Denis Chaimow

Multivariate pattern analysis (MVPA) methods are a versatile tool to retrieve information from neurophysiological data obtained with functional magnetic resonance imaging (fMRI) techniques. Since fMRI is based on measuring the hemodynamic response following neural activation, the spatial specificity of the fMRI signal is inherently limited by contributions of macrovascular compartments that drain the signal from the actual location of neural activation, making it challenging to image cortical structures at the spatial scale of cortical columns and layers. By relying on information from multiple voxels, MVPA has shown promising results in retrieving information encoded in fine-grained spatial patterns. We examined the spatial specificity of the signal exploited by MVPA. Over multiple sessions, we measured ocular dominance columns (ODCs) in human primary visual cortex (V1) with different acquisition techniques at 7 T. For measurements with blood oxygenation level dependent (BOLD) contrast, we included both gradient echo- (GE-BOLD) and spin echo-based (SE-BOLD) sequences. Furthermore, we acquired data using the vascular-space-occupancy (VASO) fMRI technique, which is sensitive to cerebral blood volume (CBV) changes. We used the data to decode eye-of-origin from signals across cortical layers. While ocularity information can be decoded with all imaging techniques, laminar profiles reveal that macrovascular contributions affect all acquisition methods, limiting their specificity across cortical depth. Therefore, although MVPA is a promising approach for investigating the mesoscopic circuitry of the human cerebral cortex, careful consideration of macrovascular contributions is needed that render the spatial specificity of the extracted signal.

Laminar fMRI using GE-BOLD is vulnerable to spatial blurring from intracortical veins, while multivariate pattern analysis (MVPA) is often assumed to mitigate these biases. Yet, this assumption has not been systematically investigated. We thus developed a mechanistic laminar response model that simulates voxel-wise patterns across cortical depths, incorporating a vascular draining model. We conducted simulations in which the ground-truth signal originated in a single, several, or across all layers, and applied standard MVPA decoding before and after deconvolution of the draining effect. Decoding accuracies were consistently influenced by draining veins: deep-origin signals yielded above-chance decoding in superficial layers, and null scenarios produced false positives in middle or deep layers. Vascular deconvolution enhanced specificity in single-layer cases but did not resolve ambiguities in null decoding profiles. Simulating six thinner layers improved decoding accuracies, especially in the deconvolved signal scenarios. These findings demonstrate that multivariate techniques are not inherently immune to vascular biases, but also demonstrate that careful modeling can help correct draining effects.

Functional magnetic resonance imaging (fMRI) at high magnetic field strengths (≥ 7 T) is a promising technique to study the functioning of the human brain at the spatial scale of cortical columns and layers. However, measurements most often rely on the blood oxygenation level dependent (BOLD) response sampled with a gradient echo (GE) sequence, which is known to be most sensitive to macrovascular contributions that limit their effective spatial resolution. Alternatively, a spin echo (SE) sequence can be used to increase the weighting toward the microvasculature and, therefore, the location of neural activation. In addition, due to the heterogeneous structure of the cortical cerebrovascular system, the effective spatial resolution can change across cortical depth. For high-resolution fMRI applications, it is hence important to know how much the effective spatial resolution varies across cortical depth. In this study, we used flickering rotating wedge stimuli to induce traveling waves with varying spatial frequencies in the retinotopically organized primary visual cortex (V1), which allowed us to infer the modulation transfer function (MTF) of the BOLD response that characterizes the spatial specificity of the measured signal. We acquired GE- and SE-BOLD data at 7 T and compared the MTF between acquisition techniques at different cortical depths. Our results show a small but consistent increase in spatial specificity when using SE-BOLD. But across cortical depth, both acquisition techniques generally show a similar decrease of specificity toward the pial surface demonstrating the dependence on macrovascular contributions, which needs to be carefully considered when interpreting the results of high-resolution fMRI studies.

Although working memory reliably activates the dorsolateral prefrontal cortex (dlPFC), the functional contributions of its cortical layers in humans remain unclear. A seminal study reported a laminar dissociation in dlPFC—superficial layers engaged during working-memory manipulation and deeper layers during motor response execution. Like many current layer-resolved studies, that work relied on several manual and semi-manual processing steps, including the selection of regions of interest. We conducted a preregistered replication in 21 human participants using an automated and fully reproducible analysis pipeline. While we replicated the superficial-layer effect during manipulation, we found no evidence for preferential deep-layer activation during response execution. This biologically plausible result refines current models of laminar organization in the human prefrontal cortex and aligns human evidence with the more heterogeneous animal literature. By demonstrating reproducibility through preregistration and automation, this work establishes a benchmark for laminar analyses in cognitive neuroscience.

Technological advances in fMRI including ultra-high magnetic fields (≥ 7 T) and acquisition methods that increase spatial specificity have paved the way for studies of the human cortex at the scale of layers and columns. This mesoscopic scale promises an improved mechanistic understanding of human cortical function so far only accessible to invasive animal neurophysiology. In recent years, an increasing number of studies have applied such methods to better understand the cortical function in perception and cognition. This future perspective article asks whether closed-loop fMRI studies could equally benefit from these methods to achieve layer and columnar specificity. We outline potential applications and discuss the conceptual and concrete challenges, including data acquisition and volitional control of mesoscopic brain activity. We anticipate an important role of fMRI with mesoscopic resolution for closed-loop fMRI and neurofeedback, yielding new insights into brain function and potentially clinical applications. This article is part of the theme issue ‘Neurofeedback: new territories and neurocognitive mechanisms of endogenous neuromodulation’.

: The dorsolateral prefrontal cortex (dlPFC) is reliably engaged in working memory (WM) and comprises different cytoarchitectonic layers, yet their functional role in human WM is unclear. Here, participants completed a delayed-match-to-sample task while undergoing functional magnetic resonance imaging (fMRI) at ultra-high resolution. We examine layer-specific activity to manipulations in WM load and motor response. Superficial layers exhibit a preferential response to WM load during the delay and retrieval periods of a WM task, indicating a lamina-specific activation of the frontoparietal network. Multivariate patterns encoding WM load in the superficial layer dynamically change across the three periods of the task. Last, superficial and deep layers are non-differentially involved in the motor response, challenging earlier findings of a preferential deep layer activation. Taken together, our results provide new insights into the functional laminar circuitry of the dlPFC during WM and support a dynamic account of dlPFC coding.

Summary: The extraordinary number of short association fibres (SAF) connecting neighbouring cortical areas is a prominent feature of the large gyrified human brain. The contribution of SAF to the human connectome is largely unknown because of methodological challenges in mapping them. We present a method to characterise cortico–cortical connectivity mediated by SAF in topologically organised cortical areas. We introduce the ‘structural connective fields’ (sCF) metric which specifically quantifies neuronal signal propagation and integration mediated by SAF. This new metric complements functional connective field metrics integrating across contributions from short- and long-range white matter and intracortical fibres. Applying the method in the human early visual processing stream, we show that SAF preserve cortical functional topology. Retinotopic maps of V2 and V3 could be predicted from retinotopy in V1 and SAF connectivity. The sCF sizes increased along the cortical hierarchy and were smaller than their functional counterparts, in line with the latter being additionally broadened by long-range and intracortical connections. In vivo sCF mapping provides insights into short-range cortico– cortical connectivity in humans comparable to tract tracing studies in animal research and is an essential step towards creating a complete human connectome. Highlights: Non-invasive mapping of Short Association Fibre (SAF) connectivity via diffusion-weighted MRI-based probabilistic tractography accurately predicted cortical functional neuroanatomy. The novel structural Connective Fields (sCF) concept provides a quantitative measure of cortico-cortical integration facilitated by SAF, complementing the existing functional Connective Field (CF) concept. Sub-millimeter resolution diffusion-weighted MRI enables tractography and connective field modeling of SAF, unlocking applications previously restricted to invasive tract tracing in animal studies.

The dorsolateral prefrontal cortex (dlPFC) is reliably engaged in working memory (WM). Evidence from non-human primates indicates that the dlPFC comprises different cytoarchitectonic layers that play distinct roles in WM subprocesses; yet the functional role of the dlPFC’s laminar circuitry in human WM is not well understood. In this study, participants completed a delayed-match-to-sample WM task while undergoing functional magnetic resonance imaging (fMRI) at ultra-high resolution, which allowed us to examine layer-specific responses of the dlPFC to manipulations in WM load and motor response. We conducted univariate and multivariate analyses across all periods of the WM task: encoding, delay and retrieval. First, we observed that superficial layers activate stronger than deep layers to higher WM load during the delay period. This aligns with earlier work showing preferential superficial layer activation to WM manipulation and as such may indicate lamina-specific activation of the frontoparietal network to heightened task demands more generally. Second, we found that superficial layers show higher decoding of WM load differences than deep layers during the retrieval period. In this context, we could show that decoding of WM load in the superficial layer exhibited dynamic changes across the encoding, delay and retrieval period of the task, indicative of separate WM control processes that occur on the WM content. Last, we found that superficial and deep layers are both non-differentially involved in the motor response, contradicting earlier findings of a preferential deep layer activation in humans. Taken together, our results provide new insights into the functional laminar circuitry of the dlPFC during WM and provide further support for a dynamic account of dlPFC coding.

Multivariate pattern analysis (MVPA) methods are a versatile tool to retrieve information from neurophysiological data obtained with functional magnetic resonance imaging (fMRI) techniques. Since fMRI is based on measuring the hemodynamic response following neural activation, the spatial specificity of the fMRI signal is inherently limited by contributions of macrovascular compartments that drain the signal from the actual location of neural activation, making it challenging to image cortical structures at the spatial scale of cortical columns and layers. By relying on information from multiple voxels, MVPA has shown promising results in retrieving information encoded in fine-grained spatial patterns. We examined the spatial specificity of the signal exploited by MVPA. Over multiple sessions, we measured ocular dominance columns (ODCs) in human primary visual cortex (V1) with different acquisition techniques at 7 T. For measurements with blood oxygenation level dependent (BOLD) contrast, we included both gradient echo- (GE-BOLD) and spin echo-based (SE-BOLD) sequences. Furthermore, we acquired data using the vascular-space-occupancy (VASO) fMRI technique, which is sensitive to cerebral blood volume (CBV) changes. We used the data to decode eye-of-origin from signals across cortical layers. While ocularity information can be decoded with all imaging techniques, laminar profiles reveal that macrovascular contributions affect all acquisition methods, limiting their specificity across cortical depth. Therefore, although MVPA is a promising approach for investigating the mesoscopic circuitry of the human cerebral cortex, careful consideration of macrovascular contributions is needed that render the spatial specificity of the extracted signal.

Visual imagery and veridical perception are phenomenologically related, suggesting that they share neural machinery. Previous research confirmed this assumption at the macroscale of human cortical organization in category-selective regions in ventral visual cortex. However, imagery and perception differ fundamentally in the information flow that underlies them: perception is driven by the integration of sensory feedforward and internally-generated feedback information, whereas imagery depends on feedback only. This suggests that while visual imagery and perception may activate common cortical regions, they do so in fundamentally distinctive ways. To investigate, we resolved neural activity during imagery and perception in the ventral visual cortex at the level of laminar organization that anatomically and functionally distinguishes feedforward from feedback information flow. We found distinctive laminar profiles for imagery and perception of scenes and faces in parahippocampal place area (PPA) and fusiform face area (FFA) respectively. Our findings clarify the neural basis of the phenomenology of visual imagery versus perception, and shed new light on how feedforward and feedback information processing in high-level ventral visual cortex orchestrates human object vision.