted in four levels from microstructure of EEG signals, such as trained kernels and effect of each kernel on the detected stages, to macrostructures, such as transitions between stages. The proposed system demonstrated greater performance than prior studies and the system learned information consistent with expert knowledge.

Medial temporal lobe functional network architecture supports sleep-related emotional memory processing in older adults.

Memory consolidation occurs via reactivation of a hippocampal index during non-rapid eye movement slow-wave sleep (NREM SWS) which binds attributes of an experience existing within cortical modules. For memories containing emotional content, hippocampal-amygdala dynamics facilitate consolidation over a sleep bout. This study tested if modularity and centrality-graph theoretical measures that index the level of segregation/integration in a system and the relative import of its nodes-map onto central tenets of memory consolidation theory and sleep-related processing. Findings indicate that greater network integration is tied to overnight emotional memory retention via NREM SWS expression. Greater hippocampal and amygdala influence over network organization supports emotional memory retention, and hippocampal or amygdala control over information flow are differentially associated with distinct stages of memory processing. These centrality measures are also tied to the local expression and coupling of key sleep oscillations tied to sleep-dependent memory consolidation. These findings suggest that measures of intrinsic network connectivity may predict the capacity of brain functional networks to acquire, consolidate, and retrieve emotional memories.

Slow oscillations promote long-range effective communication: The key for memory consolidation in a broken-down network.

A prominent and robust finding in cognitive neuroscience is the strengthening of memories during nonrapid eye movement (NREM) sleep, with slow oscillations (SOs;<1Hz) playing a critical role in systems-level consolidation. However, NREM generally shows a breakdown in connectivity and reduction of synaptic plasticity with increasing depth: a brain state seemingly unfavorable to memory consolidation. Here, we present an approach to address this apparent paradox that leverages an event-related causality measure to estimate directional information flow during NREM in epochs with and without SOs. Our results confirm that NREM is generally a state of dampened neural communication but reveals that SOs provide two windows of enhanced large-scale communication before and after the SO trough. These peaks in communication are significantly higher when SOs are coupled with sleep spindles compared with uncoupled SOs. To probe the functional relevance of these SO-selective peaks of information flow, we tested the temporal and topographic conditions that predict overnight episodic memory improvement. Our results show that global, long-range communication during SOs promotes sleep-dependent systems consolidation of episodic memories. A significant correlation between peaks of information flow and memory improvement lends predictive validity to our measurements of effective connectivity. In other words, we were able to predict memory improvement based on independent electrophysiological observations during sleep. This work introduces a noninvasive approach to understanding information processing during sleep and provides a mechanism for how systems-level brain communication can occur during an otherwise low connectivity sleep state. In short, SOs are a gating mechanism for large-scale neural communication, a necessary substrate for systems consolidation and long-term memory formation.

Competitive dynamics underlie cognitive improvements during sleep.

We provide evidence that human sleep is a competitive arena in which cognitive domains vie for limited resources. Using pharmacology and effective connectivity analysis, we demonstrate that long-term memory and working memory are served by distinct offline neural mechanisms that are mutually antagonistic. Specifically, we administered zolpidem to increase central sigma activity and demonstrated targeted suppression of autonomic vagal activity. With effective connectivity, we determined the central activity has greater causal influence over autonomic activity, and the magnitude of this influence during sleep produced a behavioral trade-off between offline long-term and working memory processing. These findings suggest a sleep switch mechanism that toggles between central sigma-dependent long-term memory and autonomic vagal-dependent working memory processing.

Performance analysis of EEG seizure detection features.

Automatic detection of epileptic seizures can serve as a valuable clinical tool which involves a more objective and computationally efficient method for the analysis of EEG data in order to generate increasingly accurate and reliable results. Automatic seizure detection is also an important component of closed-loop responsive cortical stimulation systems. The goal of this study is to evaluate EEG-based features recently proposed for seizure detection to discover the optimum ones for a reliable seizure detection system. We extracted seizure detection features from intracranial EEG signals that were recorded during invasive pre-surgical epilepsy monitoring of people with drug resistant focal epilepsy at the Epilepsy Center of the University Hospital of Freiburg. Features from time, frequency and phase space domains as well as similarity/dissimilarity features were considered. The performance of each feature was investigated using the statistical test ANOVA. Performance analysis was conducted separately on the recordings from the channels within the seizure-onset zone (SOZ-in) and the recordings from the channels outside the seizure-onset zone (SOZ-out). Similarity/dissimilarity features that measure dynamic properties of the EEG signal and the evolving phenomena of the seizures could significantly separate ictal (during seizure) states from pre-ictal (before seizure) states (p < 0.01). Among them, our proposed feature, Bhattacharyya-based dissimilarity index (BBDI), successfully passed Tukey's post-hoc test as well suggesting that it can distinguish both pre-ictal and post-ictal (after seizure) periods from ictal period. BBDI was further applied to detect epileptic seizures and achieved area under the curve of the receiver-operator characteristic (ROC) equal to 0.96 and 0.94 for SOZ-in and SOZ-out channels, respectively. No significant difference (p = 0.59) was observed in the performance of features between SOZ-in recordings and SOZ-out recordings. The discriminative value of EEG seizure detection features was determined by statistical tests. As a result, the best features to be selected for a reliable seizure detection system designed for people with drug-resistant focal epilepsy were suggested, which include similarity/dissimilarity indices.