Feature-space selection with banded ridge regression.

Encoding models provide a powerful framework to identify the information represented in brain recordings. In this framework, a stimulus representation is expressed within a feature space and is used in a regularized linear regression to predict brain activity. To account for a potential complementarity of different feature spaces, a joint model is fit on multiple feature spaces simultaneously. To adapt regularization strength to each feature space, ridge regression is extended to banded ridge regression, which optimizes a different regularization hyperparameter per feature space. The present paper proposes a method to decompose over feature spaces the variance explained by a banded ridge regression model. It also describes how banded ridge regression performs a feature-space selection, effectively ignoring non-predictive and redundant feature spaces. This feature-space selection leads to better prediction accuracy and to better interpretability. Banded ridge regression is then mathematically linked to a number of other regression methods with similar feature-space selection mechanisms. Finally, several methods are proposed to address the computational challenge of fitting banded ridge regressions on large numbers of voxels and feature spaces. All implementations are released in an open-source Python package called Himalaya.

Neural correlates of blood flow measured by ultrasound.

Functional ultrasound imaging (fUSI) is an appealing method for measuring blood flow and thus infer brain activity, but it relies on the physiology of neurovascular coupling and requires extensive signal processing. To establish to what degree fUSI trial-by-trial signals reflect neural activity, we performed simultaneous fUSI and neural recordings with Neuropixels probes in awake mice. fUSI signals strongly correlated with the slow (<0.3 Hz) fluctuations in the local firing rate and were closely predicted by the smoothed firing rate of local neurons, particularly putative inhibitory neurons. The optimal smoothing filter had a width of ∼3 s, matched the hemodynamic response function of awake mice, was invariant across mice and stimulus conditions, and was similar in the cortex and hippocampus. fUSI signals also matched neural firing spatially: firing rates were as highly correlated across hemispheres as fUSI signals. Thus, blood flow measured by ultrasound bears a simple and accurate relationship to neuronal firing.

Visual and linguistic semantic representations are aligned at the border of human visual cortex.

Semantic information in the human brain is organized into multiple networks, but the fine-grain relationships between them are poorly understood. In this study, we compared semantic maps obtained from two functional magnetic resonance imaging experiments in the same participants: one that used silent movies as stimuli and another that used narrative stories. Movies evoked activity from a network of modality-specific, semantically selective areas in visual cortex. Stories evoked activity from another network of semantically selective areas immediately anterior to visual cortex. Remarkably, the pattern of semantic selectivity in these two distinct networks corresponded along the boundary of visual cortex: for visual categories represented posterior to the boundary, the same categories were represented linguistically on the anterior side. These results suggest that these two networks are smoothly joined to form one contiguous map.

The Representation of Semantic Information Across Human Cerebral Cortex During Listening Versus Reading Is Invariant to Stimulus Modality.

An integral part of human language is the capacity to extract meaning from spoken and written words, but the precise relationship between brain representations of information perceived by listening versus reading is unclear. Prior neuroimaging studies have shown that semantic information in spoken language is represented in multiple regions in the human cerebral cortex, while amodal semantic information appears to be represented in a few broad brain regions. However, previous studies were too insensitive to determine whether semantic representations were shared at a fine level of detail rather than merely at a coarse scale. We used fMRI to record brain activity in two separate experiments while participants listened to or read several hours of the same narrative stories, and then created voxelwise encoding models to characterize semantic selectivity in each voxel and in each individual participant. We find that semantic tuning during listening and reading are highly correlated in most semantically selective regions of cortex, and models estimated using one modality accurately predict voxel responses in the other modality. These results suggest that the representation of language semantics is independent of the sensory modality through which the semantic information is received.SIGNIFICANCE STATEMENT Humans can comprehend the meaning of words from both spoken and written language. It is therefore important to understand the relationship between the brain representations of spoken or written text. Here, we show that although the representation of semantic information in the human brain is quite complex, the semantic representations evoked by listening versus reading are almost identical. These results suggest that the representation of language semantics is independent of the sensory modality through which the semantic information is received.

Voxelwise encoding models with non-spherical multivariate normal priors.

Predictive models for neural or fMRI data are often fit using regression methods that employ priors on the model parameters. One widely used method is ridge regression, which employs a spherical multivariate normal prior that assumes equal and independent variance for all parameters. However, a spherical prior is not always optimal or appropriate. There are many cases where expert knowledge or hypotheses about the structure of the model parameters could be used to construct a better prior. In these cases, non-spherical multivariate normal priors can be employed using a generalized form of ridge known as Tikhonov regression. Yet Tikhonov regression is only rarely used in neuroscience. In this paper we discuss the theoretical basis for Tikhonov regression, demonstrate a computationally efficient method for its application, and show several examples of how Tikhonov regression can improve predictive models for fMRI data. We also show that many earlier studies have implicitly used Tikhonov regression by linearly transforming the regressors before performing ridge regression.

Seeing the world through non rose-colored glasses: anxiety and the amygdala response to blended expressions.

Anxious individuals have a greater tendency to categorize faces with ambiguous emotional expressions as fearful (Richards et al., 2002). These behavioral findings might reflect anxiety-related biases in stimulus representation within the human amygdala. Here, we used functional magnetic resonance imaging (fMRI) together with a continuous adaptation design to investigate the representation of faces from three expression continua (surprise-fear, sadness-fear, and surprise-sadness) within the amygdala and other brain regions implicated in face processing. Fifty-four healthy adult participants completed a face expression categorization task. Nineteen of these participants also viewed the same expressions presented using type 1 index 1 sequences while fMRI data were acquired. Behavioral analyses revealed an anxiety-related categorization bias in the surprise-fear continuum alone. Here, elevated anxiety was associated with a more rapid transition from surprise to fear responses as a function of percentage fear in the face presented, leading to increased fear categorizations for faces with a mid-way blend of surprise and fear. fMRI analyses revealed that high trait anxious participants also showed greater representational similarity, as indexed by greater adaptation of the Blood Oxygenation Level Dependent (BOLD) signal, between 50/50 surprise/fear expression blends and faces from the fear end of the surprise-fear continuum in both the right amygdala and right fusiform face area (FFA). No equivalent biases were observed for the other expression continua. These findings suggest that anxiety-related biases in the processing of expressions intermediate between surprise and fear may be linked to differential representation of these stimuli in the amygdala and FFA. The absence of anxiety-related biases for the sad-fear continuum might reflect intermediate expressions from the surprise-fear continuum being most ambiguous in threat-relevance.

Unraveling the anxious mind: anxiety, worry, and frontal engagement in sustained attention versus off-task processing.

Much remains unknown regarding the relationship between anxiety, worry, sustained attention, and frontal function. Here, we addressed this using a sustained attention task adapted for functional magnetic resonance imaging. Participants responded to presentation of simple stimuli, withholding responses to an infrequent "No Go" stimulus. Dorsolateral prefrontal cortex (DLPFC) activity to "Go" trials, and dorsal anterior cingulate (dACC) activity to "No Go" trials were associated with faster error-free performance; consistent with DLPFC and dACC facilitating proactive and reactive control, respectively. Trait anxiety was linked to reduced recruitment of these regions, slower error-free performance, and decreased frontal-thalamo-striatal connectivity. This indicates an association between trait anxiety and impoverished frontal control of attention, even when external distractors are absent. In task blocks where commission errors were made, greater DLPFC-precuneus and DLPFC-posterior cingulate connectivity were associated with both trait anxiety and worry, indicative of increased off-task thought. Notably, unlike trait anxiety, worry was not linked to reduced frontal-striatal-thalamo connectivity, impoverished frontal recruitment, or slowed responding during blocks without commission errors, contrary to accounts proposing a direct causal link between worry and impoverished attentional control. This leads us to propose a new model of the relationship between anxiety, worry and frontal engagement in attentional control versus off-task thought.

Moderate threat causes longer lasting disruption to processing in anxious individuals.

Anxiety is associated with increased attentional capture by threat. Previous studies have used simultaneous or briefly separated (<1 s) presentation of threat distractors and target stimuli. Here, we tested the hypothesis that high trait anxious participants would show a longer time window within which distractors cause disruption to subsequent task processing, and that this would particularly be observed for stimuli of moderate or ambiguous threat value. A novel temporally separated emotional distractor task was used. Face or house distractors were presented for 250 ms at short (∼1.6 s) or long (∼3 s) intervals prior to a letter string comprising Xs or Ns. Trait anxiety was associated with slowed identification of letter strings presented at long intervals after face distractors with part surprise/part fear expressions. In other words, these distractors had an impact on high anxious individuals' speed of target identification seconds after their offset. This was associated with increased activity in the fusiform gyrus and amygdala and reduced dorsal anterior cingulate recruitment. This pattern of activity may reflect impoverished recruitment of reactive control mechanisms to damp down stimulus-specific processing in subcortical and higher visual regions. These findings have implications for understanding how threat-related attentional biases in anxiety may lead to dysfunction in everyday settings where stimuli of moderate, potentially ambiguous, threat value such as those used here are fairly common, and where attentional disruption lasting several seconds may have a profound impact.

Fear-conditioning mechanisms associated with trait vulnerability to anxiety in humans.

Investigations of fear conditioning in rodents and humans have illuminated the neural mechanisms underlying cued and contextual fear. A critical question is how personality dimensions such as trait anxiety act through these mechanisms to confer vulnerability to anxiety disorders, and whether humans' ability to overcome acquired fears depends on regulatory skills not characterized in animal models. In a neuroimaging study of fear conditioning in humans, we found evidence for two independent dimensions of neurocognitive function associated with trait vulnerability to anxiety. The first entailed increased amygdala responsivity to phasic fear cues. The second involved impoverished ventral prefrontal cortical (vPFC) recruitment to downregulate both cued and contextual fear prior to omission (extinction) of the aversive unconditioned stimulus. These two dimensions may contribute to symptomatology differences across anxiety disorders; the amygdala mechanism affecting the development of phobic fear and the frontal mechanism influencing the maintenance of both specific fears and generalized anxiety.