Mapping cognitive brain functions at scale

Neural activity during inhibitory control predicts suicidal ideation with machine learning

Suicide is a leading cause of death in the US and worldwide. Current strategies for preventing suicide are often focused on the identification and treatment of risk factors, especially suicidal ideation (SI). Hence, developing data-driven biomarkers of SI may be key for suicide prevention and intervention. Prior attempts at biomarker-based prediction models for SI have primarily used expensive neuroimaging technologies, yet clinically scalable and affordable biomarkers remain elusive. Here, we investigated the classification of SI using machine learning (ML) on a dataset of 76 subjects with and without SI(+/-) (n = 38 each), who completed a neuro-cognitive assessment session synchronized with electroencephalography (EEG). SI+/- groups were matched for age, sex, and mental health symptoms of depression and anxiety. EEG was recorded at rest and while subjects engaged in four cognitive tasks of inhibitory control, interference processing, working memory, and emotion bias. We parsed EEG signals in physiologically relevant theta (4-8 Hz), alpha (8-13 Hz), and beta (13-30 Hz) frequencies and performed cortical source imaging on the neural signals. These data served as SI predictors in ML models. The best ML model was obtained for beta band power during the inhibitory control (IC) task, demonstrating high sensitivity (89%), specificity (98%). Shapley explainer plots further showed top neural predictors as feedback-related power in the visual and posterior default mode networks and response-related power in the ventral attention, fronto-parietal, and sensory-motor networks. We further tested the external validity of the model in an independent clinically depressed sample (n = 35, 12 SI+) that engaged in an adaptive test version of the IC task, demonstrating 50% sensitivity and 61% specificity in this sample. Overall, the study suggests a promising, scalable EEG-based biomarker approach to predict SI that may serve as a target for risk identification and intervention.

Aperiodic and periodic components of oscillatory brain activity in relation to cognition and symptoms in pediatric ADHD

Children with attention-deficit/hyperactivity disorder show deficits in processing speed, as well as aberrant neural oscillations, including both periodic (oscillatory) and aperiodic (1/f-like) activity, reflecting the pattern of power across frequencies. Both components were suggested as underlying neural mechanisms of cognitive dysfunctions in attention-deficit/hyperactivity disorder. Here, we examined differences in processing speed and resting-state-Electroencephalogram neural oscillations and their associations between 6- and 12-year-old children with (n = 33) and without (n = 33) attention-deficit/hyperactivity disorder. Spectral analyses of the resting-state EEG signal using fast Fourier transform revealed increased power in fronto-central theta and beta oscillations for the attention-deficit/hyperactivity disorder group, but no differences in the theta/beta ratio. Using the parameterization method, we found a higher aperiodic exponent, which has been suggested to reflect lower neuronal excitation-inhibition, in the attention-deficit/hyperactivity disorder group. While fast Fourier transform-based theta power correlated with clinical symptoms for the attention-deficit/hyperactivity disorder group only, the aperiodic exponent was negatively correlated with processing speed across the entire sample. Finally, the aperiodic exponent was correlated with fast Fourier transform-based beta power. These results highlight the different and complementary contribution of periodic and aperiodic components of the neural spectrum as metrics for evaluation of processing speed in attention-deficit/hyperactivity disorder. Future studies should further clarify the roles of periodic and aperiodic components in additional cognitive functions and in relation to clinical status.

Design and Implementation of a Brief Digital Mindfulness and Compassion Training App for Health Care Professionals: Cluster Randomized Controlled Trial

BACKGROUND

Several studies show that intense work schedules make health care professionals particularly vulnerable to emotional exhaustion and burnout.

OBJECTIVE

In this scenario, promoting self-compassion and mindfulness may be beneficial for well-being. Notably, scalable, digital app-based methods may have the potential to enhance self-compassion and mindfulness in health care professionals.

METHODS

In this study, we designed and implemented a scalable, digital app-based, brief mindfulness and compassion training program called "WellMind" for health care professionals. A total of 22 adult participants completed up to 60 sessions of WellMind training, 5-10 minutes in duration each, over 3 months. Participants completed behavioral assessments measuring self-compassion and mindfulness at baseline (preintervention), 3 months (postintervention), and 6 months (follow-up). In order to control for practice effects on the repeat assessments and calculate effect sizes, we also studied a no-contact control group of 21 health care professionals who only completed the repeated assessments but were not provided any training. Additionally, we evaluated pre- and postintervention neural activity in core brain networks using electroencephalography source imaging as an objective neurophysiological training outcome.

RESULTS

Findings showed a post- versus preintervention increase in self-compassion (Cohen d=0.57; P=.007) and state-mindfulness (d=0.52; P=.02) only in the WellMind training group, with improvements in self-compassion sustained at follow-up (d=0.8; P=.01). Additionally, WellMind training durations correlated with the magnitude of improvement in self-compassion across human participants (ρ=0.52; P=.01). Training-related neurophysiological results revealed plasticity specific to the default mode network (DMN) that is implicated in mind-wandering and rumination, with DMN network suppression selectively observed at the postintervention time point in the WellMind group (d=-0.87; P=.03). We also found that improvement in self-compassion was directly related to the extent of DMN suppression (ρ=-0.368; P=.04).

CONCLUSIONS

Overall, promising behavioral and neurophysiological findings from this first study demonstrate the benefits of brief digital mindfulness and compassion training for health care professionals and compel the scale-up of the digital intervention.

TRIAL REGISTRATION

Trial Registration: International Standard Randomized Controlled Trial Number Registry ISRCTN94766568, https://www.isrctn.com/ISRCTN94766568.

Towards Personalised Mood Prediction and Explanation for Depression from Biophysical Data

Digital health applications using Artificial Intelligence (AI) are a promising opportunity to address the widening gap between available resources and mental health needs globally. Increasingly, passively acquired data from wearables are augmented with carefully selected active data from depressed individuals to develop Machine Learning (ML) models of depression based on mood scores. However, most ML models are black box in nature, and hence the outputs are not explainable. Depression is also multimodal, and the reasons for depression may vary significantly between individuals. Explainable and personalised models will thus be beneficial to clinicians to determine the main features that lead to a decline in the mood state of a depressed individual, thus enabling suitable personalised therapy. This is currently lacking. Therefore, this study presents a methodology for developing personalised and accurate Deep Learning (DL)-based predictive mood models for depression, along with novel methods for identifying the key facets that lead to the exacerbation of depressive symptoms. We illustrate our approach by using an existing multimodal dataset containing longitudinal Ecological Momentary Assessments of depression, lifestyle data from wearables and neurocognitive assessments for 14 mild to moderately depressed participants over one month. We develop classification- and regression-based DL models to predict participants' current mood scores-a discrete score given to a participant based on the severity of their depressive symptoms. The models are trained inside eight different evolutionary-algorithm-based optimisation schemes that optimise the model parameters for a maximum predictive performance. A five-fold cross-validation scheme is used to verify the DL model's predictive performance against 10 classical ML-based models, with a model error as low as 6% for some participants. We use the best model from the optimisation process to extract indicators, using SHAP, ALE and Anchors from explainable AI literature to explain why certain predictions are made and how they affect mood. These feature insights can assist health professionals in incorporating personalised interventions into a depressed individual's treatment regimen.

Distinct neural activations correlate with maximization of reward magnitude versus frequency

Choice selection strategies and decision-making are typically investigated using multiple-choice gambling paradigms that require participants to maximize expected value of rewards. However, research shows that performance in such paradigms suffers from individual biases towards the frequency of gains such that users often choose smaller frequent gains over larger rarely occurring gains, also referred to as melioration. To understand the basis of this subjective tradeoff, we used a simple 2-choice reward task paradigm in 186 healthy human adult subjects sampled across the adult lifespan. Cortical source reconstruction of simultaneously recorded electroencephalography suggested distinct neural correlates for maximizing reward magnitude versus frequency. We found that activations in the parahippocampal and entorhinal areas, which are typically linked to memory function, specifically correlated with maximization of reward magnitude. In contrast, maximization of reward frequency was correlated with activations in the lateral orbitofrontal cortices and operculum, typical areas involved in reward processing. These findings reveal distinct neural processes serving reward frequency versus magnitude maximization that can have clinical translational utility to optimize decision-making.

A generalized reinforcement learning based deep neural network agent model for diverse cognitive constructs

Human cognition is characterized by a wide range of capabilities including goal-oriented selective attention, distractor suppression, decision making, response inhibition, and working memory. Much research has focused on studying these individual components of cognition in isolation, whereas in several translational applications for cognitive impairment, multiple cognitive functions are altered in a given individual. Hence it is important to study multiple cognitive abilities in the same subject or, in computational terms, model them using a single model. To this end, we propose a unified, reinforcement learning-based agent model comprising of systems for representation, memory, value computation and exploration. We successfully modeled the aforementioned cognitive tasks and show how individual performance can be mapped to model meta-parameters. This model has the potential to serve as a proxy for cognitively impaired conditions, and can be used as a clinical testbench on which therapeutic interventions can be simulated first before delivering to human subjects.

Parietal alpha underlies slower cognitive responses during interference processing in adolescents

Adolescence is a critical period when cognitive control is rapidly maturing across several core dimensions. Here, we evaluated how healthy adolescents (13-17 years of age, n = 44) versus young adults (18-25 years of age, n = 49) differ across a series of cognitive assessments with simultaneous electroencephalography (EEG) recordings. Cognitive tasks included selective attention, inhibitory control, working memory, as well as both non-emotional and emotional interference processing. We found that adolescents displayed significantly slower responses than young adults specifically on the interference processing tasks. Evaluation of EEG event-related spectral perturbations (ERSPs) on the interference tasks showed that adolescents consistently had greater event-related desynchronization in alpha/beta frequencies in parietal regions. Midline frontal theta activity was also greater in the flanker interference task in adolescents, suggesting greater cognitive effort. Parietal alpha activity predicted age-related speed differences during non-emotional flanker interference processing, and frontoparietal connectivity, specifically midfrontal theta - parietal alpha functional connectivity predicted speed effects during emotional interference. Overall, our neuro-cognitive results illustrate developing cognitive control in adolescents particularly for interference processing, predicted by differential alpha band activity and connectivity in parietal brain regions.

EEG source derived salience network coupling supports real-world attention switching

While the brain mechanisms underlying selective attention have been studied in great detail in controlled laboratory settings, it is less clear how these processes function in the context of a real-world self-paced task. Here, we investigated engagement on a real-world computerized task equivalent to a standard academic test that consisted of solving high-school level problems in a self-paced manner. In this task, we used EEG-source derived estimates of effective coupling between brain sources to characterize the neural mechanisms underlying switches of sustained attention from the attentive on-task state to the distracted off-task state. Specifically, since the salience network has been implicated in sustained attention and attention switching, we conducted a hypothesis-driven analysis of effective coupling between the core nodes of the salience network, the anterior insula (AI) and the anterior cingulate cortex (ACC). As per our hypothesis, we found an increase in AI - > ACC effective coupling that occurs during the transitions of attention from on-task focused to off-task distracted state. This research may inform the development of future neural function-targeted brain-computer interfaces to enhance sustained attention.

Simultaneous Gut-Brain Electrophysiology Shows Cognition and Satiety Specific Coupling

Recent studies, using high resolution magnetoencephalography (MEG) and electrogastrography (EGG), have shown that during resting state, rhythmic gastric physiological signals are linked with cortical brain oscillations. Yet, gut-brain coupling has not been investigated with electroencephalography (EEG) during cognitive brain engagement or during hunger-related gut engagement. In this study in 14 young adults (7 females, mean ± SD age 25.71 ± 8.32 years), we study gut-brain coupling using simultaneous EEG and EGG during hunger and satiety states measured in separate visits, and compare responses both while resting as well as during a cognitively demanding working memory task. We find that EGG-EEG phase-amplitude coupling (PAC) differs based on both satiety state and cognitive effort, with greater PAC modulation observed in the resting state relative to working memory. We find a significant interaction between gut satiation levels and cognitive states in the left fronto-central brain region, with larger cognitive demand based differences in the hunger state. Furthermore, strength of PAC correlated with behavioral performance during the working memory task. Altogether, these results highlight the role of gut-brain interactions in cognition and demonstrate the feasibility of these recordings using scalable sensors.

Neural dynamics during emotional video engagement relate to anxiety

Inter-subject correlations (ISCs) of physiological data can reveal common stimulus-driven processing across subjects. ISC has been applied to passive video viewing in small samples to measure common engagement and emotional processing. Here, in a large sample study of healthy adults (N = 163) who watched an emotional film (The Lion Cage by Charlie Chaplin), we recorded electroencephalography (EEG) across participants and measured ISC in theta, alpha and beta frequency bands. Peak ISC on the emotionally engaging video was observed three-quarters into the film clip, during a time period which potentially elicited a positive emotion of relief. Peak ISC in all frequency bands was focused over centro-parietal electrodes localizing to superior parietal cortex. ISC in both alpha and beta frequencies had a significant inverse relationship with anxiety symptoms. Our study suggests that ISC measured during continuous non-event-locked passive viewing may serve as a useful marker for anxious mood.

Dissociable neural mechanisms of cognition and well-being in youth versus healthy aging

Mental health, cognition, and their underlying neural processes in healthy aging are rarely studied simultaneously. Here, in a sample of healthy younger (n = 62) and older (n = 54) adults, we compared subjective mental health as well as objective global cognition across several core cognitive domains with simultaneous electroencephalography (EEG). We found significantly greater symptoms of anxiety, depression, and loneliness in youth and in contrast, greater mental well-being in older adults. Yet, global performance across core cognitive domains was significantly worse in older adults. EEG-based source imaging of global cognitive task-evoked processing showed reduced suppression of activity in the anterior medial prefrontal default mode network (DMN) region in older adults relative to youth. Global cognitive performance efficiency was predicted by greater activity in the right dorsolateral prefrontal cortex in younger adults and in contrast, by greater activity in right inferior frontal cortex in older adults. Furthermore, greater mental well-being in older adults related to lesser global task-evoked activity in the posterior DMN. Overall, these results suggest dissociated neural mechanisms underlying global cognition and mental well-being in youth versus healthy aging. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Utility of Cognitive Neural Features for Predicting Mental Health Behaviors

Cognitive dysfunction underlies common mental health behavioral symptoms including depression, anxiety, inattention, and hyperactivity. In this study of 97 healthy adults, we aimed to classify healthy vs. mild-to-moderate self-reported symptoms of each disorder using cognitive neural markers measured with an electroencephalography (EEG). We analyzed source-reconstructed EEG data for event-related spectral perturbations in the theta, alpha, and beta frequency bands in five tasks, a selective attention and response inhibition task, a visuospatial working memory task, a Flanker interference processing task, and an emotion interference task. From the cortical source activation features, we derived augmented features involving co-activations between any two sources. Logistic regression on the augmented feature set, but not the original feature set, predicted the presence of psychiatric symptoms, particularly for anxiety and inattention with >80% sensitivity and specificity. We also computed current flow closeness and betweenness centralities to identify the “hub” source signal predictors. We found that the Flanker interference processing task was the most useful for assessing the connectivity hubs in general, followed by the inhibitory control go-nogo paradigm. Overall, these interpretable machine learning analyses suggest that EEG biomarkers collected on a rapid suite of cognitive assessments may have utility in classifying diverse self-reported mental health symptoms.

Personalized machine learning of depressed mood using wearables

Depression is a multifaceted illness with large interindividual variability in clinical response to treatment. In the era of digital medicine and precision therapeutics, new personalized treatment approaches are warranted for depression. Here, we use a combination of longitudinal ecological momentary assessments of depression, neurocognitive sampling synchronized with electroencephalography, and lifestyle data from wearables to generate individualized predictions of depressed mood over a 1-month time period. This study, thus, develops a systematic pipeline for N-of-1 personalized modeling of depression using multiple modalities of data. In the models, we integrate seven types of supervised machine learning (ML) approaches for each individual, including ensemble learning and regression-based methods. All models were verified using fourfold nested cross-validation. The best-fit as benchmarked by the lowest mean absolute percentage error, was obtained by a different type of ML model for each individual, demonstrating that there is no one-size-fits-all strategy. The voting regressor, which is a composite strategy across ML models, was best performing on-average across subjects. However, the individually selected best-fit models still showed significantly less error than the voting regressor performance across subjects. For each individual's best-fit personalized model, we further extracted top-feature predictors using Shapley statistics. Shapley values revealed distinct feature determinants of depression over time for each person ranging from co-morbid anxiety, to physical exercise, diet, momentary stress and breathing performance, sleep times, and neurocognition. In future, these personalized features can serve as targets for a personalized ML-guided, multimodal treatment strategy for depression.

Electrophysiological Correlates of Rodent Default-Mode Network Suppression Revealed by Large-Scale Local Field Potential Recordings

The default-mode network (DMN) in humans consists of a set of brain regions that, as measured with functional magnetic resonance imaging (fMRI), show both intrinsic correlations with each other and suppression during externally oriented tasks. Resting-state fMRI studies have previously identified similar patterns of intrinsic correlations in overlapping brain regions in rodents (A29C/posterior cingulate cortex, parietal cortex, and medial temporal lobe structures). However, due to challenges with performing rodent behavior in an MRI machine, it is still unclear whether activity in rodent DMN regions are suppressed during externally oriented visual tasks. Using distributed local field potential measurements in rats, we have discovered that activity in DMN brain regions noted above show task-related suppression during an externally oriented visual task at alpha and low beta-frequencies. Interestingly, this suppression (particularly in posterior cingulate cortex) was linked with improved performance on the task. Using electroencephalography recordings from a similar task in humans, we identified a similar suppression of activity in posterior cingulate cortex at alpha/low beta-frequencies. Thus, we have identified a common electrophysiological marker of DMN suppression in both rodents and humans. This observation paves the way for future studies using rodents to probe circuit-level functioning of DMN function.

SIGNIFICANCE

Here we show that alpha/beta frequency oscillations in rats show key features of DMN activity, including intrinsic correlations between DMN brain regions, task-related suppression, and interference with attention/decision-making. We found similar task-related suppression at alpha/low beta-frequencies of DMN activity in humans.

Mapping Large-Scale Networks Associated with Action, Behavioral Inhibition and Impulsivity

A key aspect of behavioral inhibition is the ability to wait before acting. Failures in this form of inhibition result in impulsivity and are commonly observed in various neuropsychiatric disorders. Prior evidence has implicated medial frontal cortex, motor cortex, orbitofrontal cortex (OFC), and ventral striatum in various aspects of inhibition. Here, using distributed recordings of brain activity [with local-field potentials (LFPs)] in rodents, we identified oscillatory patterns of activity linked with action and inhibition. Low-frequency (δ) activity within motor and premotor circuits was observed in two distinct networks, the first involved in cued, sensory-based responses and the second more generally in both cued and delayed actions. By contrast, θ activity within prefrontal and premotor regions (medial frontal cortex, OFC, ventral striatum, and premotor cortex) was linked with inhibition. Connectivity at θ frequencies was observed within this network of brain regions. Interestingly, greater connectivity between primary motor cortex (M1) and other motor regions was linked with greater impulsivity, whereas greater connectivity between M1 and inhibitory brain regions (OFC, ventral striatum) was linked with improved inhibition and diminished impulsivity. We observed similar patterns of activity on a parallel task in humans: low-frequency activity in sensorimotor cortex linked with action, θ activity in OFC/ventral prefrontal cortex (PFC) linked with inhibition. Thus, we show that δ and θ oscillations form distinct large-scale networks associated with action and inhibition, respectively.