2019-2020 Doctoral Awardee: Jacob Westerberg

“Bridging the Gap Between Cognitive Signals from Single Neurons to EEG”

Jacob Westerberg, Vanderbilt University, Department of Psychology

Abstract

The brain is a highly dynamic system. Cognitive and perceptual processes that are often studied in human participants manifest the same dynamicity through behavioral observations. Behavior is not the only useful measure employed by psychophysicists. Electrophysiology can also be noninvasively employed in humans to understand the neural correlates of cognitive and perceptual dynamics. My dissertation research has explored the dynamics intracranially while being designed to directly relate back to and explain the behavioral and electrophysiological observations in humans. To do so, we employ simultaneous EEG and high-density neurophysiological recordings in macaque monkeys performing cognitive and perceptual tasks identical to those performed by human participants. In this way, we explore the sources in the brain that give rise to behavioral changes and noninvasive signals measured in humans. Specifically, my dissertation research has provided insights into where noninvasive indicators of visual selective attention manifest in the brain, how selective attention can be biased to improve behavioral performance, how behavioral performance monitoring is unaffected by such attentional biases, and how predictability might be important in perceptual adaptation.

Introduction

Sophisticated brain-machine interface (BMI) is rapidly becoming a reality with modern advances in neural recording technologies. Major companies like Neuralink aim to implant humans with high-density intracranial recording devices within the next few years. However, to intelligently use BMI, it will be important to know where to place them to record and possibly affect areas in the brain that are of interest. An example of this might be a device to readout attentional state which may be useful in cases of cognitive disruption such as schizophrenia or attention-deficit/hyperactivity disorder. For example, the electroencephalogram (EEG) can be used to readout and, coupled with transcranial direct current stimulation, affect performance in cognitive tasks. While we know we can measure and manipulate these cognitive abilities, our understanding is in its infancy. Beginning to understand in more detail where sources of these cognitive signals in the brain are localized will allow us to more effectively use BMI moving forward.

While a great deal of the previous work in the domain of cognitive neurophysiology has identified signals in the brain that relate to cognition, they were limited in scope, primarily due to technological reasons. With the recent developments in high density recordings and precise causal manipulation of neural circuits, it has become possible to functionally trace the origins for these signals more precisely. The goal of my dissertation research is to identify the electrophysiological sources for cognitive and perceptual signals in the brain through tasks that alter said signals in a temporally- and spatially-dependent manner. In this way, changes in the spatiotemporal profile of these signals can used to trace their origins. To measure these spatiotemporal dependencies, we use a technique that is both spatially and temporally precise. High density neural recordings coupled with electroencephalography (EEG) in macaque monkeys allow for the necessary precision. It also allows us to relate our findings back to the human literature as the EEG in monkeys is highly similar to EEG in humans and monkeys can be trained to perform cognitive tasks that are identical to those performed by humans.

Methods

To discover the neural sources for cognitive changes in the brain, we use a translational approach. Macaque monkeys can perform many of the basic cognitive and perceptual tasks that we humans can and are perhaps our best model organism to investigate the neural correlates of these tasks. This is because we can directly measure the neural activity in the brain of monkeys through neurophysiological recordings that are typically off limits in humans. Additionally, monkeys also have electrophysiological homologues of signals that can be measured. That is, electroencephalography (EEG) in humans identifies signals generated in the brain that index cognitive and perceptual processes and EEG in monkeys finds many of the same signals. These noninvasive signals can then be related to the direct neural recordings to determine their origins.

One task, which has been the primary task used in my dissertation project, is the pop-out visual search task. Specifically, a priming variant known as priming of pop-out. In this task, monkeys identify a salient target stimulus in an array of homogenous distractors. Monkeys perform many trials of this task each day. These trials are organized in such a way that the search conditions remain consistent before switching after several trials. These blocks of consistently structured trials induce a change in behavior known as attentional priming. By performing these blocks, response times decrease, and performance improves. Simply put, repetition leads to behavioral benefits. However, switching the search conditions leads to slower, poorer performance of the task. I have also trained monkeys to a simple perceptual adaptation task to observe how the brain changes encoding with repetition. In my dissertation work, I have investigated the neural bases for these cognitive and perceptual improvements and shortcomings.

Priming of pop-out visual search task. A. Cartoon task design. Monkeys view a stimulus array following a fixation period. They saccade to the oddball stimulus to receive a reward. B. Trials are organized into blocks of unchanging search conditions to elicit changes in selective attention with priming. C. Two monkey data showing speeded responses times with priming of pop-out. D. Two monkey data showing improved accuracy with priming of pop-out.

Once monkeys have been trained to perform the task and the expected behavioral changes manifest, we introduce microelectrodes into the brain and place EEG electrodes outside the brain to measure neural dynamics at both levels simultaneously. Recent developments in electrode technology allow for high-density recordings of neural activity to more precisely identify the circuitry underlying these cognitive operations. My dissertation work has spanned generations of recording technology and continues to explore the cutting edge from single electrodes to 24-, 32-, and 64-channel electrodes, and now, 960-channel devices. We use evidence from human EEG studies and monkey neurophysiology studies to determine where to place these electrodes to best localize the potential regions for these cognitive operations. My dissertation work has investigated cognitive and perceptual changes in several brain regions from areas like V1 at the earliest stages of cortical visual processing to area V4, a mid-level visual processing region, to areas SEF and now FEF in the frontal cortex where many cognitive processes are thought to first manifest.

To then interpret those signals and relate them in a quantitatively rigorous manner, I employ Bayesian statistics, signal detection theory, information theory, and modeling. Through these methods, my dissertation work has discovered neural mechanisms for changes in perception and cognition that can be directly related to human behavior and electrophysiology.

Results

We began working towards identifying neural sources of cognitive and perceptual signals by demonstrating the utility of investigating the dissociations between signal inputs and outputs of neural circuits. Through this technique, we identified whether brain areas inherit signals or are the source of signals. To do so we investigated whether adaptation through repetition suppression (Westerberg & Maier in prep) in the brain manifests through a top-down or bottom-up interactions. We recorded neural activity across the layers of V1 in monkeys while they viewed repetitive images. By recording across cortical layers, we identified that repetition-induced adaptation occurred in the top-down, feedback-associated laminar compartment. This suggests this sort of perceptual adaptation is initiated through a top-down process and that we can identify neural sources of signals through these methods (Westerberg et al. 2019 J Neurophysiol; Tovar & Westerberg et al. under review J Neurosci).

In a second study, we sought to use a multilevel approach to investigate the origins of such cognitive processes. Specifically, we wanted to discover the underlying neural mechanism and source for a change in selective attention. Monkeys were trained to perform to perform a variant of pop-out visual search where priming of attentional selection occurs while neural activity in frontal cortical area SEF and concurrent EEG was recorded. We found that while there were changes in EEG associated with priming, the neural activity in SEF (a known contributor to EEG), did not. This suggests that the SEF neuronal microcircuit does not induce the changes associated with priming and that there are multiple contributors to EEG as SEF activity did not reflect the concomitant changes in EEG. This dissociation between a known contributor to EEG and the EEG itself demonstrates the necessity of precisely identifying sources of cognitive signals (Westerberg et al. 2020 J Cogn Neurosci). Priming of pop-out visual search task. A. Cartoon task design. Monkeys view a stimulus array following a fixation period. They saccade to the oddball stimulus to receive a reward. B. Trials are organized into blocks of unchanging search conditions to elicit changes in selective attention with priming. C. Two monkey data showing speeded responses times with priming of pop-out. D. Two monkey data showing improved accuracy with priming of pop-out.

Relating signals in the brain to the EEG. A. (left) MRI of a monkey brain with area V4 highlighted in red. Inset shows a sagittal section of the brain with the labels for sulci in the vicinity of V4. (right) Neun and SMI-32 histological stains showing the layers of V4 with scales bars, layer labels, and recording electrode cartoon for reference. B. Cartoon of EEG electrode placements on the monkey head. C. Depiction of the relationship between the signals in the brain to the EEG. Activity across the layers of V4 relays information regarding the visual scene to which can be measured and related to the EEG. D. Exemplar EEG event-related potential during the visual search task. Inset depicts the N2pc when attended to stimulus is contralaterally- vs. ipsilaterally-presented. E. Synaptic activity (current source density [CSD]) in V4 when the attended to stimulus is in the response field, outside the response field, and the difference of those conditions. CSD is thought to generate the EEG and, in this example, shows differences when the concurrently recorded EEG does (relative to panel D). F. The measured contribution of area V4 to the EEG during visual search. V4 upper layers (S, blue) relay information regarding the visual display to the EEG signal during the time of the N2pc.

To follow up our first study in source localizing the neural mechanism of priming of pop-out, we measured activity in visual cortex. This served two purposes. We could determine whether the hypothesized contributor of the most commonly used EEG component indexing spatial attention (known as the N2pc) was indeed generated by V4 and investigate a potential source for the changes in selective attention due to priming of pop-out. Previous work implicated visual cortex, but there were no direct recordings supporting conclusions. Also, previous EEG work in humans indicated visual cortex is involved in priming as changes were found in an EEG event-related potential indexing selective visual attention. Monkeys were trained to perform the task while activity was measured in area V4. We found that area V4 contributes to the generation of the N2pc (Westerberg et al. in prep) and is involved in the changes associated with priming of pop-out (Westerberg et al. 2020 eNeuro), however the same changes have been observed in previous study in frontal cortex and laminar specificity of the effect indicated either V4 as the source of the changes or frontal feedback initiating the change (Westerberg et al. in prep). My ongoing work is exploring this relationship through simultaneous recordings in V4 and FEF (the implicated frontal cortical area) to determine which is the actual source.

Discussion

My dissertation research has primarily focused on localizing and understanding the neural mechanisms of complex cognitive and perceptual processes. My goal is to find and understand these sources so that we can effectively read out and affect neural computations in humans. By understanding the signal generators in the brain through means we currently have Relating signals in the brain to the EEG. A. (left) MRI of a monkey brain with area V4 highlighted in red. Inset shows a sagittal section of the brain with the labels for sulci in the vicinity of V4. (right) Neun and SMI-32 histological stains showing the layers of V4 with scales bars, layer labels, and recording electrode cartoon for reference. B. Cartoon of EEG electrode placements on the monkey head. C. Depiction of the relationship between the signals in the brain to the EEG. Activity across the layers of V4 relays information regarding the visual scene to which can be measured and related to the EEG. D. Exemplar EEG event-related potential during the visual search task. Inset depicts the N2pc when attended to stimulus is contralaterally- vs. ipsilaterally-presented. E. Synaptic activity (current source density [CSD]) in V4 when the attended to stimulus is in the response field, outside the response field, and the difference of those conditions. CSD is thought to generate the EEG and, in this example, shows differences when the concurrently recorded EEG does (relative to panel D). F. The measured contribution of area V4 to the EEG during visual search. V4 upper layers (S, blue) relay information regarding the visual display to the EEG signal during the time of the N2pc. (e.g. EEG) and through means that are becoming more and more feasible in humans (e.g. high-density neurophysiological recordings) we can begin to translate cognitive neurophysiological findings to actionable targets in cases of cognitive impairment or human enhancement. Understanding attentional biasing is a salient example of this. By localizing the source for attentional priming in the brain and understanding its underlying mechanism, we can begin to pursue avenues for manipulating this change. This may be especially important in disease cases such as attention-deficit/hyperactivity disorder where attentional selection is either too frequent or too infrequent.

My dissertation work also has implications in a retrospective sense. Understanding where signals like the N2pc manifest in the brain provides further insight into previous findings using the N2pc as their index into cognition. The knowledge of the N2pc generator contextualizes findings in a neurophysiological sense where we can begin to more carefully marry the findings between monkey neurophysiology and human electrophysiology.

Impact Statement

We are on the cusp of a neurophysiological recording revolution for brain-machine interfaces in humans. The advent of major companies like Neuralink are indicators of this. Brain-machine interface promises incredible change socially and poses significant ethical considerations that will have an impact on society and policy decisions, such as in cases of human enhancement. My dissertation research is designed to better understand the signals that come about from the sort of high-density recordings that will be at the heart of this revolution. It also explores the limitations of these recordings and what we can truly infer from the signals we are recording.