Testing cognitive theories with multivariate pattern analysis of neuroimaging data

Neural circuit basis of pathological anxiety

Anxiety disorders are the most prevalent mental health conditions worldwide. Unfortunately, the understanding of the precise neurobiological mechanisms that underlie these disorders remains limited. Current diagnostic classifications, based on observable symptoms rather than underlying pathophysiology, do not capture the heterogeneity within and across anxiety disorders. Recent advances in functional neuroimaging have provided new insights into the neural circuits implicated in pathological anxiety, revealing dysfunctions that cut across traditional diagnostic boundaries. In this Review, we synthesize evidence that highlights abnormalities in neurobehavioural systems related to negative valence, positive valence, cognitive systems and social processes. We emphasize that pathological anxiety arises not only from heightened reactivity in acute threat ('fear') circuits but also from alterations in circuits that mediate distant (potential) and sustained threat, reward processing, cognitive control and social processing. We discuss how circuit vulnerabilities can lead to the emergence and maintenance of pathological anxiety. Once established, these neural abnormalities can be exacerbated by maladaptive behaviours that prevent extinction learning and perpetuate anxiety disorders. By delineating the specific neural mechanisms in each neurobiological system, we aim to contribute to a more comprehensive understanding of the neurobiology of anxiety disorders, potentially informing future research directions in this field.

Aperiodic and oscillatory systems underpinning human domain-general cognition

Domain-general cognitive systems are essential for adaptive human behaviour, supporting various cognitive tasks through flexible neural mechanisms. While fMRI studies link frontoparietal network activation to increasing demands across various tasks, the electrophysiological mechanisms underlying this domain-general response to demand remain unclear. Here, we used MEG/EEG, and separated the aperiodic and oscillatory components of the signals to examine their roles in domain-general cognition across three cognitive tasks using multivariate analysis. We found that both aperiodic (broadband power, slope, and intercept) and oscillatory (theta, alpha, and beta power) components coded task demand and content across all subtasks. Aperiodic broadband power in particular strongly coded task demand, in a manner that generalised across all subtasks. Source estimation suggested that increasing cognitive demand decreased aperiodic broadband power across the brain, with the strongest modulations overlapping with the frontoparietal network. In contrast, oscillatory activity showed more localised patterns of modulation, primarily in frontal or occipital regions. These results provide insights into the electrophysiological underpinnings of human domain-general cognition, highlighting the critical role of aperiodic broadband power.

Encoding models for developmental cognitive computational neuroscience: Promise, challenges, and potential

Cognitive computational neuroscience has received broad attention in recent years as an emerging area integrating cognitive science, neuroscience, and artificial intelligence. At the heart of this field, approaches using encoding models allow for explaining brain activity from latent and high-dimensional features, including artificial neural networks. With the notable exception of temporal response function models that are applied to electroencephalography, most prior studies have focused on adult subjects, making it difficult to capture how brain representations change with learning and development. Here, we argue that future developmental cognitive neuroscience studies would benefit from approaches relying on encoding models. We provide an overview of encoding models used in adult functional magnetic resonance imaging research. This research has notably used data with a small number of subjects, but with a large number of samples per subject. Studies using encoding models also generally require task-based neuroimaging data. Though these represent challenges for developmental studies, we argue that these challenges may be overcome by using functional alignment techniques and naturalistic paradigms. These methods would facilitate encoding model analysis in developmental neuroimaging research, which may lead to important theoretical advances.

Enhancing prediction of human traits and behaviors through ensemble learning of traditional and novel resting-state fMRI connectivity analyses

Recent advances in cognitive neuroscience have focused on using resting-state functional connectivity (RSFC) data from fMRI scans to more accurately predict human traits and behaviors. Traditional approaches generally analyze RSFC by correlating averaged time-series data across regions of interest (ROIs) or networks, which may overlook important spatial signal patterns. To address this limitation, we introduced a novel linear regression technique that estimates RSFC by predicting spatial brain activity patterns in a target ROI from those in a seed ROI. We applied both traditional and our novel RSFC estimation methods to a large-scale dataset from the Human Connectome Project and the Brain Genomics Superstruct Project, analyzing resting-state fMRI data to predict sex, age, personality traits, and psychological task performance. To enhance prediction accuracy, we developed an ensemble learner that combines these qualitatively different methods using a weighted average approach. Our findings revealed that hierarchical clustering of RSFC patterns using our novel method displays distinct whole-brain grouping patterns compared to the traditional approach. Importantly, the ensemble model, integrating these diverse weak learners, outperformed the traditional RSFC method in predicting human traits and behaviors. Notably, the predictions from the traditional and novel methods showed relatively low similarity, indicating that our novel approach captures unique and previously undetected information about human traits and behaviors through fine-grained local spatial patterns of neural activation. These results highlight the potential of combining traditional and innovative RSFC analysis techniques to enrich our understanding of the neural basis of human traits and behaviors.

Neural Predictors of Fear Depend on the Situation

The extent to which neural representations of fear experience depend on or generalize across the situational context has remained unclear. We systematically manipulated variation within and across three distinct fear-evocative situations including fear of heights, spiders, and social threats. Participants (n = 21; 10 females and 11 males) viewed ∼20 s clips depicting spiders, heights, or social encounters and rated fear after each video. Searchlight multivoxel pattern analysis was used to identify whether and which brain regions carry information that predicts fear experience and the degree to which the fear-predictive neural codes in these areas depend on or generalize across the situations. The overwhelming majority of brain regions carrying information about fear did so in a situation-dependent manner. These findings suggest that local neural representations of fear experience are unlikely to involve a singular pattern but rather a collection of multiple heterogeneous brain states.

A shared threat-anticipation circuit is dynamically engaged at different moments by certain and uncertain threat

Temporal dynamics play a central role in models of emotion: "fear" is widely conceptualized as a phasic response to certain-and-imminent danger, whereas "anxiety" is a sustained response to uncertain-or-distal harm. Yet the underlying neurobiology remains contentious. Leveraging a translationally relevant fMRI paradigm and theory-driven modeling approach, we demonstrate that certain- and uncertain-threat anticipation recruit a shared circuit that encompasses the central extended amygdala (EAc), periaqueductal gray, midcingulate, and anterior insula. This circuit exhibits persistently elevated activation when threat is uncertain and distal, and transient bursts of activation just before certain encounters with threat. Although there is widespread agreement that the EAc plays a critical role in orchestrating responses to threat, confusion persists about the respective contributions of its major subdivisions, the bed nucleus of the stria terminalis (BST) and central nucleus of the amygdala (Ce). Here we used anatomical regions-of-interest to demonstrate that the BST and Ce exhibit statistically indistinguishable threat dynamics, with moderate-to-strong Bayesian evidence for the null hypothesis. Both regions exhibited activation dynamics that run counter to popular models, with the Ce showing sustained responses to uncertain-and-distal threat and the BST showing phasic responses to certain-and-imminent threat. For many scientists, feelings are the hallmark of fear and anxiety. Here we used an independently validated brain 'signature' to covertly decode the moment-by-moment dynamics of anticipatory distress for the first time. Results mirrored the dynamics of neural activation. These observations provide fresh insights into the neurobiology of threat-elicited emotions and set the stage for more ambitious clinical and mechanistic research.

Decoding cognition in neurodevelopmental, psychiatric and neurological conditions with multivariate pattern analysis of EEG data

Multivariate pattern analysis (MVPA) of electroencephalographic (EEG) data represents a revolutionary approach to investigate how the brain encodes information. By considering complex interactions among spatio-temporal features at the individual level, MVPA overcomes the limitations of univariate techniques, which often fail to account for the significant inter- and intra-individual neural variability. This is particularly relevant when studying clinical populations, and therefore MVPA of EEG data has recently started to be employed as a tool to study cognition in brain disorders. Here, we review the insights offered by this methodology in the study of anomalous patterns of neural activity in conditions such as autism, ADHD, schizophrenia, dyslexia, neurological and neurodegenerative disorders, within different cognitive domains (perception, attention, memory, consciousness). Despite potential drawbacks that should be attentively addressed, these studies reveal a peculiar sensitivity of MVPA in unveiling dysfunctional and compensatory neurocognitive dynamics of information processing, which often remain blind to traditional univariate approaches. Such higher sensitivity in characterizing individual neurocognitive profiles can provide unique opportunities to optimise assessment and promote personalised interventions.

Spatiotemporal Properties of Common Semantic Categories for Words and Pictures

The timing of semantic processing during object recognition in the brain is a topic of ongoing discussion. One way of addressing this question is by applying multivariate pattern analysis to human electrophysiological responses to object images of different semantic categories. However, although multivariate pattern analysis can reveal whether neuronal activity patterns are distinct for different stimulus categories, concerns remain on whether low-level visual features also contribute to the classification results. To circumvent this issue, we applied a cross-decoding approach to magnetoencephalography data from stimuli from two different modalities: images and their corresponding written words. We employed items from three categories and presented them in a randomized order. We show that if the classifier is trained on words, pictures are classified between 150 and 430 msec after stimulus onset, and when training on pictures, words are classified between 225 and 430 msec. The topographical map, identified using a searchlight approach for cross-modal activation in both directions, showed left lateralization, confirming the involvement of linguistic representations. These results point to semantic activation of pictorial stimuli occurring at ∼150 msec, whereas for words, the semantic activation occurs at ∼230 msec.

A neurofunctional signature of subjective disgust generalizes to oral distaste and socio-moral contexts

While disgust originates in the hard-wired mammalian distaste response, the conscious experience of disgust in humans strongly depends on subjective appraisal and may even extend to socio-moral contexts. Here, in a series of studies, we combined functional magnetic resonance imaging with machine-learning-based predictive modelling to establish a comprehensive neurobiological model of subjective disgust. The developed neurofunctional signature accurately predicted momentary self-reported subjective disgust across discovery (n = 78) and pre-registered validation (n = 30) cohorts and generalized across core disgust (n = 34 and n = 26), gustatory distaste (n = 30) and socio-moral (unfair offers; n = 43) contexts. Disgust experience was encoded in distributed cortical and subcortical systems, and exhibited distinct and shared neural representations with subjective fear or negative affect in interoceptive-emotional awareness and conscious appraisal systems, while the signatures most accurately predicted the respective target experience. We provide an accurate functional magnetic resonance imaging signature for disgust with a high potential to resolve ongoing evolutionary debates.

Predictions and errors are distinctly represented across V1 layers

Popular accounts of mind and brain propose that the brain continuously forms predictions about future sensory inputs and combines predictions with inputs to determine what we perceive.1,2,3,4,5,6 Under "predictive processing" schemes, such integration is supported by the hierarchical organization of the cortex, whereby feedback connections communicate predictions from higher-level deep layers to agranular (superficial and deep) lower-level layers.7,8,9,10 Predictions are compared with input to compute the "prediction error," which is transmitted up the hierarchy from superficial layers of lower cortical regions to the middle layers of higher areas, to update higher-level predictions until errors are reconciled.11,12,13,14,15 In the primary visual cortex (V1), predictions have thereby been proposed to influence representations in deep layers while error signals may be computed in superficial layers. Despite the framework's popularity, there is little evidence for these functional distinctions because, to our knowledge, unexpected sensory events have not previously been presented in human laminar paradigms to contrast against expected events. To this end, this 7T fMRI study contrasted V1 responses to expected (75% likely) and unexpected (25%) Gabor orientations. Multivariate decoding analyses revealed an interaction between expectation and layer, such that expected events could be decoded with comparable accuracy across layers, while unexpected events could only be decoded in superficial laminae. Although these results are in line with these accounts that have been popular for decades, such distinctions have not previously been demonstrated in humans. We discuss how both prediction and error processes may operate together to shape our unitary perceptual experiences.

The relationship between the resting state functional connectivity and social cognition in schizophrenia: Results from the Italian Network for Research on Psychoses

Deficits in social cognition (SC) interfere with recovery in schizophrenia (SZ) and may be related to resting state brain connectivity. This study aimed at assessing the alterations in the relationship between resting state functional connectivity and the social-cognitive abilities of patients with SZ compared to healthy subjects. We divided the brain into 246 regions of interest (ROI) following the Human Healthy Volunteers Brainnetome Atlas. For each participant, we calculated the resting-state functional connectivity (rsFC) in terms of degree centrality (DC), which evaluates the total strength of the most powerful coactivations of every ROI with all other ROIs during rest. The rs-DC of the ROIs was correlated with five measures of SC assessing emotion processing and mentalizing in 45 healthy volunteers (HVs) chosen as a normative sample. Then, controlling for symptoms severity, we verified whether these significant associations were altered, i.e., absent or of opposite sign, in 55 patients with SZ. We found five significant differences between SZ patients and HVs: in the patients' group, the correlations between emotion recognition tasks and rsFC of the right entorhinal cortex (R-EC), left superior parietal lobule (L-SPL), right caudal hippocampus (R-c-Hipp), and the right caudal (R-c) and left rostral (L-r) middle temporal gyri (MTG) were lost. An altered resting state functional connectivity of the L-SPL, R-EC, R-c-Hipp, and bilateral MTG in patients with SZ may be associated with impaired emotion recognition. If confirmed, these results may enhance the development of non-invasive brain stimulation interventions targeting those cerebral regions to reduce SC deficit in SZ.

A neural signature for the subjective experience of threat anticipation under uncertainty

Uncertainty about potential future threats and the associated anxious anticipation represents a key feature of anxiety. However, the neural systems that underlie the subjective experience of threat anticipation under uncertainty remain unclear. Combining an uncertainty-variation threat anticipation paradigm that allows precise modulation of the level of momentary anxious arousal during functional magnetic resonance imaging (fMRI) with multivariate predictive modeling, we train a brain model that accurately predicts subjective anxious arousal intensity during anticipation and test it across 9 samples (total n = 572, both gender). Using publicly available datasets, we demonstrate that the whole-brain signature specifically predicts anxious anticipation and is not sensitive in predicting pain, general anticipation or unspecific emotional and autonomic arousal. The signature is also functionally and spatially distinguishable from representations of subjective fear or negative affect. We develop a sensitive, generalizable, and specific neuroimaging marker for the subjective experience of uncertain threat anticipation that can facilitate model development.

Scene-selectivity in CA1/subicular complex: Multivoxel pattern analysis at 7T

Prior univariate functional magnetic resonance imaging (fMRI) studies in humans suggest that the anteromedial subicular complex of the hippocampus is a hub for scene-based cognition. However, it is possible that univariate approaches were not sufficiently sensitive to detect scene-related activity in other subfields that have been implicated in spatial processing (e.g., CA1). Further, as connectivity-based functional gradients in the hippocampus do not respect classical subfield boundary definitions, category selectivity may be distributed across anatomical subfields. Region-of-interest approaches, therefore, may limit our ability to observe category selectivity across discrete subfield boundaries. To address these issues, we applied searchlight multivariate pattern analysis to 7T fMRI data of healthy adults who undertook a simultaneous visual odd-one-out discrimination task for scene and non-scene (including face) visual stimuli, hypothesising that scene classification would be possible in multiple hippocampal regions within, but not constrained to, anteromedial subicular complex and CA1. Indeed, we found that the scene-selective searchlight map overlapped not only with anteromedial subicular complex (distal subiculum, pre/para subiculum), but also inferior CA1, alongside posteromedial (including retrosplenial) and parahippocampal cortices. Probabilistic overlap maps revealed gradients of scene category selectivity, with the strongest overlap located in the medial hippocampus, converging with searchlight findings. This was contrasted with gradients of face category selectivity, which had stronger overlap in more lateral hippocampus, supporting ideas of parallel processing streams for these two categories. Our work helps to map the scene, in contrast to, face processing networks within, and connected to, the human hippocampus.