Justin Brooks

Estimating direction in brain-behavior interactions: Proactive and reactive brain states in driving

Abstract Conventional neuroimaging analyses have ascribed function to particular brain regions, exploiting the power of the subtraction technique in fMRI and event‐related potential analyses in EEG. Moving beyond this convention, many researchers have begun exploring network‐based neurodynamics and coordination between brain regions as a function of behavioral parameters or environmental statistics; however, most approaches average evoked activity across the experimental session to study task‐dependent networks. Here, we examined on‐going oscillatory activity as measured with EEG and use a methodology to estimate directionality in brain‐behavior interactions. After source reconstruction, activity within specific frequency bands (delta: 2–3 Hz; theta: 4–7 Hz; alpha: 8–12 Hz; beta: 13–25 Hz) in a priori regions of interest was linked to continuous behavioral measurements, and we used a predictive filtering scheme to estimate the asymmetry between brain‐to‐behavior and behavior‐to‐brain prediction using a variant of Granger causality. We applied this approach to a simulated driving task and examined directed relationships between brain activity and continuous driving performance (steering behavior or vehicle heading error). Our results indicated that two neuro‐behavioral states may be explored with this methodology: a Proactive brain state that actively plans the response to the sensory information and is characterized by delta‐beta activity, and a Reactive brain state that processes incoming information and reacts to environmental statistics primarily within the alpha band. Graphical abstract Figure. No Caption available. HighlightsTraditional neuroscience studies investigate localized task‐evoked responsesOur approach examines continuous tracking of brain‐behavior interactions in oscillatory activityBrain leads behavior in a Proactive state, while brain follows behavior in a Reactive stateReactive states are largely carried by alpha activity in regions sensitive to environmental statisticsProactive states rely more on a diffuse delta‐beta network, particularly when linked with steering behavior

A low-power multi-physiological monitoring processor for stress detection

Personal monitoring systems can offer effective solutions for human health and performance. These systems require sampling and significant processing on multiple streams of physiological signals. The processing typically consists of feature extraction, data fusion, and classification stages that require a large number of digital signal processing and machine learning kernels. In order to be functional, however, the processing architecture needs to be low-power and have a low-area footprint. In this paper we present such a design for a personalized stress monitoring system with a flexible, multi-modal design. Various physiological and behavioral features were explored to maximize detection accuracy with both SVM and KNN machine learning classifiers. Among 17 different features from 5 sensors, heart rate and accelerometer features were found to have the highest classification accuracy to detect stress in the given dataset. While KNN classifier accuracy outperforms by 2%, it requires significantly larger memory and computation compared to the SVM classifier. Therefore, we chose the SVM classifier for hardware implementation. The post-layout implementation results in 130 nm CMOS technology show that the SVM processor occupies 0.2 mm2 and dissipates 20.2 mW at 125 MHz. The proposed processor takes 800 ns to classify each input and consumes 16.2 nJ. The overall classification accuracy of this system is 96%.

Differential Functionality of Right and Left Parietal Activity in Controlling a Motor Vehicle

Driving a motor vehicle is an inherently complex task that requires robust control to avoid catastrophic accidents. Drivers must maintain their vehicle in the middle of the travel lane to avoid high speed collisions with other traffic. Interestingly, while a vehicle’s lane deviation (LD) is critical, studies have demonstrated that heading error (HE) is one of the primary variables drivers use to determine a steering response, which directly controls the position of the vehicle in the lane. In this study, we examined how the brain represents the dichotomy between control/response parameters (heading, reaction time (RT), and steering wheel corrections) and task-critical parameters (LD). Specifically, we examined electroencephalography (EEG) alpha band power (8–13 Hz) from estimated sources in right and left parietal regions, and related this activity to four metrics of driving performance. Our results demonstrate differential task involvement between the two hemispheres: right parietal activity was most closely related to LD, whereas left parietal activity was most closely related to HE, RT and steering responses. Furthermore, HE, RT and steering wheel corrections increased over the duration of the experiment while LD did not. Collectively, our results suggest that the brain uses differential monitoring and control strategies in the right and left parietal regions to control a motor vehicle. Our results suggest that the regulation of this control changes over time while maintaining critical task performance. These results are interpreted in two complementary theoretical frameworks: the uncontrolled manifold and compensatory control theories. The central tenet of these frameworks permits performance variability in parameters (i.e., HE, RT and steering) so far as it does not interfere with critical task execution (i.e., LD). Our results extend the existing research by demonstrating potential neural substrates for this phenomenon which may serve as potential targets for brain-computer interfaces that predict poor driving performance.

Event-related alpha perturbations related to the scaling of steering wheel corrections

Previously we derived a new measure relating the drivers steering wheel responses to the vehicles heading error velocity. This measure, the relative steering wheel compensation (RSWC), changes at times coincident with an alerting stimulus, possibly representing shifts in control strategy as measured by a change in the gain between visual input and motor output. In the present study, we sought to further validate this novel measure by determining the relationship between the RSWC and electroencephalogram (EEG) activity in brain regions associated with sensorimotor transformation processes. These areas have been shown to exhibit event-related spectral perturbation (ERSP) in the alpha frequency band that occurs with the onset of corrective steering wheel maneuvers in response to vehicle perturbations. We hypothesized that these regions would show differential alpha activity depending on whether the RSWC was high or low, reflecting changes in gain between visual input and motor output. Interestingly, we find that low RSWC is associated with significantly less peak desynchronization than larger RSWC. In addition we demonstrate that these differences are not attributable to the amount the steering wheel is turned nor the heading error velocity independently. Collectively these results suggest that neural activity in these sensorimotor regions scales with alertness and may represent differential utilization of multisensory information to control the steering wheel.

Target Class Induction through Image Feedback Manipulation in Rapid Serial Visual Presentation Experiments

Prior research has shown the utility of labeling images by rapidly displaying them to humans via a Rapid Serial Visual Presentation (RSVP) paradigm, classifying the resulting neural data, and integrating the results with computer vision. However, there is currently very little research on providing feedback to the human interacting with one of these systems. To explore this question, an RSVP task was developed to examine the effectiveness of feedback to induce changes in target category in near-real time. Three different factors involved in image presentation were explored: image presentation duration, target/distract or similarity, and feedback modality. Significant, nonlinear changes in performance were related to these independent variables. These results demonstrate the complexity inherent to human category learning and will guide future use of image presentation parameters to optimize human performance within a human-assisted computing system that is focused on image analysis.

Different profiles of decision making and physiology under varying levels of stress in trained military personnel

Decision making is one of the most vital processes we use every day, ranging from mundane decisions about what to eat to life-threatening choices such as how to avoid a car collision. Thus, the context in which our decisions are made is critical, and our physiology enables adaptive responses that account for how environmental stress influences our performance. The relationship between stress and decision making can additionally be affected by ones expertise in making decisions in high-threat environments, where experts can develop an adaptive response that mitigates the negative impacts of stress. In the present study, 26 male military personnel made friend/foe discriminations in an environment where we manipulated the level of stress. In the high-stress condition, participants received a shock when they incorrectly shot a friend or missed shooting a foe; in the low-stress condition, participants received a vibration for an incorrect decision. We characterized performance using signal detection theory to investigate whether a participant changed their decision criterion to avoid making an error. Results showed that under high-stress, participants made more false alarms, mistaking friends as foes, and this co-occurred with increased high frequency heart rate variability. Finally, we examined the relationship between decision making and physiology, and found that participants exhibited adaptive behavioral and physiological profiles under different stress levels. We interpret this adaptive profile as a marker of an experts ingrained training that does not require top down control, suggesting a way that expert training in high-stress environments helps to buffer negative impacts of stress on performance.

Overlapping brain network and alpha power changes suggest visuospatial attention effects on driving performance.

When humans perform prolonged, continuous tasks, their performance fluctuates. The etiology of these fluctuations is multifactorial, but they are influenced by changes in attention reflected in underlying neural dynamics. Previous work with electroencephalography has suggested that prestimulus alpha power is a neural signature of attention allocation with higher power portending relatively poorer performance. The functional mechanisms subserving these changes in alpha power and behavior are postulated to be the result of networked neural activity that permits flexibility in the allocation of attention. Here, we directly examine the similarity between prestimulus alpha connectivity and power in relation to performance fluctuations in a continuous driving task. Participants were asked to maintain their vehicle in the center of a simulated highway, and we evaluated their performance by randomly perturbing the vehicle and assessing their steering correction. We then used the 3 seconds of neural activity before the unexpected event to derive alpha functional connectivity in the first analysis and alpha power in the second analysis, and we employed linear regression to separately investigate their relationship to 3 metrics of driving performance (lane deviation, reaction time (RT), and heading error). We find that the locations involved in our network analysis also show the strongest modulation of alpha activity. Interestingly, the network pattern suggests a posterior to anterior directionality, consistent with bottom-up theories of attention, and these results may reflect a gain control model of attention in which ongoing attention is modulated through coordinated, network activity.