Methods for computing the maximum performance of computational models of fMRI responses

Modeling short visual events through the BOLD moments video fMRI dataset and metadata

Studying the neural basis of human dynamic visual perception requires extensive experimental data to evaluate the large swathes of functionally diverse brain neural networks driven by perceiving visual events. Here, we introduce the BOLD Moments Dataset (BMD), a repository of whole-brain fMRI responses to over 1000 short (3 s) naturalistic video clips of visual events across ten human subjects. We use the videos' extensive metadata to show how the brain represents word- and sentence-level descriptions of visual events and identify correlates of video memorability scores extending into the parietal cortex. Furthermore, we reveal a match in hierarchical processing between cortical regions of interest and video-computable deep neural networks, and we showcase that BMD successfully captures temporal dynamics of visual events at second resolution. With its rich metadata, BMD offers new perspectives and accelerates research on the human brain basis of visual event perception.

Disentangling signal and noise in neural responses through generative modeling

Measurements of neural responses to identically repeated experimental events often exhibit large amounts of variability. This noise is distinct from signal , operationally defined as the average expected response across repeated trials for each given event. Accurately distinguishing signal from noise is important, as each is a target that is worthy of study (many believe noise reflects important aspects of brain function) and it is important not to confuse one for the other. Here, we introduce a principled modeling approach in which response measurements are explicitly modeled as the sum of samples from multivariate signal and noise distributions. In our proposed method-termed Generative Modeling of Signal and Noise (GSN)-the signal distribution is estimated by subtracting the estimated noise distribution from the estimated data distribution. We validate GSN using ground-truth simulations and demonstrate the application of GSN to empirical fMRI data. In doing so, we illustrate a simple consequence of GSN: by disentangling signal and noise components in neural responses, GSN denoises principal components analysis and improves estimates of dimensionality. We end by discussing other situations that may benefit from GSN's characterization of signal and noise, such as estimation of noise ceilings for computational models of neural activity. A code toolbox for GSN is provided with both MATLAB and Python implementations.

Implicit racial biases are lower in more populous more diverse and less segregated US cities

Implicit biases - differential attitudes towards members of distinct groups - are pervasive in human societies and create inequities across many aspects of life. Recent research has revealed that implicit biases are generally driven by social contexts, but not whether they are systematically influenced by the ways that humans self-organize in cities. We leverage complex system modeling in the framework of urban scaling theory to predict differences in these biases between cities. Our model links spatial scales from city-wide infrastructure to individual psychology to predict that cities that are more populous, more diverse, and less segregated are less biased. We find empirical support for these predictions in U.S. cities with Implicit Association Test data spanning a decade from 2.7 million individuals and U.S. Census demographic data. Additionally, we find that changes in cities' social environments precede changes in implicit biases at short time-scales, but this relationship is bi-directional at longer time-scales. We conclude that the social organization of cities may influence the strength of these biases.

Driving and suppressing the human language network using large language models

Transformer models such as GPT generate human-like language and are highly predictive of human brain responses to language. Here, using fMRI-measured brain responses to 1,000 diverse sentences, we first show that a GPT-based encoding model can predict the magnitude of brain response associated with each sentence. Then, we use the model to identify new sentences that are predicted to drive or suppress responses in the human language network. We show that these model-selected novel sentences indeed strongly drive and suppress activity of human language areas in new individuals. A systematic analysis of the model-selected sentences reveals that surprisal and well-formedness of linguistic input are key determinants of response strength in the language network. These results establish the ability of neural network models to not only mimic human language but also noninvasively control neural activity in higher-level cortical areas, like the language network.

Neural encoding with unsupervised spiking convolutional neural network

Accurately predicting the brain responses to various stimuli poses a significant challenge in neuroscience. Despite recent breakthroughs in neural encoding using convolutional neural networks (CNNs) in fMRI studies, there remain critical gaps between the computational rules of traditional artificial neurons and real biological neurons. To address this issue, a spiking CNN (SCNN)-based framework is presented in this study to achieve neural encoding in a more biologically plausible manner. The framework utilizes unsupervised SCNN to extract visual features of image stimuli and employs a receptive field-based regression algorithm to predict fMRI responses from the SCNN features. Experimental results on handwritten characters, handwritten digits and natural images demonstrate that the proposed approach can achieve remarkably good encoding performance and can be utilized for "brain reading" tasks such as image reconstruction and identification. This work suggests that SNN can serve as a promising tool for neural encoding.

Voxelwise encoding models of body stimuli reveal a representational gradient from low-level visual features to postural features in occipitotemporal cortex

Research on body representation in the brain has focused on category-specific representation, using fMRI to investigate the response pattern to body stimuli in occipitotemporal cortex without so far addressing the issue of the specific computations involved in body selective regions, only defined by higher order category selectivity. This study used ultra-high field fMRI and banded ridge regression to investigate the coding of body images, by comparing the performance of three encoding models in predicting brain activity in occipitotemporal cortex and specifically the extrastriate body area (EBA). Our results suggest that bodies are encoded in occipitotemporal cortex and in the EBA according to a combination of low-level visual features and postural features.

Testing, explaining, and exploring models of facial expressions of emotions

Models are the hallmark of mature scientific inquiry. In psychology, this maturity has been reached in a pervasive question-what models best represent facial expressions of emotion? Several hypotheses propose different combinations of facial movements [action units (AUs)] as best representing the six basic emotions and four conversational signals across cultures. We developed a new framework to formalize such hypotheses as predictive models, compare their ability to predict human emotion categorizations in Western and East Asian cultures, explain the causal role of individual AUs, and explore updated, culture-accented models that improve performance by reducing a prevalent Western bias. Our predictive models also provide a noise ceiling to inform the explanatory power and limitations of different factors (e.g., AUs and individual differences). Thus, our framework provides a new approach to test models of social signals, explain their predictive power, and explore their optimization, with direct implications for theory development.

Accurate predictions of individual differences in task-evoked brain activity from resting-state fMRI using a sparse ensemble learner

Modelling and predicting individual differences in task-fMRI activity can have a wide range of applications from basic to clinical neuroscience. It has been shown that models based on resting-state activity can have high predictive accuracy. Here we propose several improvements to such models. Using a sparse ensemble learner, we show that (i) features extracted using Stochastic Probabilistic Functional Modes (sPROFUMO) outperform the previously proposed dual-regression approach, (ii) that the shape and overall intensity of individualised task activations can be modelled separately and explicitly, (iii) training the model on predicting residual differences in brain activity further boosts individualised predictions. These results hold for both surface-based analyses of the Human Connectome Project data as well as volumetric analyses of UK-biobank data. Overall, our model achieves state of the art prediction accuracy on par with the test-retest reliability of task-fMRI scans, suggesting that it has potential to supplement traditional task localisers.

Predicting brain activation maps for arbitrary tasks with cognitive encoding models

A deep understanding of the neural architecture of mental function should enable the accurate prediction of a specific pattern of brain activity for any psychological task, based only on the cognitive functions known to be engaged by that task. Encoding models (EMs), which predict neural responses from known features (e.g., stimulus properties), have succeeded in circumscribed domains (e.g., visual neuroscience), but implementing domain-general EMs that predict brain-wide activity for arbitrary tasks has been limited mainly by availability of datasets that 1) sufficiently span a large space of psychological functions, and 2) are sufficiently annotated with such functions to allow robust EM specification. We examine the use of EMs based on a formal specification of psychological function, to predict cortical activation patterns across a broad range of tasks. We utilized the Multi-Domain Task Battery, a dataset in which 24 subjects completed 32 ten-minute fMRI scans, switching tasks every 35 s and engaging in 44 total conditions of diverse psychological manipulations. Conditions were annotated by a group of experts using the Cognitive Atlas ontology to identify putatively engaged functions, and region-wise cognitive EMs (CEMs) were fit, for individual subjects, on neocortical responses. We found that CEMs predicted cortical activation maps of held-out tasks with high accuracy, outperforming a permutation-based null model while approaching the noise ceiling of the data, without being driven solely by either cognitive or perceptual-motor features. Hierarchical clustering on the similarity structure of CEM generalization errors revealed relationships amongst psychological functions. Spatial distributions of feature importances systematically overlapped with large-scale resting-state functional networks (RSNs), supporting the hypothesis of functional specialization within RSNs while grounding their function in an interpretable data-driven manner. Our implementation and validation of CEMs provides a proof of principle for the utility of formal ontologies in cognitive neuroscience and motivates the use of CEMs in the further testing of cognitive theories.

Self-supervised Natural Image Reconstruction and Large-scale Semantic Classification from Brain Activity

Reconstructing natural images and decoding their semantic category from fMRI brain recordings is challenging. Acquiring sufficient pairs of images and their corresponding fMRI responses, which span the huge space of natural images, is prohibitive. We present a novel self-supervised approach that goes well beyond the scarce paired data, for achieving both: (i) state-of-the art fMRI-to-image reconstruction, and (ii) first-ever large-scale semantic classification from fMRI responses. By imposing cycle consistency between a pair of deep neural networks (from image-to-fMRI & from fMRI-to-image), we train our image reconstruction network on a large number of "unpaired" natural images (images without fMRI recordings) from many novel semantic categories. This enables to adapt our reconstruction network to a very rich semantic coverage without requiring any explicit semantic supervision. Specifically, we find that combining our self-supervised training with high-level perceptual losses, gives rise to new reconstruction & classification capabilities. In particular, this perceptual training enables to classify well fMRIs of never-before-seen semantic classes, without requiring any class labels during training. This gives rise to: (i) Unprecedented image-reconstruction from fMRI of never-before-seen images (evaluated by image metrics and human testing), and (ii) Large-scale semantic classification of categories that were never-before-seen during network training. Such large-scale (1000-way) semantic classification from fMRI recordings has never been demonstrated before. Finally, we provide evidence for the biological consistency of our learned model.

Neural similarity between mentalizing and live social interaction during the transition to adolescence

Social interactions are essential for human development, yet little neuroimaging research has examined their underlying neurocognitive mechanisms using socially interactive paradigms during childhood and adolescence. Recent neuroimaging research has revealed activity in the mentalizing network when children engage with a live social partner, even when mentalizing is not required. While this finding suggests that social‐interactive contexts may spontaneously engage mentalizing, it is not a direct test of how similarly the brain responds to these two contexts. The current study used representational similarity analysis on data from 8‐ to 14‐year‐olds who made mental and nonmental judgments about an abstract character and a live interaction partner during fMRI. A within‐subject, 2 (Mental/Nonmental) × 2 (Peer/Character) design enabled us to examine response pattern similarity between conditions, and estimate fit to three conceptual models of how the two contexts relate: (1) social interaction and mentalizing about an abstract character are represented similarly; (2) interactive peers and abstract characters are represented differently regardless of the evaluation type; and (3) mental and nonmental states are represented dissimilarly regardless of target. We found that the temporal poles represent mentalizing and peer interactions similarly (Model 1), suggesting a neurocognitive link between the two in these regions. Much of the rest of the social brain exhibits different representations of interactive peers and abstract characters (Model 2). Our findings highlight the importance of studying social‐cognitive processes using interactive approaches, and the utility of pattern‐based analyses for understanding how social‐cognitive processes relate to each other.

Brain networks are decoupled from external stimuli during internal cognition

Our cognition can be directed to external stimuli or to internal information. While there are many different forms of internal cognition (mind-wandering, recall, imagery etc.), their essential feature is independence from the immediate sensory input, conceptually referred to as perceptual decoupling. Perceptual decoupling is thought to be reflected in brain activity transitioning from a stimulus-processing to internally-processing mode, but a direct investigation of this remains outstanding. Here we present a conceptual and analysis framework that quantifies the extent to which brain networks reflect stimulus processing. We tested this framework by presenting subjects with an audiovisual stimulus and instructing them to either attend to the stimulus (external task) or engage in mental imagery, recall or arithmetic (internal tasks) while measuring the evoked brain activity using functional MRI. We found that stimulus responses were generally attenuated for the internal tasks, though they increased in a subset of tasks and brain networks. However, using our new framework, we showed that brain networks became less reflective of stimulus processing, even in the subset of tasks and brain networks in which stimulus responses increased. These results quantitatively demonstrate that during internal cognition brain networks become decoupled from the external stimuli, opening the door for a fundamental and quantitative understanding of internal cognition.

The 3d Mind Model characterizes how people understand mental states across modern and historical cultures

Humans rely on social interaction to achieve many important goals. These interactions rely in turn on people's capacity to understand others' mental states: their thoughts and feelings. Do different cultures understand minds in different ways, or do widely shared principles describe how different cultures understand mental states? Extensive data suggest that the mind organizes mental state concepts using the 3d Mind Model, composed of the psychological dimensions: rationality (vs. emotionality), social impact (states which affect others more vs. less), and valence (positive vs. negative states). However, this evidence comes primarily from English-speaking individuals in the United States. Here we investigated mental state representation in 57 contemporary countries, using 163 million English language tweets; in 17 languages, using billions of words of text from internet webpages; and across more than 2000 years of history, using curated texts from four historical societies. We quantified mental state meaning by analyzing the text produced by each culture using word embeddings. We then tested whether the 3d Mind Model could explain which mental states were similar in meaning within each culture. We found that the 3d Mind Model significantly explained mental state meaning in every country, language, and historical society that we examined. These results suggest that rationality, social impact, and valence form a generalizable conceptual backbone for mental state representation.

Left hemisphere dominance for bilateral kinematic encoding in the human brain

Neurophysiological studies in humans and nonhuman primates have revealed movement representations in both the contralateral and ipsilateral hemispheres. Inspired by clinical observations, we ask if this bilateral representation differs for the left and right hemispheres. Electrocorticography was recorded in human participants during an instructed-delay reaching task, with movements produced with either the contralateral or ipsilateral arm. Using a cross-validated kinematic encoding model, we found stronger bilateral encoding in the left hemisphere, an effect that was present during preparation and was amplified during execution. Consistent with this asymmetry, we also observed better across-arm generalization in the left hemisphere, indicating similar neural representations for right and left arm movements. Notably, these left hemisphere electrodes were centered over premotor and parietal regions. The more extensive bilateral encoding in the left hemisphere adds a new perspective to the pervasive neuropsychological finding that the left hemisphere plays a dominant role in praxis.

Time-resolved multivariate pattern analysis of infant EEG data: A practical tutorial

Time-resolved multivariate pattern analysis (MVPA), a popular technique for analyzing magneto- and electro-encephalography (M/EEG) neuroimaging data, quantifies the extent and time-course by which neural representations support the discrimination of relevant stimuli dimensions. As EEG is widely used for infant neuroimaging, time-resolved MVPA of infant EEG data is a particularly promising tool for infant cognitive neuroscience. MVPA has recently been applied to common infant imaging methods such as EEG and fNIRS. In this tutorial, we provide and describe code to implement time-resolved, within-subject MVPA with infant EEG data. An example implementation of time-resolved MVPA based on linear SVM classification is described, with accompanying code in Matlab and Python. Results from a test dataset indicated that in both infants and adults this method reliably produced above-chance accuracy for classifying stimuli images. Extensions of the classification analysis are presented including both geometric- and accuracy-based representational similarity analysis, implemented in Python. Common choices of implementation are presented and discussed. As the amount of artifact-free EEG data contributed by each participant is lower in studies of infants than in studies of children and adults, we also explore and discuss the impact of varying participant-level inclusion thresholds on resulting MVPA findings in these datasets.

A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence

Extensive sampling of neural activity during rich cognitive phenomena is critical for robust understanding of brain function. Here we present the Natural Scenes Dataset (NSD), in which high-resolution functional magnetic resonance imaging responses to tens of thousands of richly annotated natural scenes were measured while participants performed a continuous recognition task. To optimize data quality, we developed and applied novel estimation and denoising techniques. Simple visual inspections of the NSD data reveal clear representational transformations along the ventral visual pathway. Further exemplifying the inferential power of the dataset, we used NSD to build and train deep neural network models that predict brain activity more accurately than state-of-the-art models from computer vision. NSD also includes substantial resting-state and diffusion data, enabling network neuroscience perspectives to constrain and enhance models of perception and memory. Given its unprecedented scale, quality and breadth, NSD opens new avenues of inquiry in cognitive neuroscience and artificial intelligence.

Individualized cognitive neuroscience needs 7T: Comparing numerosity maps at 3T and 7T MRI

The field of cognitive neuroscience is weighing evidence about whether to move from the current standard field strength of 3 Tesla (3T) to ultra-high field (UHF) of 7T and above. The present study contributes to the evidence by comparing a computational cognitive neuroscience paradigm at 3T and 7T. The goal was to evaluate the practical effects, i.e. model predictive power, of field strength on a numerosity task using accessible pre-processing and analysis tools. Previously, using 7T functional magnetic resonance imaging and biologically-inspired analyses, i.e. population receptive field modelling, we discovered topographical organization of numerosity-selective neural populations in human parietal cortex. Here we show that these topographic maps are also detectable at 3T. However, averaging of many more functional runs was required at 3T to reliably reconstruct numerosity maps. On average, one 7T run had about four times the model predictive power of one 3T run. We believe that this amount of scanning would have made the initial discovery of the numerosity maps on 3T highly infeasible in practice. Therefore, we suggest that the higher signal-to-noise ratio and signal sensitivity of UHF MRI is necessary to build mechanistic models of the organization and function of our cognitive abilities in individual participants.

End-to-end neural system identification with neural information flow

Neural information flow (NIF) provides a novel approach for system identification in neuroscience. It models the neural computations in multiple brain regions and can be trained end-to-end via stochastic gradient descent from noninvasive data. NIF models represent neural information processing via a network of coupled tensors, each encoding the representation of the sensory input contained in a brain region. The elements of these tensors can be interpreted as cortical columns whose activity encodes the presence of a specific feature in a spatiotemporal location. Each tensor is coupled to the measured data specific to a brain region via low-rank observation models that can be decomposed into the spatial, temporal and feature receptive fields of a localized neuronal population. Both these observation models and the convolutional weights defining the information processing within regions are learned end-to-end by predicting the neural signal during sensory stimulation. We trained a NIF model on the activity of early visual areas using a large-scale fMRI dataset recorded in a single participant. We show that we can recover plausible visual representations and population receptive fields that are consistent with empirical findings.

Do intrinsic brain functional networks predict working memory from childhood to adulthood?

Working memory (WM) is defined as the ability to maintain a representation online to guide goal‐directed behavior. Its capacity in early childhood predicts academic achievements in late childhood and its deficits are found in various neurodevelopmental disorders. We employed resting‐state fMRI (rs‐fMRI) of 468 participants aged from 4 to 55 years and connectome‐based predictive modeling (CPM) to explore the potential predictive power of intrinsic functional networks to WM in preschoolers, early and late school‐age children, adolescents, and adults. We defined intrinsic functional networks among brain regions identified by activation likelihood estimation (ALE) meta‐analysis on existing WM functional studies (ALE‐based intrinsic functional networks) and intrinsic functional networks generated based on the whole brain (whole‐brain intrinsic functional networks). We employed the CPM on these networks to predict WM in each age group. The CPM using the ALE‐based and whole‐brain intrinsic functional networks predicted WM of individual adults, while the prediction power of the ALE‐based intrinsic functional networks was superior to that of the whole‐brain intrinsic functional networks. Nevertheless, the CPM using the whole‐brain but not the ALE‐based intrinsic functional networks predicted WM in adolescents. And, the CPM using neither the ALE‐based nor whole‐brain networks predicted WM in any of the children groups. Our findings showed the trend of the prediction power of the intrinsic functional networks to cognition in individuals from early childhood to adulthood.

Temporal dynamics of visual representations in the infant brain

Tools from computational neuroscience have facilitated the investigation of the neural correlates of mental representations. However, access to the representational content of neural activations early in life has remained limited. We asked whether patterns of neural activity elicited by complex visual stimuli (animals, human body) could be decoded from EEG data gathered from 12-15-month-old infants and adult controls. We assessed pairwise classification accuracy at each time-point after stimulus onset, for individual infants and adults. Classification accuracies rose above chance in both groups, within 500 ms. In contrast to adults, neural representations in infants were not linearly separable across visual domains. Representations were similar within, but not across, age groups. These findings suggest a developmental reorganization of visual representations between the second year of life and adulthood and provide a promising proof-of-concept for the feasibility of decoding EEG data within-subject to assess how the infant brain dynamically represents visual objects.

Investigating the Reliability of Population Receptive Field Size Estimates Using fMRI

In functional MRI (fMRI), population receptive field (pRF) models allow a quantitative description of the response as a function of the features of the stimuli that are relevant for each voxel. The most popular pRF model used in fMRI assumes a Gaussian shape in the features space (e.g., the visual field) reducing the description of the voxel’s pRF to the Gaussian mean (the pRF preferred feature) and standard deviation (the pRF size). The estimation of the pRF mean has been proven to be highly reliable. However, the estimate of the pRF size has been shown not to be consistent within and between subjects. While this issue has been noted experimentally, here we use an optimization theory perspective to describe how the inconsistency in estimating the pRF size is linked to an inherent property of the Gaussian pRF model. When fitting such models, the goodness of fit is less sensitive to variations in the pRF size than to variations in the pRF mean. We also show how the same issue can be considered from a bias-variance perspective. We compare different estimation procedures in terms of the reliability of their estimates using simulated and real fMRI data in the visual (using the Human Connectome Project database) and auditory domain. We show that, the reliability of the estimate of the pRF size can be improved considering a linear combination of those pRF models with similar goodness of fit or a permutation based approach. This increase in reliability of the pRF size estimate does not affect the reliability of the estimate of the pRF mean and the prediction accuracy.