Closed-loop modulation of remote hippocampal representations with neurofeedback

Humans can remember specific events without acting on them and can influence which memories are retrieved based on internal goals. However, current animal models of memory typically present sensory cues to trigger retrieval and assess retrieval based on action 1-5 . As a result, it is difficult to determine whether measured patterns of neural activity relate to the cue(s), the retrieved memory, or the behavior. We therefore asked whether we could develop a paradigm to isolate retrieval-related neural activity in animals without retrieval cues or the requirement of a behavioral report. To do this, we focused on hippocampal "place cells." These cells primarily emit spiking patterns that represent the animal's current location (local representations), but they can also generate representations of previously visited locations distant from the animal's current location (remote representations) 6-13 . It is not known whether animals can deliberately engage specific remote representations, and if so, whether this engagement would occur during specific brain states. So, we used a closed-loop neurofeedback system to reward expression of remote representations that corresponded to uncued, experimenter-selected locations, and found that rats could increase the prevalence of these specific remote representations over time; thus, demonstrating memory retrieval modulated by internal goals in an animal model. These representations occurred predominately during periods of immobility but outside of hippocampal sharp-wave ripple (SWR) 13-15 events. This paradigm enables future direct studies of memory retrieval mechanisms in the healthy brain and in models of neurological disorders.

Thalamic epileptic spikes disrupt sleep spindles in patients with epileptic encephalopathy

In severe epileptic encephalopathies, epileptic activity contributes to progressive cognitive dysfunction. Epileptic encephalopathies share the trait of spike-wave activation during non-rapid eye movement sleep (EE-SWAS), a sleep stage dominated by sleep spindles, brain oscillations known to coordinate offline memory consolidation. Epileptic activity has been proposed to hijack the circuits driving these thalamocortical oscillations, thereby contributing to cognitive impairment. Using a unique dataset of simultaneous human thalamic and cortical recordings in subjects with and without EE-SWAS, we provide evidence for epileptic spike interference of thalamic sleep spindle production in patients with EE-SWAS. First, we show that epileptic spikes and sleep spindles are both predicted by slow oscillations during stage two sleep (N2), but at different phases of the slow oscillation. Next, we demonstrate that sleep activated cortical epileptic spikes propagate to the thalamus (thalamic spike rate increases after a cortical spike, p≈0). We then show that epileptic spikes in the thalamus increase the thalamic spindle refractory period (p≈0). Finally, we show that in three patients with EE-SWAS, there is a downregulation of sleep spindles for 30 seconds after each thalamic spike (p<0.01). These direct human thalamocortical observations support a proposed mechanism for epileptiform activity to impact cognitive function, wherein epileptic spikes inhibit thalamic sleep spindles in epileptic encephalopathy with spike and wave activation during sleep.

Photothrombosis induced cortical stroke produces electrographic epileptic biomarkers in mice

Objective

Interictal epileptiform spikes, high-frequency ripple oscillations, and their co-occurrence (spike ripples) in human scalp or intracranial voltage recordings are well-established epileptic biomarkers. While clinically significant, the neural mechanisms generating these electrographic biomarkers remain unclear. To reduce this knowledge gap, we introduce a novel photothrombotic stroke model in mice that reproduces focal interictal electrographic biomarkers observed in human epilepsy.

Methods

We induced a stroke in the motor cortex of C57BL/6 mice unilaterally (N=7) using a photothrombotic procedure previously established in rats. We then implanted intracranial electrodes (2 ipsilateral and 2 contralateral) and obtained intermittent local field potential (LFP) recordings over several weeks in awake, behaving mice. We evaluated the LFP for focal slowing and epileptic biomarkers - spikes, ripples, and spike ripples - using both automated and semi-automated procedures.

Results

Delta power (1-4 Hz) was higher in the stroke hemisphere than the non-stroke hemisphere in all mice ( p <0.001). Automated detection procedures indicated that compared to the non-stroke hemisphere, the stroke hemisphere had an increased spike ripple ( p =0.006) and spike rates ( p =0.039), but no change in ripple rate ( p =0.98). Expert validation confirmed the observation of elevated spike ripple rates ( p =0.008) and a trend of elevated spike rate ( p =0.055) in the stroke hemisphere. Interestingly, the validated ripple rate in the stroke hemisphere was higher than the non-stroke hemisphere ( p =0.031), highlighting the difficulty of automatically detecting ripples. Finally, using optimal performance thresholds, automatically detected spike ripples classified the stroke hemisphere with the best accuracy (sensitivity 0.94, specificity 0.94).

Significance

Cortical photothrombosis-induced stroke in commonly used C57BL/6 mice produces electrographic biomarkers as observed in human epilepsy. This model represents a new translational cortical epilepsy model with a defined irritative zone, which can be broadly applied in transgenic mice for cell type specific analysis of the cellular and circuit mechanisms of pathologic interictal activity.

Key Points

Cortical photothrombosis in mice produces stroke with characteristic intermittent focal delta slowing.Cortical photothrombosis stroke in mice produces the epileptic biomarkers spikes, ripples, and spike ripples.All biomarkers share morphological features with the corresponding human correlate.Spike ripples better lateralize to the lesional cortex than spikes or ripples.This cortical model can be applied in transgenic mice for mechanistic studies.

Individualized temporal patterns dominate cortical upstate and sleep depth in driving human sleep spindle timing

Sleep spindles are critical for memory consolidation and strongly linked to neurological disease and aging. Despite their significance, the relative influences of factors like sleep depth, cortical up/down states, and spindle temporal patterns on individual spindle production remain poorly understood. Moreover, spindle temporal patterns are typically ignored in favor of an average spindle rate. Here, we analyze spindle dynamics in 1008 participants from the Multi-Ethnic Study of Atherosclerosis using a point process framework. Results reveal fingerprint-like temporal patterns, characterized by a refractory period followed by a period of increased spindle activity, which are highly individualized yet consistent night-to-night. We observe increased timing variability with age and distinct gender/age differences. Strikingly, and in contrast to the prevailing notion, individualized spindle patterns are the dominant determinant of spindle timing, accounting for over 70% of the statistical deviance explained by all of the factors we assessed, surpassing the contribution of slow oscillation (SO) phase (~14%) and sleep depth (~16%). Furthermore, we show spindle/SO coupling dynamics with sleep depth are preserved across age, with a global negative shift towards the SO rising slope. These findings offer novel mechanistic insights into spindle dynamics with direct experimental implications and applications to individualized electroencephalography biomarker identification.

Spike ripples localize the epileptogenic zone best: an international intracranial study

We evaluated whether spike ripples, the combination of epileptiform spikes and ripples, provide a reliable and improved biomarker for the epileptogenic zone (EZ) compared to other leading interictal biomarkers in a multicenter, international study. We first validated an automated spike ripple detector on intracranial EEG recordings. We then applied this detector to subjects from four centers who subsequently underwent surgical resection with known 1-year outcomes. We evaluated the spike ripple rate in subjects cured after resection (ILAE 1 outcome) and those with persistent seizures (ILAE 2-6) across sites and recording types. We also evaluated available interictal biomarkers: spike, spike-gamma, wideband high frequency oscillation (HFO, 80-500 Hz), ripple (80-250 Hz), and fast ripple (250-500 Hz) rates using previously validated automated detectors. The proportion of resected events was computed and compared across subject outcomes and biomarkers. 109 subjects were included. Most spike ripples were removed in subjects with ILAE 1 outcome (P < 0.001), and this was qualitatively observed across all sites and for depth and subdural electrodes (P < 0.001, P < 0.001). Among ILAE 1 subjects, the mean spike ripple rate was higher in the RV (0.66/min) than in the non-removed tissue (0.08/min, P < 0.001). A higher proportion of spike ripples were removed in subjects with ILAE 1 outcomes compared to ILAE 2-6 outcomes (P = 0.06). Among ILAE 1 subjects, the proportion of spike ripples removed was higher than the proportion of spikes (P < 0.001), spike-gamma (P < 0.001), wideband HFOs (P < 0.001), ripples (P = 0.009) and fast ripples (P = 0.009) removed. At the individual level, more subjects with ILAE 1 outcomes had the majority of spike ripples removed (79%, 38/48) than spikes (69%, P = 0.12), spike-gamma (69%, P = 0.12), wideband HFOs (63%, P = 0.03), ripples (45%, P = 0.01), or fast ripples (36%, P < 0.001) removed. Thus, in this large, multicenter cohort, when surgical resection was successful, the majority of spike ripples were removed. Further, automatically detected spike ripples have improved specificity for epileptogenic tissue compared to spikes, spike-gamma, wideband HFOs, ripples, and fast ripples.

Different Methods to Estimate the Phase of Neural Rhythms Agree But Only During Times of Low Uncertainty

Rhythms are a common feature of brain activity. Across different types of rhythms, the phase has been proposed to have functional consequences, thus requiring its accurate specification from noisy data. Phase is conventionally specified using techniques that presume a frequency band-limited rhythm. However, in practice, observed brain rhythms are typically nonsinusoidal and amplitude modulated. How these features impact methods to estimate phase remains unclear. To address this, we consider three phase estimation methods, each with different underlying assumptions about the rhythm. We apply these methods to rhythms simulated with different generative mechanisms and demonstrate inconsistency in phase estimates across the different methods. We propose two improvements to the practice of phase estimation: (1) estimating confidence in the phase estimate, and (2) examining the consistency of phase estimates between two (or more) methods.

Different methods to estimate the phase of neural rhythms agree, but only during times of low uncertainty

Rhythms are a common feature of brain activity. Across different types of rhythms, the phase has been proposed to have functional consequences, thus requiring its accurate specification from noisy data. Phase is conventionally specified using techniques that presume a frequency band-limited rhythm. However, in practice, observed brain rhythms are typically non-sinusoidal and amplitude modulated. How these features impact methods to estimate phase remains unclear. To address this, we consider three phase estimation methods, each with different underlying assumptions about the rhythm. We apply these methods to rhythms simulated with different generative mechanisms and demonstrate inconsistency in phase estimates across the different methods. We propose two improvements to the practice of phase estimation: (1) estimating confidence in the phase estimate, and (2) examining the consistency of phase estimates between two (or more) methods.

Sleep spindles in the healthy brain from birth through 18 years

STUDY OBJECTIVE

Sleep spindles are present from birth and reflect cognitive functions across the lifespan, but normative values for this cognitive biomarker across development are lacking. This study aims to establish normative spindle features over development.

METHODS

All available normal 19-channel electroencephalograms from developmentally normal children between February 2002 and June 2021 in the MGH EEG lab were analyzed. Approximately, 20 000 spindles were hand-marked to train and validate an automated spindle detector across ages. Normative values for spindle rate, duration, frequency, refractory period, and interhemispheric lag are provided for each channel and each age.

RESULTS

Sleep EEGs from 567 developmentally normal children (range 0 days to 18 years) were included. The detector had excellent performance (F1 = 0.47). Maximal spindle activity is seen over central regions during infancy and adolescence and frontopolar regions during childhood. Spindle rate and duration increase nonlinearly, with the most rapid changes during the first 4 months of life and between ages 3 and 14 years. Peak spindle frequency follows a U-shaped curve and discrete frontal slow and central fast spindles are evident by 18 months. Spindle refractory periods decrease between ages 1 and 14 years while interhemispheric asynchrony decreases over the first 3 months of life and between ages 1 and 14 years.

CONCLUSIONS

These data provide age- and region-specific normative values for sleep spindles across development, where measures that deviate from these values can be considered pathological. As spindles provide a noninvasive biomarker for cognitive function across the lifespan, these normative measures can accelerate the discovery and diagnosis in neurodevelopmental disorders.

Closed-loop enhancement and neural decoding of cognitive control in humans

Deficits in cognitive control-that is, in the ability to withhold a default pre-potent response in favour of a more adaptive choice-are common in depression, anxiety, addiction and other mental disorders. Here we report proof-of-concept evidence that, in participants undergoing intracranial epilepsy monitoring, closed-loop direct stimulation of the internal capsule or striatum, especially the dorsal sites, enhances the participants' cognitive control during a conflict task. We also show that closed-loop stimulation upon the detection of lapses in cognitive control produced larger behavioural changes than open-loop stimulation, and that task performance for single trials can be directly decoded from the activity of a small number of electrodes via neural features that are compatible with existing closed-loop brain implants. Closed-loop enhancement of cognitive control might remediate underlying cognitive deficits and aid the treatment of severe mental disorders.

Integrating Statistical and Machine Learning Approaches for Neural Classification

Neurons can code for multiple variables simultaneously and neuroscientists are often interested in classifying neurons based on their receptive field properties. Statistical models provide powerful tools for determining the factors influencing neural spiking activity and classifying individual neurons. However, as neural recording technologies have advanced to produce simultaneous spiking data from massive populations, classical statistical methods often lack the computational efficiency required to handle such data. Machine learning (ML) approaches are known for enabling efficient large scale data analyses; however, they typically require massive training sets with balanced data, along with accurate labels to fit well. Additionally, model assessment and interpretation are often more challenging for ML than for classical statistical methods. To address these challenges, we develop an integrated framework, combining statistical modeling and machine learning approaches to identify the coding properties of neurons from large populations. In order to demonstrate this framework, we apply these methods to data from a population of neurons recorded from rat hippocampus to characterize the distribution of spatial receptive fields in this region.

Multiple Sources of Fast Traveling Waves during Human Seizures: Resolving a Controversy

During human seizures, organized waves of voltage activity rapidly sweep across the cortex. Two contradictory theories describe the source of these fast traveling waves: either a slowly advancing narrow region of multiunit activity (an ictal wavefront) or a fixed cortical location. Limited observations and different analyses prevent resolution of these incompatible theories. Here we address this disagreement by combining the methods and microelectrode array recordings (N = 11 patients, 2 females, N = 31 seizures) from previous human studies to analyze the traveling wave source. We find, inconsistent with both existing theories, a transient relationship between the ictal wavefront and traveling waves, and multiple stable directions of traveling waves in many seizures. Using a computational model that combines elements of both existing theories, we show that interactions between an ictal wavefront and fixed source reproduce the traveling wave dynamics observed in vivo We conclude that combining both existing theories can generate the diversity of ictal traveling waves.SIGNIFICANCE STATEMENT The source of voltage discharges that propagate across cortex during human seizures remains unknown. Two candidate theories exist, each proposing a different discharge source. Support for each theory consists of observations from a small number of human subject recordings, analyzed with separately developed methods. How the different, limited data and different analysis methods impact the evidence for each theory is unclear. To resolve these differences, we combine the unique, human microelectrode array recordings collected separately for each theory and analyze these combined data with a unified approach. We show that neither existing theory adequately describes the data. We then propose a new theory that unifies existing proposals and successfully reproduces the voltage discharge dynamics observed in vivo.

Dynamic models of obstructive sleep apnea provide robust prediction of respiratory event timing and a statistical framework for phenotype exploration

Obstructive sleep apnea (OSA), in which breathing is reduced or ceased during sleep, affects at least 10% of the population and is associated with numerous comorbidities. Current clinical diagnostic approaches characterize severity and treatment eligibility using the average respiratory event rate over total sleep time (apnea-hypopnea index). This approach, however, does not characterize the time-varying and dynamic properties of respiratory events that can change as a function of body position, sleep stage, and previous respiratory event activity. Here, we develop a statistical model framework based on point process theory that characterizes the relative influences of all these factors on the moment-to-moment rate of event occurrence. Our results provide new insights into the temporal dynamics of respiratory events, suggesting that most adults have a characteristic event pattern that involves a period of normal breathing followed by a period of increased probability of respiratory event occurrence, while significant differences in event patterns are observed among gender, age, and race/ethnicity groups. Statistical goodness-of-fit analysis suggests consistent and substantial improvements in our ability to capture the timing of individual respiratory events using our modeling framework. Overall, we demonstrate a more statistically robust approach to characterizing sleep disordered breathing that can also serve as a basis for identifying future patient-specific respiratory phenotypes, providing an improved pathway towards developing individualized treatments.

Direct Discriminative Decoder Models for Analysis of High-Dimensional Dynamical Neural Data

With the accelerated development of neural recording technology over the past few decades, research in integrative neuroscience has become increasingly reliant on data analysis methods that are scalable to high-dimensional recordings and computationally tractable. Latent process models have shown promising results in estimating the dynamics of cognitive processes using individual models for each neuron's receptive field. However, scaling these models to work on high-dimensional neural recordings remains challenging. Not only is it impractical to build receptive field models for individual neurons of a large neural population, but most neural data analyses based on individual receptive field models discard the local history of neural activity, which has been shown to be critical in the accurate inference of the underlying cognitive processes. Here, we propose a novel, scalable latent process model that can directly estimate cognitive process dynamics without requiring precise receptive field models of individual neurons or brain nodes. We call this the direct discriminative decoder (DDD) model. The DDD model consists of (1) a discriminative process that characterizes the conditional distribution of the signal to be estimated, or state, as a function of both the current neural activity and its local history, and (2) a state transition model that characterizes the evolution of the state over a longer time period. While this modeling framework inherits advantages of existing latent process modeling methods, its computational cost is tractable. More important, the solution can incorporate any information from the history of neural activity at any timescale in computing the estimate of the state process. There are many choices in building the discriminative process, including deep neural networks or gaussian processes, which adds to the flexibility of the framework. We argue that these attributes of the proposed methodology, along with its applicability to different modalities of neural data, make it a powerful tool for high-dimensional neural data analysis. We also introduce an extension of these methods, called the discriminative-generative decoder (DGD). The DGD includes both discriminative and generative processes in characterizing observed data. As a result, we can combine physiological correlates like behavior with neural data to better estimate underlying cognitive processes. We illustrate the methods, including steps for inference and model identification, and demonstrate applications to multiple data analysis problems with high-dimensional neural recordings. The modeling results demonstrate the computational and modeling advantages of the DDD and DGD methods.

Source EEG reveals that Rolandic epilepsy is a regional epileptic encephalopathy

•

Children with RE have fewer spindles but they have typical time–frequency features.

•

Spindle deficits extend to multiple cortical regions in Rolandic epilepsy.

•

Cognitive deficits are predicted by spindle rate in Rolandic epilepsy.

•

Regional spindle rate predicts motor deficits better than Rolandic spindle deficit.

•

Spindle features in RE identify a regional thalamocortical epileptic encephalopathy.

Rolandic epilepsy is the most common form of epileptic encephalopathy, characterized by sleep-potentiated inferior Rolandic epileptiform spikes, seizures, and cognitive deficits in school-age children that spontaneously resolve by adolescence. We recently identified a paucity of sleep spindles, physiological thalamocortical rhythms associated with sleep-dependent learning, in the Rolandic cortex during the active phase of this disease. Because spindles are generated in the thalamus and amplified through regional thalamocortical circuits, we hypothesized that: 1) deficits in spindle rate would involve but extend beyond the inferior Rolandic cortex in active epilepsy and 2) regional spindle deficits would better predict cognitive function than inferior Rolandic spindle deficits alone. To test these hypotheses, we obtained high-resolution MRI, high-density EEG recordings, and focused neuropsychological assessments in children with Rolandic epilepsy during active (n = 8, age 9–14.7 years, 3F) and resolved (seizure free for > 1 year, n = 10, age 10.3–16.7 years, 1F) stages of disease and age-matched controls (n = 8, age 8.9–14.5 years, 5F). Using a validated spindle detector applied to estimates of electrical source activity in 31 cortical regions, including the inferior Rolandic cortex, during stages 2 and 3 of non-rapid eye movement sleep, we compared spindle rates in each cortical region across groups. Among detected spindles, we compared spindle features (power, duration, coherence, bilateral synchrony) between groups. We then used regression models to examine the relationship between spindle rate and cognitive function (fine motor dexterity, phonological processing, attention, and intelligence, and a global measure of all functions). We found that spindle rate was reduced in the inferior Rolandic cortices in active but not resolved disease (active P = 0.007; resolved P = 0.2) compared to controls. Spindles in this region were less synchronous between hemispheres in the active group (P = 0.005; resolved P = 0.1) compared to controls; but there were no differences in spindle power, duration, or coherence between groups. Compared to controls, spindle rate in the active group was also reduced in the prefrontal, insular, superior temporal, and posterior parietal regions (i.e., “regional spindle rate”, P < 0.039 for all). Independent of group, regional spindle rate positively correlated with fine motor dexterity (P < 1e-3), attention (P = 0.02), intelligence (P = 0.04), and global cognitive performance (P < 1e-4). Compared to the inferior Rolandic spindle rate alone, models including regional spindle rate trended to improve prediction of global cognitive performance (P = 0.052), and markedly improved prediction of fine motor dexterity (P = 0.006). These results identify a spindle disruption in Rolandic epilepsy that extends beyond the epileptic cortex and a potential mechanistic explanation for the broad cognitive deficits that can be observed in this epileptic encephalopathy.

Hippocampal replay reflects specific past experiences rather than a plan for subsequent choice

Executing memory-guided behavior requires storage of information about experience and later recall of that information to inform choices. Awake hippocampal replay, when hippocampal neural ensembles briefly reactivate a representation related to prior experience, has been proposed to critically contribute to these memory-related processes. However, it remains unclear whether awake replay contributes to memory function by promoting the storage of past experiences, facilitating planning based on evaluation of those experiences, or both. We designed a dynamic spatial task that promotes replay before a memory-based choice and assessed how the content of replay related to past and future behavior. We found that replay content was decoupled from subsequent choice and instead was enriched for representations of previously rewarded locations and places that had not been visited recently, indicating a role in memory storage rather than in directly guiding subsequent behavior.

A state space modeling approach to real-time phase estimation

Brain rhythms have been proposed to facilitate brain function, with an especially important role attributed to the phase of low-frequency rhythms. Understanding the role of phase in neural function requires interventions that perturb neural activity at a target phase, necessitating estimation of phase in real-time. Current methods for real-time phase estimation rely on bandpass filtering, which assumes narrowband signals and couples the signal and noise in the phase estimate, adding noise to the phase and impairing detections of relationships between phase and behavior. To address this, we propose a state space phase estimator for real-time tracking of phase. By tracking the analytic signal as a latent state, this framework avoids the requirement of bandpass filtering, separately models the signal and the noise, accounts for rhythmic confounds, and provides credible intervals for the phase estimate. We demonstrate in simulations that the state space phase estimator outperforms current state-of-the-art real-time methods in the contexts of common confounds such as broadband rhythms, phase resets, and co-occurring rhythms. Finally, we show applications of this approach to in vivo data. The method is available as a ready-to-use plug-in for the Open Ephys acquisition system, making it widely available for use in experiments.

Hippocampal replay of experience at real-world speeds

Representations related to past experiences play a critical role in memory and decision-making processes. The rat hippocampus expresses these types of representations during sharp-wave ripple (SWR) events, and previous work identified a minority of SWRs that contain ‘replay’ of spatial trajectories at ∼20x the movement speed of the animal. Efforts to understand replay typically make multiple assumptions about which events to examine and what sorts of representations constitute replay. We therefore lack a clear understanding of both the prevalence and the range of representational dynamics associated with replay. Here, we develop a state space model that uses a combination of movement dynamics of different speeds to capture the spatial content and time evolution of replay during SWRs. Using this model, we find that the large majority of replay events contain spatially coherent, interpretable content. Furthermore, many events progress at real-world, rather than accelerated, movement speeds, consistent with actual experiences.

Application of a convolutional neural network for fully-automated detection of spike ripples in the scalp electroencephalogram

BACKGROUND

A reliable biomarker to identify cortical tissue responsible for generating epileptic seizures is required to guide prognosis and treatment in epilepsy. Combined spike ripple events are a promising biomarker for epileptogenic tissue that currently require expert review for accurate identification. This expert review is time consuming and subjective, limiting reproducibility and high-throughput applications.

NEW METHOD

To address this limitation, we develop a fully-automated method for spike ripple detection. The method consists of a convolutional neural network trained to compute the probability that a spectrogram image contains a spike ripple.

RESULTS

We validate the proposed spike ripple detector on expert-labeled data and show that this detector accurately separates subjects with low and high seizure risks.

COMPARISON WITH EXISTING METHOD

The proposed method performs as well as existing methods that require manual validation of candidate spike ripple events. The introduction of a fully automated method reduces subjectivity and increases rigor and reproducibility of this epilepsy biomarker.

CONCLUSION

We introduce and validate a fully-automated spike ripple detector to support utilization of this epilepsy biomarker in clinical and translational work.

Real-Time Point Process Filter for Multidimensional Decoding Problems Using Mixture Models

There is an increasing demand for a computationally efficient and accurate point process filter solution for real-time decoding of population spiking activity in multidimensional spaces. Real-time tools for neural data analysis, specifically real-time neural decoding solutions open doors for developing experiments in a closed-loop setting and more versatile brain-machine interfaces. Over the past decade, the point process filter has been successfully applied in the decoding of behavioral and biological signals using spiking activity of an ensemble of cells; however, the filter solution is computationally expensive in multi-dimensional filtering problems. Here, we propose an approximate filter solution for a general point-process filter problem when the conditional intensity of a cell's spiking activity is characterized using a Mixture of Gaussians. We propose the filter solution for a broader class of point process observation called marked point-process, which encompasses both clustered - mainly, called sorted - and clusterless - generally called unsorted or raw- spiking activity. We assume that the posterior distribution on each filtering time-step can be approximated using a Gaussian Mixture Model and propose a computationally efficient algorithm to estimate the optimal number of mixture components and their corresponding weights, mean, and covariance estimates. This algorithm provides a real-time solution for multi-dimensional point-process filter problem and attains accuracy comparable to the exact solution. Our solution takes advantage of mixture dropping and merging algorithms, which collectively control the growth of mixture components on each filtering time-step. We apply this methodology in decoding a rat's position in both 1-D and 2-D spaces using clusterless spiking data of an ensemble of rat hippocampus place cells. The approximate solution in 1-D and 2-D decoding is more than 20 and 4,000 times faster than the exact solution, while their accuracy in decoding a rat position only drops by less than 9% and 4% in RMSE and 95% highest probability coverage area (HPD) performance metrics. Though the marked-point filter solution is better suited for real-time decoding problems, we discuss how the filter solution can be applied to sorted spike data to better reflect the proposed methodology versatility.

Assessing Goodness-of-Fit in Marked Point Process Models of Neural Population Coding via Time and Rate Rescaling

Marked point process models have recently been used to capture the coding properties of neural populations from multiunit electrophysiological recordings without spike sorting. These clusterless models have been shown in some instances to better describe the firing properties of neural populations than collections of receptive field models for sorted neurons and to lead to better decoding results. To assess their quality, we previously proposed a goodness-of-fit technique for marked point process models based on time rescaling, which for a correct model produces a set of uniform samples over a random region of space. However, assessing uniformity over such a region can be challenging, especially in high dimensions. Here, we propose a set of new transformations in both time and the space of spike waveform features, which generate events that are uniformly distributed in the new mark and time spaces. These transformations are scalable to multidimensional mark spaces and provide uniformly distributed samples in hypercubes, which are well suited for uniformity tests. We discuss the properties of these transformations and demonstrate aspects of model fit captured by each transformation. We also compare multiple uniformity tests to determine their power to identify lack-of-fit in the rescaled data. We demonstrate an application of these transformations and uniformity tests in a simulation study. Proofs for each transformation are provided in the appendix.

Continuous Prediction of Cognitive State Using A Marked-Point Process Modeling Framework

Behavioral outcomes in many cognitive tasks are often recorded in a trial structure at discrete times. To adapt to this structure, neural encoder and decoder models have been built to take into account the trial organization to characterize the connection between brain dynamics and behavior, e.g. through latent dynamical models. The challenge of these models is that they are limited to discrete trial times while neural data is continuous. Here, we propose a marked-point process framework to characterize multivariate behavioral outcomes recorded during a trial-structured cognitive task, to build an estimation of cognitive state at a fine time resolution. We propose a state-space marked-point process modeling framework to characterize the relationship between observed behavior and underlying dynamical cognitive processes. We define the framework for a class of behavioral readouts by a response time and a discrete mark signifying an observed binary decision, and develop the state estimation and system identification steps. We define the filter and smoother for the marked-point process observation and develop an EM algorithm to estimate the model's free parameters. We demonstrate this modeling approach in a behavioral readout captured while participants perform an emotional conflict resolution task (ECR) and show that we can estimate underlying cognitive processes at a fine temporal resolution beyond the trial by trial approach.

State-Space Global Coherence to Estimate the Spatio-Temporal Dynamics of the Coordinated Brain Activity

Characterizing coordinated brain dynamics present in high-density neural recordings is critical for understanding the neurophysiology of healthy and pathological brain states and to develop principled strategies for therapeutic interventions. In this research, we propose a new modeling framework called State Space Global Coherence (SSGC), which allows us to estimate neural synchrony across distributed brain activity with fine temporal resolution. In this modeling framework, the cross-spectral matrix of neural activity at a specific frequency is defined as a function of a dynamical state variable representing a measure of Global Coherence (GC); we then combine filter-smoother and Expectation-Maximization (EM) algorithms to estimate GC and the model parameters. We demonstrate a SSGC analysis in a 64-channel EEG recording of a human subject under general anesthesia and compare the modeling result with empirical measures of GC. We show that SSGC not only attains a finer time resolution but also provides more accurate estimation of GC.

Parameter Estimation in Synaptic Coupling Model Using a Point Process Modeling Framework

Biophysical models are widely used to characterize temporal dynamics of the brain networks on different topological and spatial scales. In parallel, the state-space modeling framework with point process observations has been successfully applied in characterizing spiking activity of neuronal ensembles in response to different dynamical covariates. Parameter estimation in biophysical models is generally done heuristically, which hampers their applicability and interpretability. Heuristic parameter estimation becomes an intractable problem when the number of model parameters grows. Here, we propose an algorithm for estimating biophysical model parameters using point-process models and a state-space framework. The framework provides methods for parameter estimation as well as model validation. We demonstrate the application of this methodology to the problem of estimating the parameters of a dynamic synapse model. We generate simulation data for the dynamic synapse across a range of parameters values and assess the estimation accuracy of our method using a combination of goodness-of-fit measures. The proposed methodology can be applied broadly to parameter estimation problems across a broad range of biophysical models, including Hodgkin-Huxley models and network models.

A statistical framework to assess cross-frequency coupling while accounting for confounding analysis effects

Cross frequency coupling (CFC) is emerging as a fundamental feature of brain activity, correlated with brain function and dysfunction. Many different types of CFC have been identified through application of numerous data analysis methods, each developed to characterize a specific CFC type. Choosing an inappropriate method weakens statistical power and introduces opportunities for confounding effects. To address this, we propose a statistical modeling framework to estimate high frequency amplitude as a function of both the low frequency amplitude and low frequency phase; the result is a measure of phase-amplitude coupling that accounts for changes in the low frequency amplitude. We show in simulations that the proposed method successfully detects CFC between the low frequency phase or amplitude and the high frequency amplitude, and outperforms an existing method in biologically-motivated examples. Applying the method to in vivo data, we illustrate examples of CFC during a seizure and in response to electrical stimuli.

Decoding Hidden Cognitive States From Behavior and Physiology Using a Bayesian Approach

Cognitive processes, such as learning and cognitive flexibility, are both difficult to measure and to sample continuously using objective tools because cognitive processes arise from distributed, high-dimensional neural activity. For both research and clinical applications, that dimensionality must be reduced. To reduce dimensionality and measure underlying cognitive processes, we propose a modeling framework in which a cognitive process is defined as a low-dimensional dynamical latent variable—called a cognitive state, which links high-dimensional neural recordings and multidimensional behavioral readouts. This framework allows us to decompose the hard problem of modeling the relationship between neural and behavioral data into separable encoding-decoding approaches. We first use a state-space modeling framework, the behavioral decoder, to articulate the relationship between an objective behavioral readout (e.g., response times) and cognitive state. The second step, the neural encoder, involves using a generalized linear model (GLM) to identify the relationship between the cognitive state and neural signals, such as local field potential (LFP). We then use the neural encoder model and a Bayesian filter to estimate cognitive state using neural data (LFP power) to generate the neural decoder. We provide goodness-of-fit analysis and model selection criteria in support of the encoding-decoding result. We apply this framework to estimate an underlying cognitive state from neural data in human participants ([Formula: see text]) performing a cognitive conflict task. We successfully estimated the cognitive state within the 95% confidence intervals of that estimated using behavior readout for an average of 90% of task trials across participants. In contrast to previous encoder-decoder models, our proposed modeling framework incorporates LFP spectral power to encode and decode a cognitive state. The framework allowed us to capture the temporal evolution of the underlying cognitive processes, which could be key to the development of closed-loop experiments and treatments.

Dysmature superficial white matter microstructure in developmental focal epilepsy

Benign epilepsy with centrotemporal spikes is a common childhood epilepsy syndrome that predominantly affects boys, characterized by self-limited focal seizures arising from the perirolandic cortex and fine motor abnormalities. Concurrent with the age-specific presentation of this syndrome, the brain undergoes a developmentally choreographed sequence of white matter microstructural changes, including maturation of association u-fibres abutting the cortex. These short fibres mediate local cortico-cortical communication and provide an age-sensitive structural substrate that could support a focal disease process. To test this hypothesis, we evaluated the microstructural properties of superficial white matter in regions corresponding to u-fibres underlying the perirolandic seizure onset zone in children with this epilepsy syndrome compared with healthy controls. To verify the spatial specificity of these features, we characterized global superficial and deep white matter properties. We further evaluated the characteristics of the perirolandic white matter in relation to performance on a fine motor task, gender and abnormalities observed on EEG. Children with benign epilepsy with centrotemporal spikes (n = 20) and healthy controls (n = 14) underwent multimodal testing with high-resolution MRI including diffusion tensor imaging sequences, sleep EEG recordings and fine motor assessment. We compared white matter microstructural characteristics (axial, radial and mean diffusivity, and fractional anisotropy) between groups in each region. We found distinct abnormalities corresponding to the perirolandic u-fibre region, with increased axial, radial and mean diffusivity and fractional anisotropy values in children with epilepsy (P = 0.039, P = 0.035, P = 0.042 and P = 0.017, respectively). Increased fractional anisotropy in this region, consistent with decreased integrity of crossing sensorimotor u-fibres, correlated with inferior fine motor performance (P = 0.029). There were gender-specific differences in white matter microstructure in the perirolandic region; males and females with epilepsy and healthy males had higher diffusion and fractional anisotropy values than healthy females (P ≤ 0.035 for all measures), suggesting that typical patterns of white matter development disproportionately predispose boys to this developmental epilepsy syndrome. Perirolandic white matter microstructure showed no relationship to epilepsy duration, duration seizure free, or epileptiform burden. There were no group differences in diffusivity or fractional anisotropy in superficial white matter outside of the perirolandic region. Children with epilepsy had increased radial diffusivity (P = 0.022) and decreased fractional anisotropy (P = 0.027) in deep white matter, consistent with a global delay in white matter maturation. These data provide evidence that atypical maturation of white matter microstructure is a basic feature in benign epilepsy with centrotemporal spikes and may contribute to the epilepsy, male predisposition and clinical comorbidities observed in this disorder.

Quantifying statistical uncertainty in metrics of sleep disordered breathing

BACKGROUND

The apnea-hypopnea index (AHI) (or one of its derivatives) is the primary clinical metric for characterizing sleep disordered breathing-the value of which with respect to a threshold determines severity of diagnosis and eligibility for treatment reimbursement. The index value, however, is taken as a perfect point estimate, with no measure of statistical uncertainty. Thus, current practice does not robustly account for variability in diagnosis/eligibility due to chance. In this paper, we quantify the statistical uncertainty associated with respiratory event indices for sleep disordered breathing and the effect of uncertainty on treatment eligibility.

METHODS

We develop an empirical estimate of uncertainty using a non-parametric bootstrap on the interevent times, as well as a theoretical Poisson estimate reflecting the current formulation of the AHI. We then apply these methods to estimate AHI uncertainty for 2049 subjects (954/1095 M/F, age: mean 69 ± 9.1) from the Multi-Ethnic Study of Atherosclerosis (MESA).

RESULTS AND CONCLUSIONS

The mean 95% empirical confidence interval width was 11.500 ± 6.208 events per hour and the mean 95% theoretical Poisson confidence interval width was 5.998 ± 2.897 events per hour, suggesting that uncertainty is likely a major confounding factor within the current diagnostic framework. Of the 278 subjects in the symptomatic population (ESS>10), 27% (76/278) had uncertain diagnoses given the 95% empirical confidence interval. Of the 2049 subjects in the full population, 43% (880/2049) had uncertain diagnoses given the 95% empirical confidence interval. The inclusion of subjects with uncertain diagnoses increases the number of eligible patients by 21.3% for the symptomatic population and by 84.8% for the full population. The exclusion of subjects with uncertain diagnoses given the 95% empirical confidence interval decreases the number of eligible patients by 12.4% for the symptomatic population and by 34.8% for full population. Additional analyses suggest that it is practically infeasible to gain diagnostic statistical significance through additional testing for a broad range of borderline cases. Overall, these results suggest that AHI uncertainty is a vital additional piece of information that would greatly benefit clinical practice, and that the inclusion of uncertainty in epidemiological analysis might help improve the ability for researchers to robustly link AHI with co-morbidities and long-term outcomes.

Scalp recorded spike ripples predict seizure risk in childhood epilepsy better than spikes

In the past decade, brief bursts of fast oscillations in the ripple range have been identified in the scalp EEG as a promising non-invasive biomarker for epilepsy. However, investigation and clinical application of this biomarker have been limited because standard approaches to identify these brief, low amplitude events are difficult, time consuming, and subjective. Recent studies have demonstrated that ripples co-occurring with epileptiform discharges ('spike ripple events') are easier to detect than ripples alone and have greater pathological significance. Here, we used objective techniques to quantify spike ripples and test whether this biomarker predicts seizure risk in childhood epilepsy. We evaluated spike ripples in scalp EEG recordings from a prospective cohort of children with a self-limited epilepsy syndrome, benign epilepsy with centrotemporal spikes, and healthy control children. We compared the rate of spike ripples between children with epilepsy and healthy controls, and between children with epilepsy during periods of active disease (active, within 1 year of seizure) and after a period of sustained seizure-freedom (seizure-free, >1 year without seizure), using semi-automated and automated detection techniques. Spike ripple rate was higher in subjects with active epilepsy compared to healthy controls (P = 0.0018) or subjects with epilepsy who were seizure-free ON or OFF medication (P = 0.0018). Among epilepsy subjects with spike ripples, each month seizure-free decreased the odds of a spike ripple by a factor of 0.66 [95% confidence interval (0.47, 0.91), P = 0.021]. Comparing the diagnostic accuracy of the presence of at least one spike ripple versus a classic spike event to identify group, we found comparable sensitivity and negative predictive value, but greater specificity and positive predictive value of spike ripples compared to spikes (P = 0.016 and P = 0.006, respectively). We found qualitatively consistent results using a fully automated spike ripple detector, including comparison with an automated spike detector. We conclude that scalp spike ripple events identify disease and track with seizure risk in this epilepsy population, using both semi-automated and fully automated detection methods, and that this biomarker outperforms analysis of spikes alone in categorizing seizure risk. These data provide evidence that spike ripples are a specific non-invasive biomarker for seizure risk in benign epilepsy with centrotemporal spikes and support future work to evaluate the utility of this biomarker to guide medication trials and tapers in these children and predict seizure risk in other at-risk populations.

Efficient Decoding of Multi-Dimensional Signals From Population Spiking Activity Using a Gaussian Mixture Particle Filter

New recording technologies and the potential for closed-loop experiments have led to an increasing demand for computationally efficient and accurate algorithms to decode population spiking activity in multi-dimensional spaces. Exact point process filters can accurately decode low-dimensional signals, but are computationally intractable for high-dimensional signals. Approximate Gaussian filters are computationally efficient, but are inaccurate when the signals have complex distributions and nonlinear dynamics. Even particle filter methods tend to become inefficient and inaccurate when the filter distribution has multiple peaks. Here, we develop a new point process filter algorithm that combines the computational efficiency of approximate Gaussian methods with a numerical accuracy that exceeds standard particle filters. We use a mixture of Gaussian model for the posterior at each time step, allowing for an analytic solution to the computationally expensive filter integration step. During non-spike intervals, the filter needs only to update the mean, covariance, and mixture weight of each component. At spike times, a sampling procedure is used to update the filtering distribution and find the number of Gaussian mixture components necessary to maintain an accurate approximation. We illustrate the application of this algorithm to the problem of decoding a rat's position and velocity in a maze from hippocampal place cell data using both 2-D and 4-D decoders.

COMPASS: An Open-Source, General-Purpose Software Toolkit for Computational Psychiatry

Mathematical modeling of behavior during a psychophysical task, referred to as "computational psychiatry," could greatly improve our understanding of mental disorders. One barrier to the broader adoption of computational methods, is that they often require advanced statistical modeling and mathematical skills. Biological and behavioral signals often show skewed or non-Gaussian distributions, and very few toolboxes and analytical platforms are capable of processing such signal categories. We developed the Computational Psychiatry Adaptive State-Space (COMPASS) toolbox, an open-source MATLAB-based software package. This toolbox is easy to use and capable of integrating signals with a variety of distributions. COMPASS has the tools to process signals with continuous-valued and binary measurements, or signals with incomplete-missing or censored-measurements, which makes it well-suited for processing those signals captured during a psychophysical task. After specifying a few parameters in a small set of user-friendly functions, COMPASS allows users to efficiently apply a wide range of computational behavioral models. The model output can be analyzed as an experimental outcome or used as a regressor for neural data and can also be tested using the goodness-of-fit measurement. Here, we demonstrate that COMPASS can replicate two computational behavioral analyses from different groups. COMPASS replicates and can slightly improve on the original modeling results. We also demonstrate the use of COMPASS application in a censored-data problem and compare its performance result with naïve estimation methods. This flexible, general-purpose toolkit should accelerate the use of computational modeling in psychiatric neuroscience.

A common goodness-of-fit framework for neural population models using marked point process time-rescaling

A critical component of any statistical modeling procedure is the ability to assess the goodness-of-fit between a model and observed data. For spike train models of individual neurons, many goodness-of-fit measures rely on the time-rescaling theorem and assess model quality using rescaled spike times. Recently, there has been increasing interest in statistical models that describe the simultaneous spiking activity of neuron populations, either in a single brain region or across brain regions. Classically, such models have used spike sorted data to describe relationships between the identified neurons, but more recently clusterless modeling methods have been used to describe population activity using a single model. Here we develop a generalization of the time-rescaling theorem that enables comprehensive goodness-of-fit analysis for either of these classes of population models. We use the theory of marked point processes to model population spiking activity, and show that under the correct model, each spike can be rescaled individually to generate a uniformly distributed set of events in time and the space of spike marks. After rescaling, multiple well-established goodness-of-fit procedures and statistical tests are available. We demonstrate the application of these methods both to simulated data and real population spiking in rat hippocampus. We have made the MATLAB and Python code used for the analyses in this paper publicly available through our Github repository at https://github.com/Eden-Kramer-Lab/popTRT .

A procedure to increase the power of Granger-causal analysis through temporal smoothing

BACKGROUND

How the human brain coordinates network activity to support cognition and behavior remains poorly understood. New high-resolution recording modalities facilitate a more detailed understanding of the human brain network. Several approaches have been proposed to infer functional networks, indicating the transient coordination of activity between brain regions, from neural time series. One category of approach is based on statistical modeling of time series recorded from multiple sensors (e.g., multivariate Granger causality). However, fitting such models remains computationally challenging as the history structure may be long in neural activity, requiring many model parameters to fully capture the dynamics.

NEW METHOD

We develop a method based on Granger causality that makes the assumption that the history dependence varies smoothly. We fit multivariate autoregressive models such that the coefficients of the lagged history terms are smooth functions. We do so by modelling the history terms with a lower dimensional spline basis, which requires many fewer parameters than the standard approach and increases the statistical power of the model.

RESULTS

We show that this procedure allows accurate estimation of brain dynamics and functional networks in simulations and examples of brain voltage activity recorded from a patient with pharmacoresistant epilepsy.

COMPARISON WITH EXISTING METHOD

The proposed method has more statistical power than the Granger method for networks of signals that exhibit extended and smooth history dependencies.

CONCLUSIONS

The proposed tool permits conditional inference of functional networks from many brain regions with extended history dependence, furthering the applicability of Granger causality to brain network science.

Timing matters: Impact of anticonvulsant drug treatment and spikes on seizure risk in benign epilepsy with centrotemporal spikes

Objective

Benign epilepsy with centrotemporal spikes (BECTS) is a common, self‐limited epilepsy syndrome affecting school‐age children. Classic interictal epileptiform discharges (IEDs) confirm diagnosis, and BECTS is presumed to be pharmacoresponsive. As seizure risk decreases in time with this disease, we hypothesize that the impact of IEDs and anticonvulsive drug (ACD) treatment on the risk of subsequent seizure will differ based on disease duration.

Methods

We calculate subsequent seizure risk following diagnosis in a large retrospective cohort of children with BECTS (n = 130), evaluating the impact of IEDs and ACD treatment in the first, second, third, and fourth years of disease. We use a Kaplan–Meier survival analysis and logistic regression models. Patients were censored if they were lost to follow‐up or if they changed group status.

Results

Two‐thirds of children had a subsequent seizure within 2 years of diagnosis. The majority of children had a subsequent seizure within 3 years despite treatment. The presence of IEDs on electroencephalography (EEG) did not impact subsequent seizure risk early in the disease. By the fourth year of disease, all children without IEDs remained seizure free, whereas one‐third of children with IEDs at this stage had a subsequent seizure. Conversely, ACD treatment corresponded with lower risk of seizure early in the disease but did not impact seizure risk in later years.

Significance

In this cohort, the majority of children with BECTS had a subsequent seizure despite treatment. In addition, ACD treatment and IEDs predicted seizure risk at specific points of disease duration. Future prospective studies are needed to validate these exploratory findings.

A Comparison Study of Point-Process Filter and Deep Learning Performance in Estimating Rat Position Using an Ensemble of Place Cells

The emergence of deep learning techniques has provided new tools for the analysis of complex data in the field of neuroscience. In parallel, advanced statistical approaches like point-process modeling provide powerful tools for analyzing the spiking activity of neural populations. How statistical and machine learning techniques compare when applied to neural data remains largely unclear. In this research, we compare the performance of a point-process filter and a long short-term memory (LSTM) network in decoding the 2D movement trajectory of a rat using the neural activity recorded from an ensemble of hippocampal place cells. We compute the least absolute error (LAE), a measure of accuracy of prediction, and the coefficient of determination (R2), a measure of prediction consistency, to compare the performance of these two methods. We show that the LSTM and point-process filter provide comparable accuracy in predicting the position; however, the point-process provides further information about the prediction which is unavailable for LSTM. Though previous results report better performance using deep learning techniques, our results indicate that this is not universally the case. We also investigate how these techniques encode information carried by place cell activity and compare the computational efficiency of the two methods. While the point-process model is built using the receptive field for each place cell, we show that LSTM does not necessarily encode receptive fields, but instead decodes the movement trajectory using other features of neural activity. Although it is less robust, LSTM runs more than 7 times faster than the fastest point-process filter in this research, providing a strong advantage in computational efficiency. Together, these results suggest that the point-process filters and LSTM approaches each provide distinct advantages; the choice of model should be informed by the specific scientific question of interest.

Computational Neuroscience: Mathematical and Statistical Perspectives

Mathematical and statistical models have played important roles in neuroscience, especially by describing the electrical activity of neurons recorded individually, or collectively across large networks. As the field moves forward rapidly, new challenges are emerging. For maximal effectiveness, those working to advance computational neuroscience will need to appreciate and exploit the complementary strengths of mechanistic theory and the statistical paradigm.

Predicting learning dynamics in Multiple-Choice Decision-Making Tasks using a variational Bayes technique

Multiple-Choice Decision-Making Tasks are widely used to analyze behavior and infer underlying cognitive states that shape the decision and learning processes. The behavioral signals recorded in these tasks are dynamic and often non-Gaussian - for instance, when learning a multiple choice association task. Previously developed estimation algorithms for latent behavioral variables do not address multiple-choice responses. In this research, we use a state-space modeling framework to predict a cognitive learning state related to multiple choice decisions, which are best described by a multinomial distribution. The proposed algorithm combines a multinomial filter/smoother and a variational Bayes technique to estimate the dynamics of a learning state vector. The algorithm is applied to decision response data recorded from non-human primates (NHPs) performing a Multiple-Choice Decision Task.

Capturing Spike Variability in Noisy Izhikevich Neurons Using Point Process Generalized Linear Models

To understand neural activity, two broad categories of models exist: statistical and dynamical. While statistical models possess rigorous methods for parameter estimation and goodness-of-fit assessment, dynamical models provide mechanistic insight. In general, these two categories of models are separately applied; understanding the relationships between these modeling approaches remains an area of active research. In this letter, we examine this relationship using simulation. To do so, we first generate spike train data from a well-known dynamical model, the Izhikevich neuron, with a noisy input current. We then fit these spike train data with a statistical model (a generalized linear model, GLM, with multiplicative influences of past spiking). For different levels of noise, we show how the GLM captures both the deterministic features of the Izhikevich neuron and the variability driven by the noise. We conclude that the GLM captures essential features of the simulated spike trains, but for near-deterministic spike trains, goodness-of-fit analyses reveal that the model does not fit very well in a statistical sense; the essential random part of the GLM is not captured.

Estimating Dynamic Signals From Trial Data With Censored Values

Censored data occur commonly in trial-structured behavioral experiments and many other forms of longitudinal data. They can lead to severe bias and reduction of statistical power in subsequent analyses. Principled approaches for dealing with censored data, such as data imputation and methods based on the complete data's likelihood, work well for estimating fixed features of statistical models but have not been extended to dynamic measures, such as serial estimates of an underlying latent variable over time. Here we propose an approach to the censored-data problem for dynamic behavioral signals. We developed a state-space modeling framework with a censored observation process at the trial timescale. We then developed a filter algorithm to compute the posterior distribution of the state process using the available data. We showed that special cases of this framework can incorporate the three most common approaches to censored observations: ignoring trials with censored data, imputing the censored data values, or using the full information available in the data likelihood. Finally, we derived a computationally efficient approximate Gaussian filter that is similar in structure to a Kalman filter, but that efficiently accounts for censored data. We compared the performances of these methods in a simulation study and provide recommendations of approaches to use, based on the expected amount of censored data in an experiment. These new techniques can broadly be applied in many research domains in which censored data interfere with estimation, including survival analysis and other clinical trial applications.

Rapid classification of hippocampal replay content for real-time applications

Sharp-wave ripple (SWR) events in the hippocampus replay millisecond-timescale patterns of place cell activity related to the past experience of an animal. Interrupting SWR events leads to learning and memory impairments, but how the specific patterns of place cell spiking seen during SWRs contribute to learning and memory remains unclear. A deeper understanding of this issue will require the ability to manipulate SWR events based on their content. Accurate real-time decoding of SWR replay events requires new algorithms that are able to estimate replay content and the associated uncertainty, along with software and hardware that can execute these algorithms for biological interventions on a millisecond timescale. Here we develop an efficient estimation algorithm to categorize the content of replay from multiunit spiking activity. Specifically, we apply real-time decoding methods to each SWR event and then compute the posterior probability of the replay feature. We illustrate this approach by classifying SWR events from data recorded in the hippocampus of a rat performing a spatial memory task into four categories: whether they represent outbound or inbound trajectories and whether the activity is replayed forward or backward in time. We show that our algorithm can classify the majority of SWR events in a recording epoch within 20 ms of the replay onset with high certainty, which makes the algorithm suitable for a real-time implementation with short latencies to incorporate into content-based feedback experiments.

Cognitive state prediction using an EM algorithm applied to Gamma distributed data

Behavioral tests are widely used to quantify features of cognitive processing. For a large class of behavioral signals, the observed variables are non-Gaussian and dynamic; classical estimation algorithms are ill-suited to modeling such data. In this research, we propose a mathematical framework to predict a cognitive state variable related to behavioral signals, which are best modeled using a Gamma distribution. The proposed algorithm combines a Gamma Smoother and EM algorithm in the prediction process. The algorithm is applied to reaction time recorded from subjects performing a Multi-Source Interference Task (MSIT) to dynamically quantify their cognitive flexibility through the course of the experiment.

Tracking non-stationary spectral peak structure in EEG data

We develop a particle filter algorithm to simultaneously estimate and track the instantaneous peak frequency, amplitude, and bandwidth of multiple concurrent non-stationary components of an EEG signal in the time-frequency domain. We use this method to characterize human EEG activity during anesthesia-induced unconsciousness.

Clusterless Decoding of Position from Multiunit Activity Using a Marked Point Process Filter

Point process filters have been applied successfully to decode neural signals and track neural dynamics. Traditionally these methods assume that multiunit spiking activity has already been correctly spike-sorted. As a result, these methods are not appropriate for situations where sorting cannot be performed with high precision, such as real-time decoding for brain-computer interfaces. Because the unsupervised spike-sorting problem remains unsolved, we took an alternative approach that takes advantage of recent insights into clusterless decoding. Here we present a new point process decoding algorithm that does not require multiunit signals to be sorted into individual units. We use the theory of marked point processes to construct a function that characterizes the relationship between a covariate of interest (in this case, the location of a rat on a track) and features of the spike waveforms. In our example, we use tetrode recordings, and the marks represent a four-dimensional vector of the maximum amplitudes of the spike waveform on each of the four electrodes. In general, the marks may represent any features of the spike waveform. We then use Bayes's rule to estimate spatial location from hippocampal neural activity. We validate our approach with a simulation study and experimental data recorded in the hippocampus of a rat moving through a linear environment. Our decoding algorithm accurately reconstructs the rat's position from unsorted multiunit spiking activity. We then compare the quality of our decoding algorithm to that of a traditional spike-sorting and decoding algorithm. Our analyses show that the proposed decoding algorithm performs equivalent to or better than algorithms based on sorted single-unit activity. These results provide a path toward accurate real-time decoding of spiking patterns that could be used to carry out content-specific manipulations of population activity in hippocampus or elsewhere in the brain.

Estimating a dynamic state to relate neural spiking activity to behavioral signals during cognitive tasks

An important question in neuroscience is understanding the relationship between high-dimensional electrophysiological data and complex, dynamic behavioral data. One general strategy to address this problem is to define a low-dimensional representation of essential cognitive features describing this relationship. Here we describe a general state-space method to model and fit a low-dimensional cognitive state process that allows us to relate behavioral outcomes of various tasks to simultaneously recorded neural activity across multiple brain areas. In particular, we apply this model to data recorded in the lateral prefrontal cortex (PFC) and caudate nucleus of non-human primates as they perform learning and adaptation in a rule-switching task. First, we define a model for a cognitive state process related to learning, and estimate the progression of this learning state through the experiments. Next, we formulate a point process generalized linear model to relate the spiking activity of each PFC and caudate neuron to the stimated learning state. Then, we compute the posterior densities of the cognitive state using a recursive Bayesian decoding algorithm. We demonstrate that accurate decoding of a learning state is possible with a simple point process model of population spiking. Our analyses also allow us to compare decoding accuracy across neural populations in the PFC and caudate nucleus.

Examination of rhythmicity of extracellularly recorded neurons in the entorhinal cortex

A number of studies have examined the theta-rhythmic modulation of neuronal firing in the hippocampal circuit. For extracellular recordings, this is often done by examining spectral properties of the spike-time autocorrelogram, most significantly, for validating the presence or absence of theta modulation across species. These techniques can show significant rhythmicity for high firing rate, highly rhythmic neurons; however, they are substantially biased by several factors including the peak firing rate of the neuron, the amount of time spent in the neuron's receptive field, and other temporal properties of the rhythmicity such as cycle-skipping. These limitations make it difficult to examine rhythmic modulation in neurons with low firing rates or when an animal has short dwell times within the firing field and difficult to compare rhythmicity under disparate experimental conditions when these factors frequently differ. Here, we describe in detail the challenges that researchers face when using these techniques and apply our findings to recent recordings from bat entorhinal grid cells, suggesting that they may have lacked enough data to examine theta rhythmicity robustly. We describe a more sensitive and statistically rigorous method using maximum likelihood estimation (MLE) of a parametric model of the lags within the autocorrelation window, which helps to alleviate some of the problems of traditional methods and was also unable to detect rhythmicity in bat grid cells. Using large batteries of simulated data, we explored the boundaries for which the MLE technique and the theta index can detect rhythmicity. The MLE technique is less sensitive to many features of the autocorrelogram and provides a framework for statistical testing to detect rhythmicity as well as changes in rhythmicity in individual sessions providing a substantial improvement over previous methods.

Rate-adjusted spike-LFP coherence comparisons from spike-train statistics

Coherence is a fundamental tool in the analysis of neuronal data and for studying multiscale interactions of single and multiunit spikes with local field potentials. However, when the coherence is used to estimate rhythmic synchrony between spiking and any other time series, the magnitude of the coherence is dependent upon the spike rate. This property is not a statistical bias, but a feature of the coherence function. This dependence confounds cross-condition comparisons of spike-field and spike-spike coherence in electrophysiological experiments. Taking inspiration from correction methods that adjust the spike rate of a recording with bootstrapping ('thinning'), we propose a method of estimating a correction factor for the spike-field and spike-spike coherence that adjusts the coherence to account for this rate dependence. We demonstrate that the proposed rate adjustment is accurate under standard assumptions and derive distributional properties of the estimator. The reduced estimation variance serves to provide a more powerful test of cross-condition differences in spike-LFP coherence than the thinning method and does not require repeated Monte Carlo trials. We also demonstrate some of the negative consequences of failing to account for rate dependence. The proposed spike-field coherence estimator accurately adjusts the spike-field coherence with respect to rate and has well-defined distributional properties that endow the estimator with lower estimation variance than the existing adjustment method.

Assessing dynamics, spatial scale, and uncertainty in task-related brain network analyses

The brain is a complex network of interconnected elements, whose interactions evolve dynamically in time to cooperatively perform specific functions. A common technique to probe these interactions involves multi-sensor recordings of brain activity during a repeated task. Many techniques exist to characterize the resulting task-related activity, including establishing functional networks, which represent the statistical associations between brain areas. Although functional network inference is commonly employed to analyze neural time series data, techniques to assess the uncertainty-both in the functional network edges and the corresponding aggregate measures of network topology-are lacking. To address this, we describe a statistically principled approach for computing uncertainty in functional networks and aggregate network measures in task-related data. The approach is based on a resampling procedure that utilizes the trial structure common in experimental recordings. We show in simulations that this approach successfully identifies functional networks and associated measures of confidence emergent during a task in a variety of scenarios, including dynamically evolving networks. In addition, we describe a principled technique for establishing functional networks based on predetermined regions of interest using canonical correlation. Doing so provides additional robustness to the functional network inference. Finally, we illustrate the use of these methods on example invasive brain voltage recordings collected during an overt speech task. The general strategy described here-appropriate for static and dynamic network inference and different statistical measures of coupling-permits the evaluation of confidence in network measures in a variety of settings common to neuroscience.

Characterizing context-dependent differential firing activity in the hippocampus and entorhinal cortex

The rat hippocampus and entorhinal cortex have been shown to possess neurons with place fields that modulate their firing properties under different behavioral contexts. Such context-dependent changes in neural activity are commonly studied through electrophysiological experiments in which a rat performs a continuous spatial alternation task on a T-maze. Previous research has analyzed context-based differential firing during this task by describing differences in the mean firing activity between left-turn and right-turn experimental trials. In this article, we develop qualitative and quantitative methods to characterize and compare changes in trial-to-trial firing rate variability for sets of experimental contexts. We apply these methods to cells in the CA1 region of hippocampus and in the dorsocaudal medial entorhinal cortex (dcMEC), characterizing the context-dependent differences in spiking activity during spatial alternation. We identify a subset of cells with context-dependent changes in firing rate variability. Additionally, we show that dcMEC populations encode turn direction uniformly throughout the T-maze stem, whereas CA1 populations encode context at major waypoints in the spatial trajectory. Our results suggest scenarios in which individual cells that sparsely provide information on turn direction might combine in the aggregate to produce a robust population encoding.