John A. Thompson

Sleep spindles, transient 11–16 Hz oscillatory bursts, are a defining electrographic feature of non-rem (NREM) sleep and a key biomarker of sleep physiology. A need for efficient and reliable identification of spindles motivates a large literature on automated detection algorithms. In this literature, annotation by trained sleep specialists remains the gold standard against which automated methods are trained, tuned and evaluated. However, inter-scorer agreement among experts is modest, which leaves a significant role for subjective judgment in the definition of a spindle. Finding objective, scorer-independent, criteria for identifying spindles remains an unresolved challenge. We report here a robust, highly specific, and previously unrecognized signature of spindle activity in the fourth-order spectrum (trispectrum), from which we identify the presence of spindles, characterize their waveforms, and obtain an optimal detection filter through a decomposition of the trispectrum (HOSD). Although it is a strictly blind, data-driven method, HOSD-based spindle identification and detection agrees well with expert annotation (median AUROC ~ 0.9), yet identifies many more events at the native threshold than both human scorers and comparison detectors. Many of these additional detections are confirmed as meeting AASM spindle criteria by four blinded specialists, demonstrating that spindle-like oscillatory bursting is prevalent below conventional human and automated detection thresholds. We observe that N2 sleep is distinguished principally by high-amplitude bursts, while low-amplitude bursting persists throughout NREM sleep, being globally suppressed only in REM sleep. We also describe robust identification of recording-specific spindle waveform properties such as frequency deceleration. Key Messages: We introduce the first purely data-driven machine-learning algorithm for identifying and detecting sleep spindles that is independent of human judgment. This method identifies the presence of oscillatory bursting in the spindle band using information in the fourth-order spectrum (trispectrum) through a spectral decomposition of excess kurtosis in the spindle band. Detection thresholds are chosen based on a principled statistical criterion. This new method detects many more spindles than human scorers and other automated spindle detection algorithms. We present evidence that a large proportion of the additional detections represent true positives rather than false positives. We conclude that N2 sleep is primarily characterized by high-amplitude spindles, but that oscillatory bursts within the sigma band are prevalent below the conventional detection threshold in other non-REM stages and wakefulness, and uniquely suppressed during REM sleep.

Background: In Parkinson's disease (PD), sleep‐related oscillatory dynamics between basal ganglia and cortical nodes may inform sleep dysfunction. The subthalamic nucleus (STN) is highly interconnected with human sleep circuitry and is a current target of interest for sleep neuromodulation. Methods: To investigate correlated cortical–subcortical activity, we recorded simultaneous PSG and STN local field potentials (LFP) in 19 patients undergoing deep brain stimulation (DBS) for PD. We evaluated the correlation in spectral power between STN‐LFP and EEG during polysomnogram (PSG)‐defined stages of sleep. In addition, we analyzed aperiodic 1/f spectral parameters (exponent and offset), which have been identified as markers of arousal across sleep stages. Finally, we trained classifiers on aperiodic features to test whether 1/f structure alone could recover PSG‐defined sleep stages from EEG and STN‐LFP. Results: Results revealed a significant interaction between recording modality and sleep stage. During wakefulness and REM sleep, STN‐LFP exhibited a significantly steeper spectral slope compared to scalp EEG. However, this dissociation diminished with sleep depth; slopes appeared more similar during N2 and reversed during N3, where EEG became steeper than LFP. Across subjects, the aperiodic component showed robust state dependence and modality differences with STN‐LFP maintaining higher exponents overall. STN‐LFP aperiodic features alone classified sleep stages above chance and outperformed scalp EEG, approaching the combined‐modality ceiling. Conclusion: These findings highlight distinct subcortical electrophysiological signatures in sleep‐dependent regulation of neural arousal across brain networks in PD. These insights offer potential biomarkers for both invasive and non‐invasive forms of closed‐loop neuromodulation strategies targeting sleep dysfunction. Trial Registration: ClinicalTrials.gov identifier: NCT04620551

There is growing interest and need for non-invasive stimulation approaches for the treatment of Parkinson’s disease (PD) and other neurological conditions. Pilot studies indicate that vibrotactile stimulation on the fingertips may reduce PD motor symptoms (Pfeifer et al., 2021; Syrkin-Nikolau et al., 2018). PD motor symptoms (e.g., rigidity, bradykinesia) are correlated with exaggerated beta power in the subthalamic nucleus (STN), where neurons are excessively synchronized (Brown 2003; Kühn et al., 2006; Neumann et al., 2016; Yin et al., 2021), but the effect of vibrotactile stimulation on the STN has not been determined. Here, in 12 PD participants in the OFF deep brain stimulation (DBS) and OFF medication state, we investigated how unilateral vibrotactile stimulation applied to the fingertips affects local field potential (LFP) power in STN. We used a within-participants design to expose each participant to a treatment stimulation pattern, termed “Rapidly Varying Sequences” within a Coordinated Reset stimulus framework (“CR RVS”, abbreviated as RVS herein; Pfeifer et al., 2021; Zeitler & Tass, 2015), and a control stimulation pattern, with the order randomized and with intermittent acquisition of STN LFP. RVS yielded a modest but statistically significant 12% (SEM 4.6%) reduction in mean normalized STN beta power and a 48% (SEM 19%) reduction in peak beta power compared to the DBS-off baseline condition and was significantly different when compared to our control stimulus. Furthermore, we identified a biomarker in STN beta power that predicts which participants may benefit from RVS. We observed that participants that exhibited prominent beta peaks had stronger reductions in mean beta power (17% reduction, SEM 6.1%) and peak beta power (55% reduction, SEM 10%). Regressing against the magnitude of the peak in beta provides a moderate prediction of change in mean and peak beta power due to RVS (R2 = 0.58 for mean and 0.52 for peak). We then used our observations to construct a computational model where beta peaks in a simulated STN varied from prominent to diminished. We found that the efficacy of randomized treatments was dependent on the magnitude of beta peaking, mirroring our clinical findings, and showing that RVS may act by reducing intra-neuronal synaptic strengths in STN. Despite robust changes in STN LFP in our study population, we did not observe a significant change in motor symptoms. These results suggest that peripheral vibrotactile stimulation can reduce STN beta power and motivate additional studies to investigate its long-term effects on motor symptoms across a large population of participants.

Objective. Deep brain stimulation (DBS) of the subthalamic nucleus (STN) is an established therapy for Parkinson’s disease (PD), yet optimizing outcomes remains challenging due to the complexity of STN functional architecture and the subjectivity of intraoperative assessments. Furthermore, the encoding of fine versus gross movements across STN dorsoventral subregions remains poorly understood. This study aimed to characterize region- and movement-specific modulation of STN neuronal activity using synchronized electrophysiology and markerless motion tracking-derived kinematics. Approach. We recorded intraoperative electrophysiological and videographic data from 11 subjects with PD (15 hemispheres) undergoing awake STN-DBS implantation surgery. Microelectrode recordings were synchronized with high-resolution kinematics during repeated trials of continuous upper-limb motor tasks involving both distal and proximal muscle groups. Instantaneous firing rate (IFR) and neuronal responsiveness metrics were quantified using a parameter-free Zenith of Event-based Time-locked Anomalies (ZETA) statistical framework and analyzed with linear mixed-effects models. Main results. STN single-unit activity (SUA) and multi-unit activity (MUA) features exhibited robust region- and movement-specific modulation. Firing rates for both SUA and MUA were significantly higher in the dorsal STN during active movements, particularly distal fine motor tasks, compared to ventral STN and baseline activity. ZETA-test analyses revealed significant movement-specific responsiveness across STN subregions, with more dextrous, distal movements eliciting stronger responses than more coordinated, proximal movements. Temporal dynamics of neuronal responsiveness also varied with STN depth and movement type, indicating hierarchical recruitment of STN subregions in the encoding of movement-specific temporal motifs. Significance. These findings demonstrate that STN motor representations are context-dependent, anatomically organized, and temporally distinct, reinforcing the STN’s dynamic engagement during motor execution. By identifying objective electrophysiological features linked to movement context and STN region, this work provides a quantitative foundation for reducing reliance on subjective assessments to inform data-driven DBS targeting, programming, and closed-loop control strategies.

Theta oscillations, neural activity within the 4–8 Hz range, are implicated in a wide range of cognitive functions, including oculomotor and sensory processing, attention, memory, and motor planning and execution across diverse brain regions. Saccadic eye movements (SEMs), which are integral to visual perception and cognition, occur within a similar frequency range. This review explores how theta oscillations contribute to oculomotor and cognitive processing, emphasizing their role in coordinated motor and sensory functions. We synthesize foundational and contemporary studies into a working model describing neural synchronization across cognitive networks. We discuss the complex interplay between theta oscillations, SEMs, and cognition, summarizing the current state of our knowledge.

Objective: Patients with Tourette syndrome (TS) may benefit from lower DBS current during sleep to reduce side effects and habituation, yet manual adjustment can be challenging. We tested the feasibility of an adaptive deep brain stimulation (aDBS) paradigm in which circadian fluctuations in beta activity serve as a biomarker to reduce stimulation during sleep. Approach: We analyzed chronic beta-band local field potentials (LFPs) recorded from sensing-enabled DBS leads in bilateral ventral capsule/ventral striatum (VC/VS) and centromedian-parafascicular nucleus of the thalamus (CM-Pf) in a patient with refractory TS and obsessive-compulsive disorder (OCD). We used circadian beta fluctuations in the VC/VS and CM-Pf to drive a dual-threshold algorithm that we configured to function as a slowly adapting single-threshold system. Because the clinically optimal electrode configuration created sensing constraints, we linked VC/VS and CM-Pf within each hemisphere so beta activity from either target could trigger automatic nighttime stimulation reduction. Results: Multi-week recordings showed clear circadian beta rhythmicity in both VC/VS and CM-Pf. The slowly adapting single-threshold aDBS algorithm with cross-target tethering reduced stimulation during sleep while maintaining stable daytime stimulation. A 16-night period in which aDBS was inadvertently disabled created a partially triple-blinded comparison between aDBS and continuous DBS (cDBS). Clinical measures showed modest reductions in tic severity and substantial reductions in depression during aDBS compared with cDBS. Significance: This proof-of-concept study demonstrates: (1) that circadian beta rhythms in limbic and thalamic targets can serve as a biomarker for adaptive DBS; (2) that cross-target tethering is feasible (i.e., adapting stimulation in the VC/VS off of sensing in the CM-Pf and vice versa); and (3) that a dual-threshold algorithm can be configured to enable gradual transitions between maximal and minimal stimulation. The patient's observed improvements during circadian-driven aDBS compared with cDBS suggest potential clinical benefit, warranting testing in larger samples.

Remembering the temporal order of events is critical for episodic memory, allowing us to link individual events into sequences. While the medial temporal lobe and prefrontal cortex are essential for this process, the underlying neural mechanisms remain poorly understood. Here we characterized the representation of order information at the level of single neurons and field potentials recorded from human neurosurgical patients watching naturalistic videos of everyday events and later recalling the order and content of the events depicted. We found order-selective cells (OSCs) in the human hippocampus, amygdala, and orbitofrontal cortex that responded selectively to specific event orders, independent of event content or absolute time. OSCs exhibited transient theta phase precession following their preferred order during both memory encoding and retrieval, the strength of which predicted participants’ order memory accuracy. During retrieval, OSC spike timing relative to theta varied with the relative position of their preferred event within the recalled event sequence, enabling selective retrieval of relevant events. These findings reveal a neural substrate for representing, encoding into and retrieving from memory absolute and relative ordinal relationships between discrete events. OSCs tile temporal space into discrete ordinal positions, thereby weaving episodic experiences into coherent temporal narratives.