Biology subjects

Purdon, P. L.

Publications and source records attributed to Purdon, P. L..

12 recordsLinked to original sources

bioRxiv2023

A dynamic generative model can extract interpretable oscillatory components from multichannel neurophysiological recordings

Modern neurophysiological recordings are performed using multichannel sensor arrays that are able to record activity in an increasingly high number of channels numbering in the 100s to 1000s. Often, underlying lower-dimensional patterns of activity are responsible for the observed dynamics, but these representations are difficult to reliably identify using existing methods that attempt to summarize multivariate relationships in a post-hoc manner from univariate analyses, or using current blind source separation methods. While such methods can reveal appealing patterns of activity, determining the number of components to include, assessing their statistical significance, and interpreting them requires extensive manual intervention and subjective judgement in practice. These difficulties with component selection and interpretation occur in large part because these methods lack a generative model for the underlying spatio-temporal dynamics. Here we describe a novel component analysis method anchored by a generative model where each source is described by a bio-physically inspired state space representation. The parameters governing this representation readily capture the oscillatory temporal dynamics of the components, so we refer to it as Oscillation Component Analysis (OCA). These parameters - the oscillatory properties, the component mixing weights at the sensors, and the number of oscillations - all are inferred in a data-driven fashion within a Bayesian framework employing an instance of the expectation maximization algorithm. We analyze high-dimensional electroencephalography and magnetoencephalography recordings from human studies to illustrate the potential utility of this method for neuroscience data. Significance StatementNeuroscience studies often involve simultaneous recordings in a large number of sensors in which a smaller number of dynamic components generate the complex spatio-temporal patterns observed in the data. Current blind source separation techniques produce sub-optimal results and are difficult to interpret because these methods lack an appropriate generative model that can guide both statistical inference and interpretation. Here we describe a novel component analysis method employing a dynamic generative model that can decompose high-dimensional multivariate data into a smaller set of oscillatory components are learned in a data-driven way, with parameters that are immediately interpretable. We show how this method can be applied to neurophysiological recordings with millisecond precision that exhibit oscillatory activity such as electroencephalography and magnetoencephalography.

neuroscience↗

bioRxiv2022

Switching state-space modeling of neural signal dynamics

Linear parametric state-space models are a ubiquitous tool for analyzing neural time series data, providing a way to characterize the underlying brain dynamics with much greater statistical efficiency than non-parametric data analysis approaches. However, neural time series data are frequently time-varying, exhibiting rapid changes in dynamics, with transient activity that is often the key feature of interest in the data. Stationary methods can be adapted to time-varying scenarios by employing fixed-duration windows under an assumption of quasi-stationarity. But time-varying dynamics can be explicitly modeled by switching state-space models, i.e., by using a pool of state-space models with different dynamics selected by a probabilistic switching process. Unfortunately, exact solutions for state inference and parameter learning with switching state-space models are intractable. Here we revisit a switching state-space model inference approach first proposed by Ghahramani and Hinton. We provide explicit derivations for solving the inference problem iteratively after applying variational approximation on the joint posterior of the hidden states and the switching process. We introduce a novel initialization procedure using an efficient leave-one-out strategy to compare among candidate models, which significantly improves performance compared to the existing method that relies on deterministic annealing. We then utilize this state-inference solution within a generalized expectation-maximization algorithm to estimate model parameters of the switching process and the linear state-space models with dynamics potentially shared among candidate models. We perform extensive simulations under different settings to benchmark performance against existing switching inference methods and further validate the robustness of our switching inference solution outside the generative switching model class. Finally, we demonstrate the utility of our method for sleep spindle detection in real recordings, showing how switching state-space models can be used to detect and extract transient spindles from human sleep electroencephalograms in an unsupervised manner. Author summaryAn inherent aspect of brain activity is that it changes over time, but existing methods for analyzing neuroscience data typically assume that the underlying activity is strictly stationary, i.e., the properties of that activity do not change over time. One way of handling time-varying data is to break the data into smaller segments that one assumes to be quasi-stationary, but this approach only works if signals vary gradually, and tends to perform poorly when changes are rapid or the target activity is transient in nature. A class of models called linear switching state-space models can explicitly represent time-varying activity, but they pose another set of challenges: exact solutions for such models are intractable, and existing approximate solutions can be highly inaccurate. In this work we present a solution for linear switching state-space models that is able to recover the underlying hidden states and model parameters for time-varying dynamics in a way that is robust to model mis-specification and that outperforms previously proposed methods. We demonstrate the utility of our method by applying it to the problem of sleep spindle detection and show that switching state-space models can automatically detect transient spindle activity from human sleep electroencephalograms.

neuroscience↗

bioRxiv2022

An iterative search algorithm to identify oscillatory dynamics in neurophysiological time series

Neural oscillations have long been recognized for their mechanistic importance in coordinating activity within and between brain circuits. Co-occurring broad-band, non-periodic signals are also ubiquitous in neural data and are thought to reflect the characteristics of populationlevel neuronal spiking activity. Identifying oscillatory activity distinct from broadband signals is therefore an important, yet surprisingly difficult, problem in neuroscience. Commonly-used bandpass filters produce spurious oscillations when applied to broad-band noise and may be illinformed by canonical frequency bands. Curve-fitting procedures have been developed to identify peaks in the power spectrum distinct from broadband noise. Unfortunately, these ad hoc methods are prone to overfitting and are difficult to interpret in the absence of generative models to formally represent oscillatory behavior. Here we present a novel method to identify and characterize neural oscillations distinct from broad-band noise. First, we propose a new conceptual construct that makes clear, from a dynamical systems perspective, when oscillations are present or not. We then use this construct to develop generative models for neural oscillations. We show through extensive analyses of simulated and human EEG data that our approach identifies oscillations and their characteristics far more accurately than widely used methods, achieving near-perfect recovery of the number of oscillations when the SNR exceeds a very modest threshold. We also show that our method can automatically identify subject-level variations in frequency to overcome the limitations of fixed canonical frequency bands. Finally we demonstrate how our method can extract clinically-relevant neurophysiological features with greater statistical efficiency other established methods.

neuroscience↗

bioRxiv2022

Characterizing ketamine-induced dissociation using human intracranial neurophysiology: brain dynamics, network activity, and interactions with propofol

Subanesthetic doses of ketamine produce rapid and sustained anti-depressant effects in patients with treatment-resistant depression. Unfortunately, the usefulness of ketamine as a treatment is limited by its potential for abuse because of psychotropic side effects such as dissociation. Understanding the brain dynamics and the neural circuits involved in ketamines effects could lend insight into improved therapies for depression with fewer adverse effects. It is believed that ketamine acts via NMDA receptor and hyperpolarization-activated cyclic nucleotide-gated 1 (HCN1) channels to produce changes in oscillatory brain dynamics. Here we show, in humans, a detailed description of the principal oscillatory changes in cortical and subcortical structures by administration of a subanesthetic dose of ketamine. Using recordings from intracranial electrodes, we found that ketamine increased gamma oscillations within prefrontal cortical areas and the hippocampus--structures previously implicated in ketamines antidepressant effects. Furthermore, our studies provide direct evidence of a ketamine-induced 3 Hz oscillation in posteromedial cortex that has been proposed as a mechanism for its dissociative effects. By analyzing changes in neural oscillations after the addition of propofol, whose GABAergic activity antagonizes ketamines NMDA-mediated disinhibition alongside a shared HCN1 inhibitory effect, we identified brain dynamics that could be attributed to NMDA-mediated disinhibition versus HCN1 inhibition. Overall, our results imply that ketamine engages different neural circuits in distinct frequency-dependent patterns of activity to produce its antidepressant and dissociative sensory effects. These insights may help guide the development of novel brain dynamic biomarkers and therapeutics for depression.

neuroscience↗

bioRxiv2022

Propofol disrupts alpha dynamics in distinct thalamocortical networks underlying sensory and cognitive function during loss of consciousness

During propofol-induced general anesthesia, alpha rhythms undergo a striking shift from posterior to anterior, termed anteriorization. We combined human intracranial recordings with diffusion imaging to show that anteriorization occurs with opposing dynamics in two distinct thalamocortical subnetworks. The cortical and thalamic anatomy involved, as well as their known functional roles, suggest multiple means by which propofol dismantles sensory and cognitive processes to achieve loss of consciousness.

neuroscience↗

bioRxiv2022

Neuromodulation due to propofol affects anesthetic oscillatory coupling

Propofol-mediated unconsciousness elicits strong alpha/low-beta and slow oscillations in the electroencephalogram (EEG) of patients. As anesthetic dose increases, the EEG signal changes in ways that give clues to the level of unconsciousness; the network mechanisms of these changes are only partially understood. Here, we construct a biophysical thalamocortical network involving brainstem influences that reproduces transitions in dynamics seen in the EEG involving the evolution of the power and frequency of alpha/low beta and slow rhythm, as well as their interactions. Our model suggests propofol engages thalamic spindle and cortical sleep mechanisms to elicit persistent alpha/low-beta and slow rhythms, respectively. The thalamocortical network fluctuates between two mutually exclusive states on the timescale of seconds. One state is characterized by continuous alpha/low-beta frequency spiking in thalamus (C-state), while in the other, thalamic alpha spiking is interrupted by periods of co-occurring thalamic and cortical silence (I-state). In the I-state, alpha co-localizes to the peak of the slow; in the C-state, there is a variable relationship between an alpha/beta rhythm and the slow oscillation. The C-state predominates near loss of consciousness; with increasing dose, the proportion of time spent in the I-state increases, recapitulating EEG phenomenology. Cortical synchrony drives the switch to the I-state by changing the nature of the thalamocortical feedback. Brainstem influence on the strength of thalamocortical feedback mediates the amount of cortical synchrony. Our model implicates loss of low-beta, cortical synchrony, and coordinated thalamocortical silent periods as contributing to the unconscious state. New & NoteworthyGABAergic anesthetics induce alpha/low-beta and slow oscillations in the EEG, which interact in dose-dependent ways. We construct a thalamocortical model to investigate how these interdependent oscillations change with propofol dose. We find two dynamic states of thalamocortical coordination, which change on the timescale of seconds and dose-dependently mirror known changes in EEG. Thalamocortical feedback determines the oscillatory coupling and power seen in each state, and this is primarily driven by cortical synchrony and brainstem neuromodulation.

neuroscience↗

bioRxiv2021

Elementary Integrate-and-Fire Process Underlies Pulse Amplitudes in Electrodermal Activity

Electrodermal activity (EDA) is a direct read-out of sweat-induced changes in the skins electrical conductance. Sympathetically-mediated pulsatile changes in skin sweat measured as EDA resemble an integrate-and-fire process, which yields an inverse Gaussian model as the inter-pulse interval distribution. We have previously showed that the inter-pulse intervals in EDA follow an inverse Gaussian distribution. However, the statistical structure of EDA pulse amplitudes has not yet been characterized based on the physiology. Expanding upon the integrate-and-fire nature of sweat glands, we hypothesized that the amplitude of an EDA pulse is proportional to the excess volume of sweat produced compared to what is required to just reach the surface of the skin. We modeled this as the difference of two inverse Gaussian models for each pulse, one which represents the time required to produce just enough sweat to rise to the surface of the skin and one which represents the time requires to produce the actual volume of sweat. We proposed and tested a series of four simplifications of our hypothesis, ranging from a single difference of inverse Gaussians to a single simple inverse Gaussian. We also tested four additional models for comparison, including the lognormal and gamma distributions. All models were tested on EDA data from two subject cohorts, 11 healthy volunteers during 1 hour of quiet wakefulness and a different set of 11 healthy volunteers during approximately 3 hours of controlled propofol sedation. All four models represent simplifications of our hypothesis outperformed other models across all 22 subjects, as measured by Akaikes Information Criterion (AIC), as well as mean and maximum distance from the diagonal on a quantile-quantile plot. Our broader model set of four simplifications offered a useful framework to enhance further statistical descriptions of EDA pulse amplitudes. Some of the simplifications prioritize fit near the mode of the distribution, while others prioritize fit near the tail. With this new insight, we can summarize the physiologically-relevant amplitude information in EDA with at most four parameters. Our findings establish that physiologically based probability models provide parsimonious and accurate description of temporal and amplitude characteristics in EDA. AUTHOR SUMMARYElectrodermal activity (EDA) is an indirect read-out of the bodys sympathetic nervous system, or fight-or-flight response, measured as sweat-induced changes in the electrical conductance properties of the skin. Interest is growing in using EDA to track physiological conditions such as stress levels, sleep quality, and emotional states. Our previous worked showed that the times in between EDA pulses obeyed a specific statistical distribution, the inverse Gaussian, that arises from the physiology of EDA production. In this work, we build on that insight to analyze the amplitudes of EDA pulses. In an analysis of EDA data recorded in 11 healthy volunteers during quiet wakefulness and 11 different healthy volunteers during controlled propofol sedation, we establish that the amplitudes of EDA pulses also have specific statistical structure, as the differences of inverse Gaussians, that arises from the physiology of sweat production. We capture that structure using a series of progressively simpler models that each prioritize different parts of the pulse amplitude distribution. Our findings show a physiologically-based statistical model provides a parsimonious and accurate description EDA. This enables increased reliability and robustness in analyzing EDA data collected under any circumstance.

neuroscience↗

bioRxiv2020

Quantitative assessment of the relationship between behavioral and autonomic dynamics during propofol-induced unconsciousness

During general anesthesia, both behavioral and autonomic changes are caused by the administration of anesthetics such as propofol. Propofol produces unconsciousness by creating highly structured oscillations in brain circuits. The anesthetic also has autonomic effects due to its actions as a vasodilator and myocardial depressant. Understanding how autonomic dynamics change in relation to propofol-induced unconsciousness is an important scientific and clinical question since anesthesiologists often infer changes in level of unconsciousness from changes in autonomic dynamics. Therefore, we present a framework combining physiology-based statistical models that have been developed specifically for heart rate variability and electrodermal activity with a robust statistical tool to compare behavioral and multimodal autonomic changes before, during, and after propofol-induced unconsciousness. We tested this framework on physiological data recorded from nine healthy volunteers during computer-controlled administration of propofol. We studied how autonomic dynamics related to behavioral markers of unconsciousness: 1) overall, 2) during the transitions of loss and recovery of consciousness, and 3) before and after anesthesia as a whole. Our results show a strong relationship between behavioral state of consciousness and autonomic dynamics. All of our prediction models showed areas under the curve greater than 0.75 despite the presence of non-monotonic relationships among the variables during the transition periods. Our analysis highlighted the specific roles played by fast versus slow changes, parasympathetic vs sympathetic activity, heart rate variability vs electrodermal activity, and even pulse rate vs pulse amplitude information within electrodermal activity. Further advancement upon this work can quantify the complex and subject-specific relationship between behavioral changes and autonomic dynamics before, during, and after anesthesia. However, this work demonstrates the potential of a multimodal, physiologically-informed, statistical approach to characterize autonomic dynamics.

neuroscience↗

bioRxiv2020

Electroencephalogram Monitoring of Depth of Anesthesia during Office-Based Anesthesia

ObjectiveElectroencephalogram (EEG) monitors are often used to monitor depth of general anesthesia. EEG monitoring is less well developed for lighter levels of anesthesia. Here we present an automated method to monitor the depth of anesthesia for office based procedures using EEG spectral features. MethodsWe analyze EEG recordings from 30 patients undergoing sedation using a multimodal anesthesia strategy. Level of sedation during the procedure is coded using the Richmond Agitation and Sedation Scale (RASS). The power spectrum from the frontal EEG is used to infer the level of sedation, by training a logistic regression model with elastic net regularization. Area under the receiver operator characteristic curve (AUC) is used to evaluate how well the automated system distinguishes awake from sedated EEG epochs. ResultsEEG power spectral characteristics vary systematically and consistently across patients with the levels of light anesthesia and relatively healthy patients encountered during office-based anesthesia procedures. The logistic regression model using spectral EEG features distinguishes awake and sedated states with an AUC of 0.85 ({+/-} 0.14). ConclusionsOur results demonstrate that frontal EEG spectral features can reliably monitor sedation levels during office based anesthesia.

neuroscience↗

bioRxiv2020

A Model-Based Approach for Pulse Selection from Electrodermal Activity

ObjectiveThe goal of this work was to develop a physiology-based paradigm for pulse selection from electrodermal activity (EDA) data. MethodsWe aimed to use insight about the integrate-and-fire physiology of sweat gland bursts, which predicts inverse Gaussian inter-pulse interval structure. At the core of our paradigm is a subject-specific amplitude threshold selection process for pulses based on the statistical properties of four right-skewed models including the inverse Gaussian. These four models differ in their tail behavior, which reflects sweat gland physiology to varying degrees. By screening across thresholds and fitting all four models, we selected for heavier tails that reflect inverse Gaussian-like structure and verified the pulse selection with a goodness-of-fit analysis. ResultsWe tested our paradigm on two different subject cohorts recorded during different experimental conditions and using different equipment. In both cohorts, our method robustly and consistently recovered pulses that captured the inverse Gaussian-like structure predicted by physiology, despite large differences in noise level of the data. In contrast, an established EDA analysis paradigm, which assumes a constant amplitude threshold across all data, was unable to separate pulses from noise. ConclusionWe present a computationally efficient, statistically rigorous, and physiology-informed paradigm for pulse selection from EDA data that is robust across individuals and experimental conditions yet adaptable to changes in noise level. SignificanceThe robustness of our paradigm and its basis in physiology move EDA closer to serving as a clinical marker for sympathetic activity in diverse conditions such as pain, anxiety, depression, and sleep.

physiology↗

bioRxiv2019

State Space Methods for Phase Amplitude Coupling Analysis

Phase Amplitude Coupling (PAC) is thought to play a fundamental role in the dynamic coordination of brain circuits and systems. There are however growing concerns that existing methods for PAC analysis are prone to error and misinterpretation. Improper frequency band selection can render true PAC undetectable, while non-linearities or abrupt changes in the signal can produce spurious PAC. Current methods require substantial amounts of data and lack formal statistical inference tools. We describe here a novel approach for PAC analysis that substantially addresses these problems. We use a state space model to estimate the component oscillations, avoiding problems with frequency band selection, nonlinearities, and sharp signal transitions. We represent cross-frequency coupling in parametric and time-varying forms to further improve statistical efficiency and estimate the posterior distribution of the coupling parameters to derive their credible intervals. We demonstrate the method using simulated data, rat LFP data, and human EEG data.

neuroscience↗

bioRxiv2017

Electroencephalographic markers of brain development during sevoflurane anesthesia in children aged 0 to 3 years old

The general anesthetic sevoflurane acts on GABAergic-interneurons to generate stereotyped oscillations that relate fundamentally to neural circuit architecture. Each year, millions of children require general anesthesia, providing an experiment of nature that allows characterization of the developmental trajectory of GABAergic-inhibitory circuits in the human brain. We used multichannel electroencephalograph recordings in 91 children 0-40 months old. We mapped spatial power and coherence across the cortex. During sevoflurane exposure: (1) slow-delta oscillations are present in all ages; (2) theta and alpha oscillations emerge around 4 months; (3) alpha oscillations increase in power from 4 to 10 months; (4) frontal alpha oscillation predominance emerge at ~6 months; (5) frontal slow oscillations are coherent in the first 6 months of age only; and (6) frontal alpha oscillations become coherent around 10 months and persist in older ages. Our results suggest key developmental milestones are visible in the functional activity of sevoflurane-stimulated GABAergic circuits.

neuroscience↗