Robust, automated sleep scoring by a compact neural network with distributional shift correction

Workflow for the unsupervised clustering of sleep stages identifies light and deep sleep in electrophysiological recordings in mice

BACKGROUND

Sleep physiology plays a critical role in brain development and aging. Accurate sleep staging, which categorizes different sleep states, is fundamental for sleep physiology studies. Traditional methods for sleep staging rely on manual, rule-based scoring techniques, which limit their accuracy and adaptability.

NEW METHOD

We describe, test and challenge a workflow for unsupervised clustering of sleep states (WUCSS) in rodents, which uses accelerometer and electrophysiological data to classify different sleep states. WUCSS utilizes unsupervised clustering to identify sleep states using six features, extracted from 4-second epochs.

RESULTS

We gathered high-quality EEG recordings combined with accelerometer data in diverse transgenic mouse lines (male ApoE3 versus ApoE4 knockin; male CNTNAP2 KO versus wildtype littermates). WUCSS showed high recall, precision, and F1-score against manual scoring on awake, NREM, and REM sleep states. Within NREM, WUCSS consistently identified two additional clusters that qualify as deep and light sleep states.

COMPARISON WITH EXISTING METHODS

The ability of WUCSS to discriminate between deep and light sleep enhanced the precision and comprehensiveness of the current mouse sleep physiology studies. This differentiation led to the discovery of an additional sleep phenotype, notably in CNTNAP2 KO mice, showcasing the method's superiority over traditional scoring methods.

CONCLUSIONS

WUCSS, with its unsupervised approach and classification of deep and light sleep states, provides an unbiased opportunity for researchers to enhance their understanding of sleep physiology. Its high accuracy, adaptability, and ability to save time and resources make it a valuable tool for improving sleep staging in both clinical and preclinical research.

Restoration of locus coeruleus noradrenergic transmission during sleep

Sleep is indispensable for health and wellbeing, but its basic function remains elusive. The locus coeruleus (LC) powerfully promotes arousal by releasing noradrenaline. We found that noradrenaline transmission is reduced by prolonged wakefulness and restored during sleep. Fiber-photometry imaging of noradrenaline using its biosensor showed that its release evoked by optogenetic LC neuron activation was strongly attenuated by three hours of sleep deprivation and restored during subsequent sleep. This is accompanied by the reduction and recovery of the wake-promoting effect of the LC neurons. The reduction of both LC evoked noradrenaline release and wake-inducing potency is activity dependent, and the rate of noradrenaline transmission recovery depends on mammalian target of rapamycin (mTOR) signaling. The decline and recovery of noradrenaline transmission also occur in spontaneous sleep-wake cycles on a timescale of minutes. Together, these results reveal an essential role of sleep in restoring transmission of a key arousal-promoting neuromodulator.

Monitoring norepinephrine release in vivo using next-generation GRABNE sensors

Norepinephrine (NE) is an essential biogenic monoamine neurotransmitter. The first-generation NE sensor makes in vivo, real-time, cell-type-specific and region-specific NE detection possible, but its low NE sensitivity limits its utility. Here, we developed the second-generation GPCR-activation-based NE sensors (GRABNE2m and GRABNE2h) with a superior response and high sensitivity and selectivity to NE both in vitro and in vivo. Notably, these sensors can detect NE release triggered by either optogenetic or behavioral stimuli in freely moving mice, producing robust signals in the locus coeruleus and hypothalamus. With the development of a novel transgenic mouse line, we recorded both NE release and calcium dynamics with dual-color fiber photometry throughout the sleep-wake cycle; moreover, dual-color mesoscopic imaging revealed cell-type-specific spatiotemporal dynamics of NE and calcium during sensory processing and locomotion. Thus, these new GRABNE sensors are valuable tools for monitoring the precise spatiotemporal release of NE in vivo, providing new insights into the physiological and pathophysiological roles of NE.

Sleep-Deep-Learner is taught sleep-wake scoring by the end-user to complete each record in their style

Sleep-wake scoring is a time-consuming, tedious but essential component of clinical and preclinical sleep research. Sleep scoring is even more laborious and challenging in rodents due to the smaller EEG amplitude differences between states and the rapid state transitions which necessitate scoring in shorter epochs. Although many automated rodent sleep scoring methods exist, they do not perform as well when scoring new datasets, especially those which involve changes in the EEG/EMG profile. Thus, manual scoring by expert scorers remains the gold standard. Here we take a different approach to this problem by using a neural network to accelerate the scoring of expert scorers. Sleep-Deep-Learner creates a bespoke deep convolution neural network model for individual electroencephalographic or local-field-potential (LFP) records via transfer learning of GoogLeNet, by learning from a small subset of manual scores of each EEG/LFP record as provided by the end-user. Sleep-Deep-Learner then automates scoring of the remainder of the EEG/LFP record. A novel REM sleep scoring correction procedure further enhanced accuracy. Sleep-Deep-Learner reliably scores EEG and LFP data and retains sleep-wake architecture in wild-type mice, in sleep induced by the hypnotic zolpidem, in a mouse model of Alzheimer's disease and in a genetic knock-down study, when compared to manual scoring. Sleep-Deep-Learner reduced manual scoring time to 1/12. Since Sleep-Deep-Learner uses transfer learning on each independent recording, it is not biased by previously scored existing datasets. Thus, we find Sleep-Deep-Learner performs well when used on signals altered by a drug, disease model, or genetic modification.

Improved green and red GRAB sensors for monitoring spatiotemporal serotonin release in vivo

The serotonergic system plays important roles in both physiological and pathological processes, and is a therapeutic target for many psychiatric disorders. Although several genetically encoded GFP-based serotonin (5-HT) sensors were recently developed, their sensitivities and spectral profiles are relatively limited. To overcome these limitations, we optimized green fluorescent G-protein-coupled receptor (GPCR)-activation-based 5-HT (GRAB5-HT) sensors and developed a red fluorescent GRAB5-HT sensor. These sensors exhibit excellent cell surface trafficking and high specificity, sensitivity and spatiotemporal resolution, making them suitable for monitoring 5-HT dynamics in vivo. Besides recording subcortical 5-HT release in freely moving mice, we observed both uniform and gradient 5-HT release in the mouse dorsal cortex with mesoscopic imaging. Finally, we performed dual-color imaging and observed seizure-induced waves of 5-HT release throughout the cortex following calcium and endocannabinoid waves. In summary, these 5-HT sensors can offer valuable insights regarding the serotonergic system in both health and disease.

A role for δ subunit-containing GABA A receptors on parvalbumin positive neurons in maintaining electrocortical signatures of sleep states

GABA A receptors containing δ subunits have been shown to mediate tonic/slow inhibition in the CNS. These receptors are typically found extrasynaptically and are activated by relatively low levels of ambient GABA in the extracellular space. In the mouse neocortex, δ subunits are expressed on the surface of some pyramidal cells as well as on parvalbumin positive (PV+) interneurons. An important function of PV+ interneurons is the organization of coordinated network activity that can be measured by EEG; however, it remains unclear what role tonic/slow inhibitory control of PV+ neurons may play in shaping oscillatory activity. After confirming a loss of functional δ mediated tonic currents in PV cells in cortical slices from mice lacking Gabrd in PV+ neurons (PV δcKO), we performed EEG recordings to survey network activity across wake and sleep states. PV δcKO mice showed altered spectral content of EEG during NREM and REM sleep that was a result of increased oscillatory activity in NREM and the emergence of transient high amplitude bursts of theta frequency activity during REM. Viral reintroduction of Gabrd to PV+ interneurons in PV δcKO mice rescued REM EEG phenotypes, supporting an important role for δ subunit mediated inhibition of PV+ interneurons for maintaining normal REM cortical oscillations.

Significance statement

The impact on cortical EEG of inhibition on PV+ neurons was studied by deleting a GABA A receptor subunit selectively from these neurons. We discovered unexpected changes at low frequencies during sleep that were rescued by viral reintroduction.

Disrupted sleep-wake regulation in the MCI-Park mouse model of Parkinson's disease

Disrupted sleep has a profound adverse impact on lives of Parkinson's disease (PD) patients and their caregivers. Sleep disturbances are exceedingly common in PD, with substantial heterogeneity in type, timing, and severity. Among the most common sleep-related symptoms reported by PD patients are insomnia, excessive daytime sleepiness, and sleep fragmentation, characterized by interruptions and decreased continuity of sleep. Alterations in brain wave activity, as measured on the electroencephalogram (EEG), also occur in PD, with changes in the pattern and relative contributions of different frequency bands of the EEG spectrum to overall EEG activity in different vigilance states consistently observed. The mechanisms underlying these PD-associated sleep-wake abnormalities are poorly understood, and they are ineffectively treated by conventional PD therapies. To help fill this gap in knowledge, a new progressive model of PD - the MCI-Park mouse - was studied. Near the transition to the parkinsonian state, these mice exhibited significantly altered sleep-wake regulation, including increased wakefulness, decreased non-rapid eye movement (NREM) sleep, increased sleep fragmentation, reduced rapid eye movement (REM) sleep, and altered EEG activity patterns. These sleep-wake abnormalities resemble those identified in PD patients. Thus, this model may help elucidate the circuit mechanisms underlying sleep disruption in PD and identify targets for novel therapeutic approaches.

Activation of locus coeruleus noradrenergic neurons rapidly drives homeostatic sleep pressure

Homeostatic sleep regulation is essential for optimizing the amount and timing of sleep, but the underlying mechanism remains unclear. Optogenetic activation of locus coeruleus noradrenergic neurons immediately increased sleep propensity following transient wakefulness. Fiber photometry showed that repeated optogenetic or sensory stimulation caused rapid declines of locus coeruleus calcium activity and noradrenaline release. This suggests that functional fatigue of noradrenergic neurons, which reduces their wake-promoting capacity, contributes to sleep pressure.

Periodic and aperiodic changes to cortical EEG in response to pharmacological manipulation

Cortical electroencephalograms (EEGs) may help understanding of neuropsychiatric illness and new treatment mechanisms. The aperiodic component (1/f) of EEG power spectra is often treated as noise, but recent studies suggest that changes to the aperiodic exponent of power spectra may reflect changes in excitation/inhibition balance, a concept linked to antidepressant effects, epilepsy, autism, and other clinical conditions. One confound of previous studies is behavioral state, because factors associated with behavioral state other than excitation/inhibition ratio may alter EEG parameters. Thus, to test the robustness of the aperiodic exponent as a predictor of excitation/inhibition ratio, we analyzed video-EEG during active exploration in mice of both sexes during various pharmacological manipulations with the fitting oscillations and one over f (FOOOF) algorithm. We found that GABAA receptor (GABAAR)-positive allosteric modulators increased the aperiodic exponent, consistent with the hypothesis that an increased exponent signals enhanced cortical inhibition, but other drugs (ketamine and GABAAR antagonists at subconvulsive doses) did not follow the prediction. To tilt excitation/inhibition ratio more selectively toward excitation, we suppressed the activity of parvalbumin-positive interneurons with Designer Receptors Exclusively Activated by Designer Drugs (DREADDs). Contrary to our expectations, circuit disinhibition with the DREADD increased the aperiodic exponent. We conclude that the aperiodic exponent of EEG power spectra does not yield a universally reliable marker of cortical excitation/inhibition ratio.NEW & NOTEWORTHY Neuropsychiatric illness may be associated with altered excitation/inhibition balance. A single electroencephalogram (EEG) parameter, the aperiodic exponent of power spectra, may predict the ratio between excitation and inhibition. Here, we use cortical EEGs in mice to evaluate this hypothesis, using pharmacological manipulations of known mechanism. We show that the aperiodic exponent of EEG power spectra is not a reliable marker of excitation/inhibition ratio. Thus, alternative markers of this ratio must be sought.

Knocking out the LRRK2 gene increases sensitivity to wavelength information in rats

Leucine-rich repeat kinase 2 (LRRK2) is a gene related to familial Parkinson's disease (PD). It has been associated with nonmotor symptoms such as disturbances in the visual system affecting colour discrimination and contrast sensitivity. This study examined how deficiency of LRRK2 impacts visual processing in adult rats. Additionally, we investigated whether these changes can be modelled in wild-type rats by administering the LRRK2 inhibitor PFE360. Visual evoked potentials (VEPs) and steady-state visual evoked potentials (SSVEPs) were recorded in the visual cortex and superior colliculus of female LRRK2-knockout and wild-type rats to study how the innate absence of LRRK2 changes visual processing. Exposing the animals to stimulation at five different wavelengths revealed an interaction between genotype and the response to stimulation at different wavelengths. Differences in VEP amplitudes and latencies were robust and barely impacted by the presence of the LRRK2 inhibitor PFE360, suggesting a developmental effect. Taken together, these results indicate that alterations in visual processing were related to developmental deficiency of LRRK2 and not acute deficiency of LRRK2, indicating a role of LRRK2 in the functional development of the visual system and synaptic transmission.

Sleep-Deep-Net learns sleep wake scoring from the end-user and completes each record in their style

Sleep-wake scoring is a time-consuming, tedious but essential component of clinical and pre-clinical sleep research. Sleep scoring is even more laborious and challenging in rodents due to the smaller EEG amplitude differences between states and the rapid state transitions which necessitate scoring in shorter epochs. Although many automated rodent sleep scoring methods exist, they do not perform as well when scoring new data sets, especially those which involve changes in the EEG/EMG profile. Thus, manual scoring by expert scorers remains the gold-standard. Here we take a different approach to this problem by using a neural network to accelerate the scoring of expert scorers. Sleep-Deep-Net (SDN) creates a bespoke deep convolution neural network model for individual electroencephalographic or local-field-potential records via transfer learning of GoogleNet, by learning from a small subset of manual scores of each EEG/LFP record as provided by the end-user. SDN then automates scoring of the remainder of the EEG/LFP record. A novel REM scoring correction procedure further enhanced accuracy. SDN reliably scores EEG and LFP data and retains sleep-wake architecture in wild-type mice, in sleep induced by the hypnotic zolpidem, in a mouse model of Alzheimer's disease and in a genetic knock-down study, when compared to manual scoring. SDN reduced manual scoring time to 1/12. Since SDN uses transfer learning on each independent recording, it is not biased by previously scored existing data sets. Thus, we find SDN performs well when used on signals altered by a drug, disease model or genetic modification.

Prolonged sleep deprivation induces a cytokine-storm-like syndrome in mammals

Most animals require sleep, and sleep loss induces serious pathophysiological consequences, including death. Previous experimental approaches for investigating sleep impacts in mice have been unable to persistently deprive animals of both rapid eye movement sleep (REMS) and non-rapid eye movement sleep (NREMS). Here, we report a "curling prevention by water" paradigm wherein mice remain awake 96% of the time. After 4 days of exposure, mice exhibit severe inflammation, and approximately 80% die. Sleep deprivation increases levels of prostaglandin D2 (PGD2) in the brain, and we found that elevated PGD2 efflux across the blood-brain-barrier-mediated by ATP-binding cassette subfamily C4 transporter-induces both accumulation of circulating neutrophils and a cytokine-storm-like syndrome. Experimental disruption of the PGD2/DP1 axis dramatically reduced sleep-deprivation-induced inflammation. Thus, our study reveals that sleep-related changes in PGD2 in the central nervous system drive profound pathological consequences in the peripheral immune system.

Dopamine release in the nucleus accumbens promotes REM sleep and cataplexy

Patients with the sleep disorder narcolepsy suffer from excessive daytime sleepiness, disrupted nighttime sleep, and cataplexy-the abrupt loss of postural muscle tone during wakefulness, often triggered by strong emotion. The dopamine (DA) system is implicated in both sleep-wake states and cataplexy, but little is known about the function of DA release in the striatum and sleep disorders. Recording DA release in the ventral striatum revealed orexin-independent changes across sleep-wake states as well as striking increases in DA release in the ventral, but not dorsal, striatum prior to cataplexy onset. Tonic low-frequency stimulation of ventral tegmental efferents in the ventral striatum suppressed both cataplexy and rapid eye movement (REM) sleep, while phasic high-frequency stimulation increased cataplexy propensity and decreased the latency to REM sleep. Together, our findings demonstrate a functional role of DA release in the striatum in regulating cataplexy and REM sleep.

Sleep need, the key regulator of sleep homeostasis, is indicated and controlled by phosphorylation of threonine 221 in salt-inducible kinase 3

Sleep need drives sleep and plays a key role in homeostatic regulation of sleep. So far sleep need can only be inferred by animal behaviors and indicated by electroencephalography (EEG). Here we report that phosphorylation of threonine (T) 221 of the salt-inducible kinase 3 (SIK3) increased the catalytic activity and stability of SIK3. T221 phosphorylation in the mouse brain indicates sleep need: more sleep resulting in less phosphorylation and less sleep more phosphorylation during daily sleep/wake cycle and after sleep deprivation (SD). Sleep need was reduced in SIK3 loss of function (LOF) mutants and by T221 mutation to alanine (T221A). Rebound after SD was also decreased in SIK3 LOF and T221A mutant mice. By contrast, SIK1 and SIK2 do not satisfy criteria to be both an indicator and a controller of sleep need. Our results reveal SIK3-T221 phosphorylation as a chemical modification which indicates and controls sleep need.

A novel machine learning system for identifying sleep-wake states in mice

Research into sleep-wake behaviors relies on scoring sleep states, normally done by manual inspection of electroencephalogram (EEG) and electromyogram (EMG) recordings. This is a highly time-consuming process prone to inter-rater variability. When studying relationships between sleep and motor function, analyzing arousal states under a four-state system of active wake (AW), quiet wake (QW), nonrapid-eye-movement (NREM) sleep, and rapid-eye-movement (REM) sleep provides greater precision in behavioral analysis but is a more complex model for classification than the traditional three-state identification (wake, NREM, and REM sleep) usually used in rodent models. Characteristic features between sleep-wake states provide potential for the use of machine learning to automate classification. Here, we devised SleepEns, which uses a novel ensemble architecture, the time-series ensemble. SleepEns achieved 90% accuracy to the source expert, which was statistically similar to the performance of two other human experts. Considering the capacity for classification disagreements that are still physiologically reasonable, SleepEns had an acceptable performance of 99% accuracy, as determined blindly by the source expert. Classifications given by SleepEns also maintained similar sleep-wake characteristics compared to expert classifications, some of which were essential for sleep-wake identification. Hence, our approach achieves results comparable to human ability in a fraction of the time. This new machine-learning ensemble will significantly impact the ability of sleep researcher to detect and study sleep-wake behaviors in mice and potentially in humans.

Striatal dopamine regulates sleep states and narcolepsy-cataplexy

Disruptions to sleep can be debilitating and have a severe effect on daily life. Patients with the sleep disorder narcolepsy suffer from excessive daytime sleepiness, disrupted nighttime sleep, and cataplexy - the abrupt loss of postural muscle tone (atonia) during wakefulness, often triggered by strong emotion. The dopamine (DA) system is implicated in both sleep-wake states and cataplexy, but little is known about the function of DA release in the striatum - a major output region of midbrain DA neurons - and sleep disorders. To better characterize the function and pattern of DA release in sleepiness and cataplexy, we combined optogenetics, fiber photometry, and sleep recordings in a murine model of narcolepsy (orexin-/-; OX KO) and in wildtype mice. Recording DA release in the ventral striatum revealed OX-independent changes across sleep-wake states as well as striking increases in DA release in the ventral, but not dorsal, striatum prior to cataplexy onset. Tonic low frequency stimulation of ventral tegmental efferents in the ventral striatum suppressed both cataplexy and REM sleep, while phasic high frequency stimulation increased cataplexy propensity and decreased the latency to rapid eye movement (REM) sleep. Together, our findings demonstrate a functional role of DA release in the striatum in regulating cataplexy and REM sleep.

Sleep deprivation exacerbates microglial reactivity and Aβ deposition in a TREM2-dependent manner in mice

Sleep loss is associated with cognitive decline in the aging population and is a risk factor for Alzheimer's disease (AD). Considering the crucial role of immunomodulating genes such as that encoding the triggering receptor expressed on myeloid cells type 2 (TREM2) in removing pathogenic amyloid-β (Aβ) plaques and regulating neurodegeneration in the brain, our aim was to investigate whether and how sleep loss influences microglial function in mice. We chronically sleep-deprived wild-type mice and the 5xFAD mouse model of cerebral amyloidosis, expressing either the humanized TREM2 common variant, the loss-of-function R47H AD-associated risk variant, or without TREM2 expression. Sleep deprivation not only enhanced TREM2-dependent Aβ plaque deposition compared with 5xFAD mice with normal sleeping patterns but also induced microglial reactivity that was independent of the presence of parenchymal Aβ plaques. We investigated lysosomal morphology using transmission electron microscopy and found abnormalities particularly in mice without Aβ plaques and also observed lysosomal maturation impairments in a TREM2-dependent manner in both microglia and neurons, suggesting that changes in sleep modified neuro-immune cross-talk. Unbiased transcriptome and proteome profiling provided mechanistic insights into functional pathways triggered by sleep deprivation that were unique to TREM2 and Aβ pathology and that converged on metabolic dyshomeostasis. Our findings highlight that sleep deprivation directly affects microglial reactivity, for which TREM2 is required, by altering the metabolic ability to cope with the energy demands of prolonged wakefulness, leading to further Aβ deposition, and underlines the importance of sleep modulation as a promising future therapeutic approach.

Classifier for the Rapid Simultaneous Determination of Sleep-Wake States and Seizures in Mice

Independent automated scoring of sleep-wake and seizures have recently been achieved; however, the combined scoring of both states has yet to be reported. Mouse models of epilepsy typically demonstrate an abnormal electroencephalographic (EEG) background with significant variability between mice, making combined scoring a more difficult classification problem for manual and automated scoring. Given the extensive EEG variability between epileptic mice, large group sizes are needed for most studies. As large datasets are unwieldy and impractical to score manually, automatic seizure and sleep-wake classification are warranted. To this end, we developed an accurate automated classifier of sleep-wake states, seizures, and the post-ictal state. Our benchmark was a classification accuracy at or above the 93% level of human inter-rater agreement. Given the failure of parametric scoring in the setting of altered baseline EEGs, we adopted a machine-learning approach. We created several multi-layer neural network architectures that were trained on human-scored training data from an extensive repository of continuous recordings of electrocorticogram (ECoG), left and right hippocampal local field potential (HPC-L and HPC-R), and electromyogram (EMG) in the murine intra-amygdala kainic acid model of medial temporal lobe epilepsy. We then compared different network models, finding a bidirectional long short-term memory (BiLSTM) design to show the best performance with validation and test portions of the dataset. The SWISC (sleep-wake and the ictal state classifier) achieved >93% scoring accuracy in all categories for epileptic and non-epileptic mice. Classification performance was principally dependent on hippocampal signals and performed well without EMG. Additionally, performance is within desirable limits for recording montages featuring only ECoG channels, expanding its potential scope. This accurate classifier will allow for rapid combined sleep-wake and seizure scoring in mouse models of epilepsy and other neurologic diseases with varying EEG abnormalities, thereby facilitating rigorous experiments with larger numbers of mice.

Fast Detection and Quantification of Interictal Spikes and Seizures in a Rodent Model of Epilepsy Using an Automated Algorithm

The electroencephalogram (EEG) is a powerful tool for analyzing neural activity in various neurological disorders, both in animals and in humans. This technology has enabled researchers to record the brain's abrupt changes in electrical activity with high resolution, thus facilitating efforts to understand the brain's response to internal and external stimuli. The EEG signal acquired from implanted electrodes can be used to precisely study the spiking patterns that occur during abnormal neural discharges. These patterns can be analyzed in conjunction with behavioral observations and serve as an important means for accurate asses sment and quantification of behavioral and electrographic seizures. Numerous algorithms have been developed for the automated quantification of EEG data; however, many of these algorithms were developed with outdated programming languages and require robust computational hardware to run effectively. Additionally, some of these programs require substantial computation time, reducing the relative benefits of automation. Thus, we sought to develop an automated EEG algorithm that was programmed using a familiar programming language (MATLAB), and that could run efficiently without extensive computational demands. This algorithm was developed to quantify interictal spikes and seizures in mice that were subjected to traumatic brain injury. Although the algorithm was designed to be fully automated, it can be operated manually, and all the parameters for EEG activity detection can be easily modified for broad data analysis. Additionally, the algorithm is capable of processing months of lengthy EEG datasets in the order of minutes to hours, reducing both analysis time and errors introduced through manual-based processing.

Acute effect of antiseizure drugs on background oscillations in Scn1aA1783V Dravet syndrome mouse model

Dravet syndrome (Dravet) is a rare and severe form of developmental epileptic encephalopathy. Antiseizure medications (ASMs) for Dravet patients include valproic acid (VA) or clobazam (CLB), with or without stiripentol (STP), while sodium channel blockers like carbamazepine (CBZ) or lamotrigine (LTG) are contraindicated. In addition to their effect on epileptic phenotypes, ASMs were shown to modify the properties of background neuronal activity. Nevertheless, little is known about these background properties alterations in Dravet. Here, utilizing Dravet mice (DS, Scn1aA1783V/WT), we tested the acute effect of several ASMs on background electrocorticography (ECoG) activity and frequency of interictal spikes. Compared to wild-type mice, background ECoG activity in DS mice had lower power and reduced phase coherence, which was not corrected by any of the tested ASMs. However, acute administration of Dravet-recommended drugs, VA, CLB, or a combination of CLB + STP, caused, in most mice, a reduction in the frequency of interictal spikes, alongside an increase in the relative contribution of the beta frequency band. Conversely, CBZ and LTG increased the frequency of interictal spikes, with no effect on background spectral properties. Moreover, we uncovered a correlation between the reduction in interictal spike frequency, the drug-induced effect on the power of background activity, and a spectral shift toward higher frequency bands. Together, these data provide a comprehensive analysis of the effect of selected ASMs on the properties of background neuronal oscillations, and highlight a possible correlation between their effect on epilepsy and background activity.

IntelliSleepScorer, a software package with a graphic user interface for automated sleep stage scoring in mice based on a light gradient boosting machine algorithm

Machine learning has been applied in recent years to categorize sleep stages (NREM, REM, and wake) using electroencephalogram (EEG) recordings; however, a well-validated sleep scoring automatic pipeline in rodent research is still not publicly available. Here, we present IntelliSleepScorer, a software package with a graphic user interface to score sleep stages automatically in mice. IntelliSleepScorer uses the light gradient boosting machine (LightGBM) to score sleep stages for each epoch of recordings. We developed LightGBM models using a large cohort of data, which consisted of 5776 h of sleep EEG and electromyogram (EMG) signals across 519 unique recordings from 124 mice. The LightGBM model achieved an overall accuracy of 95.2% and a Cohen's kappa of 0.91, which outperforms the baseline models such as the logistic regression model (accuracy = 93.3%, kappa = 0.88) and the random forest model (accuracy = 94.3%, kappa = 0.89). The overall performance of the LightGBM model as well as the performance across different sleep stages are on par with that of the human experts. Most importantly, we validated the generalizability of the LightGBM models: (1) The LightGBM model performed well on two publicly available, independent datasets (kappa >  = 0.80), which have different sampling frequency and epoch lengths; (2) The LightGBM model performed well on data recorded at a lower sampling frequency (kappa = 0.90); (3) The performance of the LightGBM model is not affected by the light/dark cycle; and (4) A modified LightGBM model performed well on data containing only one EEG and one EMG electrode (kappa >  = 0.89). Taken together, the LightGBM models offer state-of-the-art performance for automatic sleep stage scoring in mice. Last, we implemented the IntelliSleepScorer software package based on the validated model to provide an out-of-box solution to sleep researchers (available for download at https://sites.broadinstitute.org/pan-lab/resources ).

Genetically encoded sensors for measuring histamine release both in vitro and in vivo

Histamine (HA) is a key biogenic monoamine involved in a wide range of physiological and pathological processes in both the central and peripheral nervous systems. Because the ability to directly measure extracellular HA in real time will provide important insights into the functional role of HA in complex circuits under a variety of conditions, we developed a series of genetically encoded G-protein-coupled receptor-activation-based (GRAB) HA (GRAB_HA) sensors with good photostability, sub-second kinetics, nanomolar affinity, and high specificity. Using these GRAB_HA sensors, we measured electrical-stimulation-evoked HA release in acute brain slices with high spatiotemporal resolution. Moreover, we recorded HA release in the preoptic area of the hypothalamus and prefrontal cortex during the sleep-wake cycle in freely moving mice, finding distinct patterns of HA dynamics between these specific brain regions. Thus, GRAB_HA sensors are robust tools for measuring extracellular HA transmission in both physiological and pathological processes.

Pupil Dynamics-derived Sleep Stage Classification of a Head-fixed Mouse Using a Recurrent Neural Network

The standard method for sleep state classification is thresholding the amplitudes of electroencephalography (EEG) and electromyography (EMG) data, followed by manual correction by an expert. Although popular, this method has some shortcomings: (1) the time-consuming manual correction by human experts is sometimes a bottleneck hindering sleep studies, (2) EEG electrodes on the skull interfere with wide-field imaging of the cortical activity of a head-fixed mouse under a microscope, (3) invasive surgery to fix the electrodes on the thin mouse skull risks brain tissue injury, and (4) metal electrodes for EEG and EMG recording are difficult to apply to some experimental apparatus such as that for functional magnetic resonance imaging. To overcome these shortcomings, we propose a pupil dynamics-based vigilance state classification method for a head-fixed mouse using a long short-term memory (LSTM) model, a variant of a recurrent neural network, for multi-class labeling of NREM, REM, and WAKE states. For supervisory hypnography, EEG and EMG recording were performed on head-fixed mice. This setup was combined with left eye pupillometry using a USB camera and a markerless tracking toolbox, DeepLabCut. Our open-source LSTM model with feature inputs of pupil diameter, pupil location, pupil velocity, and eyelid opening for 10 s at a 10 Hz sampling rate achieved vigilance state estimation with a higher classification performance (macro F1 score, 0.77; accuracy, 86%) than a feed-forward neural network. Findings from a diverse range of pupillary dynamics implied possible subdivision of the vigilance states defined by EEG and EMG. Pupil dynamics-based hypnography can expand the scope of alternatives for sleep stage scoring of head-fixed mice.

Cortical regulation of two-stage rapid eye movement sleep

Rapid eye movement (REM) sleep is a sleep state characterized by skeletal muscle paralysis and cerebral cortical activation. Yet, global cortical dynamics and their role in regulating REM sleep remain unclear. Here we show that in mice, REM sleep is accompanied by highly patterned cortical activity waves, with the retrosplenial cortex (RSC) as a major initiation site. Two-photon imaging of layer 2/3 pyramidal neurons of the RSC revealed two distinct patterns of population activities during REM sleep. These activities encoded two sequential REM sleep substages, characterized by contrasting facial movement and autonomic activity and by distinguishable electroencephalogram theta oscillations. Closed-loop optogenetic inactivation of RSC during REM sleep altered cortical activity dynamics and shortened REM sleep duration via inhibition of the REM substage transition. These results highlight an important role for the RSC in dictating cortical dynamics and regulating REM sleep progression.

Validation of Deep Learning-based Sleep State Classification

Deep learning methods have been developed to classify sleep states of mouse electroencephalogram (EEG) and electromyogram (EMG) recordings with accuracy reported as high as 97%. However, when applied to independent datasets, with a variety of experimental and recording conditions, sleep state classification accuracy often drops due to distributional shift. Mixture z-scoring, a pre-processing standardization of EEG/EMG signals, has been suggested to account for these variations. This study sought to validate mixture z-scoring in combination with a deep learning method on an independent dataset. The open-source software Accusleep, which implements mixture z-scoring in combination with deep learning via a convolutional neural network, was used to classify sleep states in 12, three-hour EEG/EMG recordings from mice sleeping in a head-fixed position. Mixture z-scoring with deep learning classified sleep states on two independent recordings with 85-92% accuracy and a Cohen's κ of 0.66-0.71. These results validate mixture z-scoring in combination with deep learning to classify sleep states with the potential for widespread use.

Optimization of real-time analysis of sleep-wake cycle in mice

Studying the biology of sleep requires accurate and efficient assessment of the sleep stages. However, analysis of sleep-wake cycles in mice and other laboratory animals remains a time-consuming and laborious process. In this study, we developed a Python script and a process for the streamlined analysis of sleep data that includes real-time processing of electroencephalogram (EEG) and electromyogram (EMG) signals that is compatible with commercial sleep-recording software that supports user datagram protocol (UDP) communication. The process consists of EEG/EMG data acquisition, automated threshold calculation for real-time determination of sleep stages, sleep staging and EEG power spectrum analysis. It also allows data storage in the format that facilitates further analysis of the sleep pattern in mice. The described method is aimed at increasing efficiency of sleep stage scoring and analysis in mice thus facilitating sleep research. • A process of EEG/EMG recording and streamline analysis of sleep-wake cycle in real time in mice. • The compatibility with commercial sleep-recording software that can generate a UDP stream. • The capability of further analysis of recorded data by an open-source software.

Employing a Long-Short-Term Memory Neural Network to Improve Automatic Sleep Stage Classification of Pharmaco-EEG Profiles

An increasing problem in today’s society is the spiraling number of people suffering from various sleep disorders. The research results presented in this paper support the use of a novel method that employs techniques from the classification of sleep disorders for more accurate scoring. Applying this novel method will assist researchers with better analyzing subject profiles for recommending prescriptions or to alleviate sleep disorders. In biomedical research, the use of animal models is required to experimentally test the safety and efficacy of a drug in the pre-clinical stage. We have developed a novel LSTM Recurrent Neural Network to process Pharmaco-EEG Profiles of rats to automatically score their sleep–wake stages. The results indicate improvements over the current methods; for the case of combined channels, the model accuracy improved by 1% and 3% in binary or multiclass classifications, respectively, to accuracies of 93% and 82%. In the case of using a single channel, binary and multiclass LSTM models for identifying rodent sleep stages using single or multiple electrode positions for binary or multiclass problems have not been evaluated in prior literature. The results reveal that single or combined channels, and binary or multiclass classification tasks, can be applied in the automatic sleep scoring of rodents.

LKB1 Is Physiologically Required for Sleep from Drosophila melanogaster to the Mus musculus

Liver Kinase B1 (LKB1) is known as a master kinase for 14 kinases related to the adenosine monophosphate (AMP)-activated protein kinase (AMPK). Two of them salt inducible kinase 3 (SIK3) and AMPKα have previously been implicated in sleep regulation. We generated loss-of-function (LOF) mutants for Lkb1 in both Drosophila and mice. Sleep, but not circadian rhythms, was reduced in Lkb1-mutant flies and in flies with neuronal deletion of Lkb1. Genetic interactions between Lkb1 and Threonine to Alanine mutation at residue 184 of AMPK in Drosophila sleep or those between Lkb1 and Threonine to Glutamic Acid mutation at residue 196 of SIK3 in Drosophila viability have been observed. Sleep was reduced in mice after virally mediated reduction of Lkb1 in the brain. Electroencephalography (EEG) analysis showed that non-rapid eye movement (NREM) sleep and sleep need were both reduced in Lkb1-mutant mice. These results indicate that LKB1 plays a physiological role in sleep regulation conserved from flies to mice.

Disrupted neural correlates of anesthesia and sleep reveal early circuit dysfunctions in Alzheimer models

Dysregulated homeostasis of neural activity has been hypothesized to drive Alzheimer's disease (AD) pathogenesis. AD begins with a decades-long presymptomatic phase, but whether homeostatic mechanisms already begin failing during this silent phase is unknown. We show that before the onset of memory decline and sleep disturbances, familial AD (fAD) model mice display no deficits in CA1 mean firing rate (MFR) during active wakefulness. However, homeostatic down-regulation of CA1 MFR is disrupted during non-rapid eye movement (NREM) sleep and general anesthesia in fAD mouse models. The resultant hyperexcitability is attenuated by the mitochondrial dihydroorotate dehydrogenase (DHODH) enzyme inhibitor, which tunes MFR toward lower set-point values. Ex vivo fAD mutations impair downward MFR homeostasis, resulting in pathological MFR set points in response to anesthetic drug and inhibition blockade. Thus, firing rate dyshomeostasis of hippocampal circuits is masked during active wakefulness but surfaces during low-arousal brain states, representing an early failure of the silent disease stage.

Automated sleep state classification of wide-field calcium imaging data via multiplex visibility graphs and deep learning

BACKGROUND

Wide-field calcium imaging (WFCI) allows for monitoring of cortex-wide neural dynamics in mice. When applied to the study of sleep, WFCI data are manually scored into the sleep states of wakefulness, non-REM (NREM) and REM by use of adjunct EEG and EMG recordings. However, this process is time-consuming and often suffers from low inter- and intra-rater reliability and invasiveness. Therefore, an automated sleep state classification method that operates on WFCI data alone is needed.

NEW METHOD

A hybrid, two-step method is proposed. In the first step, spatial-temporal WFCI data is mapped to multiplex visibility graphs (MVGs). Subsequently, a two-dimensional convolutional neural network (2D CNN) is employed on the MVGs to be classified as wakefulness, NREM and REM.

RESULTS

Sleep states were classified with an accuracy of 84% and Cohen's κ of 0.67. The method was also effectively applied on a binary classification of wakefulness/sleep (accuracy=0.82, κ = 0.62) and a four-class wakefulness/sleep/anesthesia/movement classification (accuracy=0.74, κ = 0.66). Gradient-weighted class activation maps revealed that the CNN focused on short- and long-term temporal connections of MVGs in a sleep state-specific manner. Sleep state classification performance when using individual brain regions was highest for the posterior area of the cortex and when cortex-wide activity was considered.

COMPARISON WITH EXISTING METHOD

On a 3-hour WFCI recording, the MVG-CNN achieved a κ of 0.65, comparable to a κ of 0.60 corresponding to the human EEG/EMG-based scoring.

CONCLUSIONS

The hybrid MVG-CNN method accurately classifies sleep states from WFCI data and will enable future sleep-focused studies with WFCI.

High-throughput visual assessment of sleep stages in mice using machine learning

STUDY OBJECTIVES

Sleep is an important biological process that is perturbed in numerous diseases, and assessment its substages currently requires implantation of electrodes to carry out electroencephalogram/electromyogram (EEG/EMG) analysis. Although accurate, this method comes at a high cost of invasive surgery and experts trained to score EEG/EMG data. Here, we leverage modern computer vision methods to directly classify sleep substages from video data. This bypasses the need for surgery and expert scoring, provides a path to high-throughput studies of sleep in mice.

METHODS

We collected synchronized high-resolution video and EEG/EMG data in 16 male C57BL/6J mice. We extracted features from the video that are time and frequency-based and used the human expert-scored EEG/EMG data to train a visual classifier. We investigated several classifiers and data augmentation methods.

RESULTS

Our visual sleep classifier proved to be highly accurate in classifying wake, non-rapid eye movement sleep (NREM), and rapid eye movement sleep (REM) states, and achieves an overall accuracy of 0.92 +/- 0.05 (mean +/- SD). We discover and genetically validate video features that correlate with breathing rates, and show low and high variability in NREM and REM sleep, respectively. Finally, we apply our methods to non-invasively detect that sleep stage disturbances induced by amphetamine administration.

CONCLUSIONS

We conclude that machine learning based visual classification of sleep is a viable alternative to EEG/EMG based scoring. Our results will enable non-invasive high-throughput sleep studies and will greatly reduce the barrier to screening mutant mice for abnormalities in sleep.

An artificial neural network for automated behavioral state classification in rats

Accurate behavioral state classification is critical for many research applications. Researchers typically rely upon manual identification of behavioral state through visual inspection of electrophysiological signals, but this approach is time intensive and subject to low inter-rater reliability. To overcome these limitations, a diverse set of algorithmic approaches have been put forth to automate the classification process. Recently, novel machine learning approaches have been detailed that produce rapid and highly accurate classifications. These approaches however, are often computationally expensive, require significant expertise to implement, and/or require proprietary software that limits broader adoption. Here we detail a novel artificial neural network that uses electrophysiological features to automatically classify behavioral state in rats with high accuracy, sensitivity, and specificity. Common parameters of interest to sleep scientists, including state-dependent power spectra and homeostatic non-REM slow wave activity, did not significantly differ when using this automated classifier as compared to manual scoring. Flexible options enable researchers to further increase classification accuracy through manual rescoring of a small subset of time intervals with low model prediction certainty or further decrease researcher time by generalizing trained networks across multiple recording days. The algorithm is fully open-source and coded within a popular, and freely available, software platform to increase access to this research tool and provide additional flexibility for future researchers. In sum, we have developed a readily implementable, efficient, and effective approach for automated behavioral state classification in rats.

A deep learning algorithm for sleep stage scoring in mice based on a multimodal network with fine-tuning technique

Sleep stage scoring is important to determine sleep structure in preclinical and clinical research. The aim of this study was to develop an automatic sleep stage classification system for mice with a new deep neural network algorithm. For the purpose of base feature extraction, wake-sleep and rapid eye movement (REM) and non- rapid eye movement (NREM) models were developed by extracting defining features from mouse-derived electromyogram (EMG) and electroencephalogram (EEG) signals, respectively. The wake-sleep model and REM-NREM sleep model were integrated into three different algorithms including a rule-based integration approach, an ensemble stacking approach, and a multimodal with fine-tuning approach. The deep learning algorithm assessing sleep stages in animal experiments by the multimodal with fine-tuning approach showed high potential for increasing accuracy in sleep stage scoring in mice and promoting sleep research.

Mice Lacking GABAA Receptor δ Subunit Have Altered Pharmaco-EEG Responses to Multiple Drugs

In the brain, extrasynaptically expressed ionotropic, δ subunit-containing γ-aminobutyric acid A-type receptors (δ-GABA_ARs) have been implicated in drug effects at both neuronal and behavioral levels. These alterations are supposed to be caused via drug-induced modulation of receptor ionophores affecting chloride ion-mediated inhibitory tonic currents. Often, a transgenic mouse model genetically lacking the δ-GABA_ARs (δ-KO) has been used to study the roles of δ-GABA_ARs in brain functions, because a specific antagonist of the δ-GABA_ARs is still lacking. We have previously observed with these δ-KO mice that activation of δ-GABA_ARs is needed for morphine-induced conditioning of place preference, and others have suggested that δ-GABA_ARs act as targets selectively for low doses of ethanol. Furthermore, activation of these receptors via drug-mediated agonism induces a robust increase in the slow-wave frequency bands of electroencephalography (EEG). Here, we tested δ-KO mice (compared to littermate wild-type controls) for the pharmaco-EEG responses of a broad spectrum of pharmacologically different drug classes, including alcohol, opioids, stimulants, and psychedelics. Gaboxadol (THIP), a known superagonist of δ-GABA_ARs, was included as the positive control, and as expected, δ-KO mice produced a blunted pharmaco-EEG response to 6 mg/kg THIP. Pharmaco-EEGs showed notable differences between treatments but also differences between δ-KO mice and their wild-type littermates. Interestingly mephedrone (4-MMC, 5 mg/kg), an amphetamine-like stimulant, had reduced effects in the δ-KO mice. The responses to ethanol (1 g/kg), LSD (0.2 mg/kg), and morphine (20 mg/kg) were similar in δ-KO and wild-type mice. Since stimulants are not known to act on δ-GABA_ARs, our findings on pharmaco-EEG effects of 4-MMC suggest that δ-GABA_ARs are involved in the secondary indirect regulation of the brain rhythms after 4-MMC.

Automated scoring of pre-REM sleep in mice with deep learning

Reliable automation of the labor-intensive manual task of scoring animal sleep can facilitate the analysis of long-term sleep studies. In recent years, deep-learning-based systems, which learn optimal features from the data, increased scoring accuracies for the classical sleep stages of Wake, REM, and Non-REM. Meanwhile, it has been recognized that the statistics of transitional stages such as pre-REM, found between Non-REM and REM, may hold additional insight into the physiology of sleep and are now under vivid investigation. We propose a classification system based on a simple neural network architecture that scores the classical stages as well as pre-REM sleep in mice. When restricted to the classical stages, the optimized network showed state-of-the-art classification performance with an out-of-sample F1 score of 0.95 in male C57BL/6J mice. When unrestricted, the network showed lower F1 scores on pre-REM (0.5) compared to the classical stages. The result is comparable to previous attempts to score transitional stages in other species such as transition sleep in rats or N1 sleep in humans. Nevertheless, we observed that the sequence of predictions including pre-REM typically transitioned from Non-REM to REM reflecting sleep dynamics observed by human scorers. Our findings provide further evidence for the difficulty of scoring transitional sleep stages, likely because such stages of sleep are under-represented in typical data sets or show large inter-scorer variability. We further provide our source code and an online platform to run predictions with our trained network.

WaveSleepNet: An interpretable deep convolutional neural network for the continuous classification of mouse sleep and wake

BACKGROUND

Recent advancement in deep learning provides a pivotal opportunity to potentially supplement or supplant the limiting step of manual sleep scoring.

NEW METHOD

In this paper, we characterize the WaveSleepNet (WSN), a deep convolutional neural network (CNN) that uses wavelet transformed images of mouse EEG/EMG signals to autoscore sleep and wake.

RESULTS

WSN achieves an epoch by epoch mean accuracy of 0.86 and mean F1 score of 0.82 compared to manual scoring by a human expert. In mice experiencing mechanically induced sleep fragmentation, an overall epoch by epoch mean accuracy of 0.80 is achieved by WSN and classification of non-REM (NREM) sleep is not compromised, but the high level of sleep fragmentation results in WSN having greater difficulty differentiating REM from NREM sleep. We also find that WSN achieves similar levels of accuracy on an independent dataset of externally acquired EEG/EMG recordings with an overall epoch by epoch accuracy of 0.91. We also compared conventional summary sleep metrics in mice sleeping ad libitum. WSN systematically biases sleep fragmentation metrics of bout number and bout length leading to an overestimated degree of sleep fragmentation.

COMPARISON WITH EXISTING METHODS

In a cross-validation, WSN has a greater macro and stage-specific accuracy compared to a conventional random forest classifier. Examining the WSN, we find that it automatically learns spectral features consistent with manual scoring criteria that are used to define each class.

CONCLUSION

These results suggest to us that WSN is capable of learning visually agreeable features and may be useful as a supplement to human manual scoring.

Application of Artificial Intelligence in the Diagnosis and Management of Corneal Diseases

Diagnosis and treatment planning in ophthalmology heavily depend on clinical examination and advanced imaging modalities, which can be time-consuming and carry the risk of human error. Artificial intelligence (AI) and deep learning (DL) are being used in different fields of ophthalmology and in particular, when running diagnostics and predicting outcomes of anterior segment surgeries. This review will evaluate the recent developments in AI for diagnostics, surgical interventions, and prognosis of corneal diseases. It also provides a brief overview of the newer AI dependent modalities in corneal diseases.

Sleep-Wake Cycle and EEG-Based Biomarkers during Late Neonate to Adult Transition

During the transition from neonate to adulthood, brain maturation establishes coherence between behavioral states—wakefulness, non-rapid eye movement, and rapid eye movement sleep. In animal models few studies have characterized and analyzed cerebral rhythms and the sleep–wake cycle in early ages, in relation to adulthood. Since the analysis of sleep in early ages can be used as a predictive model of brain development and the subsequent emergence of neural disturbances in adults, we performed a study on late neonatal mice, an age not previously characterized. We acquired longitudinal 24 h electroencephalogram and electromyogram recordings and performed time and spectral analyses. We compared both age groups and found that late neonates: (i) spent more time in wakefulness and less time in non-rapid eye movement sleep, (ii) showed an increased relative band power in delta, which, however, reduced in theta during each behavioral state, (iii) showed a reduced relative band power in beta during wakefulness and non-rapid eye movement sleep, and (iv) manifested an increased total power over all frequencies. The data presented here might have implications expanding our knowledge of cerebral rhythms in early ages for identification of potential biomarkers in preclinical models of neurodegeneration.

Machine and Deep Learning in Molecular and Genetic Aspects of Sleep Research

Epidemiological sleep research strives to identify the interactions and causal mechanisms by which sleep affects human health, and to design intervention strategies for improving sleep throughout the lifespan. These goals can be advanced by further focusing on the environmental and genetic etiology of sleep disorders, and by development of risk stratification algorithms, to identify people who are at risk or are affected by, sleep disorders. These studies rely on comprehensive sleep-related data which often contains complex multi-dimensional physiological and molecular measurements across multiple timepoints. Thus, sleep research is well-suited for the application of computational approaches that can handle high-dimensional data. Here, we survey recent advances in machine and deep learning together with the availability of large human cohort studies with sleep data that can jointly drive the next breakthroughs in the sleep-research field. We describe sleep-related data types and datasets, and present some of the tasks in the field that can be targets for algorithmic approaches, as well as the challenges and opportunities in pursuing them.

Deep learning approaches for neural decoding across architectures and recording modalities

Decoding behavior, perception or cognitive state directly from neural signals is critical for brain-computer interface research and an important tool for systems neuroscience. In the last decade, deep learning has become the state-of-the-art method in many machine learning tasks ranging from speech recognition to image segmentation. The success of deep networks in other domains has led to a new wave of applications in neuroscience. In this article, we review deep learning approaches to neural decoding. We describe the architectures used for extracting useful features from neural recording modalities ranging from spikes to functional magnetic resonance imaging. Furthermore, we explore how deep learning has been leveraged to predict common outputs including movement, speech and vision, with a focus on how pretrained deep networks can be incorporated as priors for complex decoding targets like acoustic speech or images. Deep learning has been shown to be a useful tool for improving the accuracy and flexibility of neural decoding across a wide range of tasks, and we point out areas for future scientific development.