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.

Lexical-Semantic Content, Not Syntactic Structure, Is the Main Contributor to ANN-Brain Similarity of fMRI Responses in the Language Network.

Representations from artificial neural network (ANN) language models have been shown to predict human brain activity in the language network. To understand what aspects of linguistic stimuli contribute to ANN-to-brain similarity, we used an fMRI data set of responses to n = 627 naturalistic English sentences (Pereira et al., 2018) and systematically manipulated the stimuli for which ANN representations were extracted. In particular, we (i) perturbed sentences' word order, (ii) removed different subsets of words, or (iii) replaced sentences with other sentences of varying semantic similarity. We found that the lexical-semantic content of the sentence (largely carried by content words) rather than the sentence's syntactic form (conveyed via word order or function words) is primarily responsible for the ANN-to-brain similarity. In follow-up analyses, we found that perturbation manipulations that adversely affect brain predictivity also lead to more divergent representations in the ANN's embedding space and decrease the ANN's ability to predict upcoming tokens in those stimuli. Further, results are robust as to whether the mapping model is trained on intact or perturbed stimuli and whether the ANN sentence representations are conditioned on the same linguistic context that humans saw. The critical result-that lexical-semantic content is the main contributor to the similarity between ANN representations and neural ones-aligns with the idea that the goal of the human language system is to extract meaning from linguistic strings. Finally, this work highlights the strength of systematic experimental manipulations for evaluating how close we are to accurate and generalizable models of the human language network.

Language in Brains, Minds, and Machines.

It has long been argued that only humans could produce and understand language. But now, for the first time, artificial language models (LMs) achieve this feat. Here we survey the new purchase LMs are providing on the question of how language is implemented in the brain. We discuss why, a priori, LMs might be expected to share similarities with the human language system. We then summarize evidence that LMs represent linguistic information similarly enough to humans to enable relatively accurate brain encoding and decoding during language processing. Finally, we examine which LM properties-their architecture, task performance, or training-are critical for capturing human neural responses to language and review studies using LMs as in silico model organisms for testing hypotheses about language. These ongoing investigations bring us closer to understanding the representations and processes that underlie our ability to comprehend sentences and express thoughts in language.

Driving and suppressing the human language network using large language models.

Transformer models such as GPT generate human-like language and are predictive of human brain responses to language. Here, using functional-MRI-measured brain responses to 1,000 diverse sentences, we first show that a GPT-based encoding model can predict the magnitude of the brain response associated with each sentence. We then 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 the 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 non-invasively control neural activity in higher-level cortical areas, such as the language network.

Many but not all deep neural network audio models capture brain responses and exhibit correspondence between model stages and brain regions.

Models that predict brain responses to stimuli provide one measure of understanding of a sensory system and have many potential applications in science and engineering. Deep artificial neural networks have emerged as the leading such predictive models of the visual system but are less explored in audition. Prior work provided examples of audio-trained neural networks that produced good predictions of auditory cortical fMRI responses and exhibited correspondence between model stages and brain regions, but left it unclear whether these results generalize to other neural network models and, thus, how to further improve models in this domain. We evaluated model-brain correspondence for publicly available audio neural network models along with in-house models trained on 4 different tasks. Most tested models outpredicted standard spectromporal filter-bank models of auditory cortex and exhibited systematic model-brain correspondence: Middle stages best predicted primary auditory cortex, while deep stages best predicted non-primary cortex. However, some state-of-the-art models produced substantially worse brain predictions. Models trained to recognize speech in background noise produced better brain predictions than models trained to recognize speech in quiet, potentially because hearing in noise imposes constraints on biological auditory representations. The training task influenced the prediction quality for specific cortical tuning properties, with best overall predictions resulting from models trained on multiple tasks. The results generally support the promise of deep neural networks as models of audition, though they also indicate that current models do not explain auditory cortical responses in their entirety.

Joint cortical registration of geometry and function using semi-supervised learning.

Brain surface-based image registration, an important component of brain image analysis, establishes spatial correspondence between cortical surfaces. Existing iterative and learning-based approaches focus on accurate registration of folding patterns of the cerebral cortex, and assume that geometry predicts function and thus functional areas will also be well aligned. However, structure/functional variability of anatomically corresponding areas across subjects has been widely reported. In this work, we introduce a learning-based cortical registration framework, JOSA, which jointly aligns folding patterns and functional maps while simultaneously learning an optimal atlas. We demonstrate that JOSA can substantially improve registration performance in both anatomical and functional domains over existing methods. By employing a semi-supervised training strategy, the proposed framework obviates the need for functional data during inference, enabling its use in broad neuroscientific domains where functional data may not be observed. The source code of JOSA will be released to the public at https://voxelmorph.net.

Lexical semantic content, not syntactic structure, is the main contributor to ANN-brain similarity of fMRI responses in the language network.

Representations from artificial neural network (ANN) language models have been shown to predict human brain activity in the language network. To understand what aspects of linguistic stimuli contribute to ANN-to-brain similarity, we used an fMRI dataset of responses to n=627 naturalistic English sentences (Pereira et al., 2018) and systematically manipulated the stimuli for which ANN representations were extracted. In particular, we i) perturbed sentences' word order, ii) removed different subsets of words, or iii) replaced sentences with other sentences of varying semantic similarity. We found that the lexical semantic content of the sentence (largely carried by content words) rather than the sentence's syntactic form (conveyed via word order or function words) is primarily responsible for the ANN-to-brain similarity. In follow-up analyses, we found that perturbation manipulations that adversely affect brain predictivity also lead to more divergent representations in the ANN's embedding space and decrease the ANN's ability to predict upcoming tokens in those stimuli. Further, results are robust to whether the mapping model is trained on intact or perturbed stimuli, and whether the ANN sentence representations are conditioned on the same linguistic context that humans saw. The critical result-that lexical-semantic content is the main contributor to the similarity between ANN representations and neural ones-aligns with the idea that the goal of the human language system is to extract meaning from linguistic strings. Finally, this work highlights the strength of systematic experimental manipulations for evaluating how close we are to accurate and generalizable models of the human language network.

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.

Probabilistic atlas for the language network based on precision fMRI data from >800 individuals.

Two analytic traditions characterize fMRI language research. One relies on averaging activations across individuals. This approach has limitations: because of inter-individual variability in the locations of language areas, any given voxel/vertex in a common brain space is part of the language network in some individuals but in others, may belong to a distinct network. An alternative approach relies on identifying language areas in each individual using a functional 'localizer'. Because of its greater sensitivity, functional resolution, and interpretability, functional localization is gaining popularity, but it is not always feasible, and cannot be applied retroactively to past studies. To bridge these disjoint approaches, we created a probabilistic functional atlas using fMRI data for an extensively validated language localizer in 806 individuals. This atlas enables estimating the probability that any given location in a common space belongs to the language network, and thus can help interpret group-level activation peaks and lesion locations, or select voxels/electrodes for analysis. More meaningful comparisons of findings across studies should increase robustness and replicability in language research.

Frontal language areas do not emerge in the absence of temporal language areas: A case study of an individual born without a left temporal lobe.

Language processing relies on a left-lateralized fronto-temporal brain network. How this network emerges ontogenetically remains debated. We asked whether frontal language areas emerge in the absence of temporal language areas through a 'deep-data' investigation of an individual (EG) born without her left temporal lobe. Using fMRI methods that have been validated to elicit reliable individual-level responses, we find that-as expected for early left-hemisphere damage-EG has a fully functional language network in her right hemisphere (comparable to the LH network in n = 145 controls) and intact linguistic abilities. However, we detect no response to language in EG's left frontal lobe (replicated across two sessions, 3 years apart). Another network-the multiple demand network-is robustly present in frontal lobes bilaterally, suggesting that EG's left frontal cortex can support non-linguistic cognition. The existence of temporal language areas therefore appears to be a prerequisite for the emergence of the frontal language areas.

The neural architecture of language: Integrative modeling converges on predictive processing.

The neuroscience of perception has recently been revolutionized with an integrative modeling approach in which computation, brain function, and behavior are linked across many datasets and many computational models. By revealing trends across models, this approach yields novel insights into cognitive and neural mechanisms in the target domain. We here present a systematic study taking this approach to higher-level cognition: human language processing, our species' signature cognitive skill. We find that the most powerful "transformer" models predict nearly 100% of explainable variance in neural responses to sentences and generalize across different datasets and imaging modalities (functional MRI and electrocorticography). Models' neural fits ("brain score") and fits to behavioral responses are both strongly correlated with model accuracy on the next-word prediction task (but not other language tasks). Model architecture appears to substantially contribute to neural fit. These results provide computationally explicit evidence that predictive processing fundamentally shapes the language comprehension mechanisms in the human brain.

Real-Time Decoding of Attentional States Using Closed-Loop EEG Neurofeedback.

Sustained attention is a cognitive ability to maintain task focus over extended periods of time (Mackworth, 1948; Chun, Golomb, & Turk-Browne, 2011). In this study, scalp electroencephalography (EEG) signals were processed in real time using a 32 dry-electrode system during a sustained visual attention task. An attention training paradigm was implemented, as designed in DeBettencourt, Cohen, Lee, Norman, and Turk-Browne (2015) in which the composition of a sequence of blended images is updated based on the participant's decoded attentional level to a primed image category. It was hypothesized that a single neurofeedback training session would improve sustained attention abilities. Twenty-two participants were trained on a single neurofeedback session with behavioral pretraining and posttraining sessions within three consecutive days. Half of the participants functioned as controls in a double-blinded design and received sham neurofeedback. During the neurofeedback session, attentional states to primed categories were decoded in real time and used to provide a continuous feedback signal customized to each participant in a closed-loop approach. We report a mean classifier decoding error rate of 34.3% (chance = 50%). Within the neurofeedback group, there was a greater level of task-relevant attentional information decoded in the participant's brain before making a correct behavioral response than before an incorrect response. This effect was not visible in the control group (interaction p=7.23e-4), which strongly indicates that we were able to achieve a meaningful measure of subjective attentional state in real time and control participants' behavior during the neurofeedback session. We do not provide conclusive evidence whether the single neurofeedback session per se provided lasting effects in sustained attention abilities. We developed a portable EEG neurofeedback system capable of decoding attentional states and predicting behavioral choices in the attention task at hand. The neurofeedback code framework is Python based and open source, and it allows users to actively engage in the development of neurofeedback tools for scientific and translational use.

Agonist-antagonist myoneural interface amputation preserves proprioceptive sensorimotor neurophysiology in lower limbs.

The brain undergoes marked changes in function and functional connectivity after limb amputation. The agonist-antagonist myoneural interface (AMI) amputation is a procedure that restores physiological agonist-antagonist muscle relationships responsible for proprioceptive sensory feedback to enable greater motor control. We compared results from the functional neuroimaging of individuals (n = 29) with AMI amputation, traditional amputation, and no amputation. Individuals with traditional amputation demonstrated a significant decrease in proprioceptive activity, measured by activation of Brodmann area 3a, whereas functional activation in individuals with AMIs was not significantly different from controls with no amputation (P < 0.05). The degree of proprioceptive activity in the brain strongly correlated with fascicle activity in the peripheral muscles and performance on motor tasks (P < 0.05), supporting the mechanistic basis of the AMI procedure. These results suggest that surgical techniques designed to restore proprioceptive peripheral neuromuscular constructs result in desirable central sensorimotor plasticity.

Single-Trial Decoding of Scalp EEG under Natural Conditions.

There is significant current interest in decoding mental states from electroencephalography (EEG) recordings. EEG signals are subject-specific, are sensitive to disturbances, and have a low signal-to-noise ratio, which has been mitigated by the use of laboratory-grade EEG acquisition equipment under highly controlled conditions. In the present study, we investigate single-trial decoding of natural, complex stimuli based on scalp EEG acquired with a portable, 32 dry-electrode sensor system in a typical office setting. We probe generalizability by a leave-one-subject-out cross-validation approach. We demonstrate that support vector machine (SVM) classifiers trained on a relatively small set of denoised (averaged) pseudotrials perform on par with classifiers trained on a large set of noisy single-trial samples. We propose a novel method for computing sensitivity maps of EEG-based SVM classifiers for visualization of EEG signatures exploited by the SVM classifiers. Moreover, we apply an NPAIRS resampling framework for estimation of map uncertainty, and thus show that effect sizes of sensitivity maps for classifiers trained on small samples of denoised data and large samples of noisy data are similar. Finally, we demonstrate that the average pseudotrial classifier can successfully predict the class of single trials from withheld subjects, which allows for fast classifier training, parameter optimization, and unbiased performance evaluation in machine learning approaches for brain decoding.

Comparison of Cultured Human Cardiomyocyte Clusters Obtained from Embryos/Fetuses or Derived from Human Embryonic Stem Cells.

Cardiomyocytes (CMs) derived from human embryonic stem cells (hESCs) or induced pluripotent stem cells (iPSCs) are used to study cardiogenesis and mechanisms of heart disease, and are being used in methods for toxiological screening of drugs. The phenotype of stem-cell-derived CMs should ideally resemble native CMs. Here, we compare embryonic/fetal CMs with hESC-derived CMs according to function and morphology. CM clusters were obtained from human embryonic/fetal hearts from elective terminated pregnancies before gestational week 12, and separated into atrial and ventricular tissues. Specific markers for embryonic CMs and primary cilia were visualized using immunofluorescence microscopy analysis. Contracting human embryonic cardiomyocyte (hECM) clusters morphologically and phenotypically resemble CMs in the embryonic/fetal heart. In addition, the contracting hECM clusters expressed primary cilia similar to that of cells in the embryonic/fetal heart. The electrophysiological characteristics of atrial and ventricular CMs were established by recording action potentials (APs) using sharp electrodes. In contrast to ventricular APs, atrial APs displayed a marked early repolarization followed by a plateau phase. hESC-CMs displayed a continuum of AP shapes. In all embryonic/fetal clusters, both atrial and ventricular, AP duration was prolonged by exposure to the KV11.1 channel inhibitor dofetilide (50 nM); however, the prolongation was not significant, possibly due to the relatively small number of experiments. This study provides novel information on APs and functional characteristics of atrial and ventricular CMs in first trimester hearts, and demonstrates that Kv11.1 channels play a functional role already at these early stages. These results provide information needed to validate methods being developed on the basis of in vitro-derived CMs from either hESC or iPSC, and although there was a good correlation between the morphology of the two types of CMs, differences in electrophysiological characteristics exist.