Dissociable laminar profiles of concurrent bottom-up and top-down modulation in the human visual cortex

Abstract

Recent developments in human neuroimaging make it possible to non-invasively measure neural activity from different cortical layers. This can potentially reveal not only which brain areas are engaged by a task, but also how. Specifically, bottom-up and top-down responses are associated with distinct laminar profiles. Here, we measured lamina-resolved fMRI responses during a visual task designed to induce concurrent bottom-up and top-down modulations via orthogonal manipulations of stimulus contrast and feature-based attention. BOLD responses were modulated by both stimulus contrast (bottom-up) and by engaging feature-based attention (top-down). Crucially, these effects operated at different cortical depths: Bottom-up modulations were strongest in the middle cortical layer, while top-down modulations were strong at all depths, being significantly stronger in deep and superficial layers compared to bottom-up effects. As such, we demonstrate that laminar activity profiles can discriminate between concurrent top-down and bottom-up processing, and are diagnostic of how a brain region is activated.

Recurrent Processing Drives Perceptual Plasticity

“…Here, we demonstrate layer-specific, learning-dependent changes following several control analyses for these potential confounds, suggesting that our results are unlikely to be confounded by vasculature-related artifacts. Our results are consistent with previous laminar imaging studies showing BOLD effects in superficial layers in a range of tasks [ 28 , 29 , 46 , 47 ] and could not be simply attributed to differences in attention due to task difficulty, as participant performance was matched across sessions (pre- versus post-training). Future work could exploit recent advances in cerebral blood volume (CBV) imaging using vascular space occupancy (VASO) [ 52 ] to enhance the spatial specificity of laminar imaging in the human brain.…”

“…Here, we combined several approaches to reduce this superficial bias by removing voxels with low temporal signal-to-noise ratio and high t-statistic for stimulation contrast (see STAR Methods , correcting for vascular effects for details). We then Z scored each voxel’s time course to account for possible differences in signal strength and variance due to thermal or physiological noise across layers while preserving differences between conditions [ 29 ]. These corrections resulted in similar BOLD magnitude and multi-voxel pattern classification accuracy before training across layers ( Figure 3 E), suggesting that our approach for correcting vasculature-related effects controlled substantially for the superficial bias.…”

“…Neurophysiological studies have shown that this micro-circuit is involved in a range of visual recognition [ 26 ] and attention [ 27 ] tasks. Recent laminar fMRI studies provide evidence for the involvement of this circuit in the context of sensory processing [ 28 ] and visual attention [ 29 ].

Figure 1 Laminar Brain Circuits Schematic representation of the hypotheses tested; training may modify (1) feedforward processing (blue arrows) between LGN and V1, (2) recurrent processing via horizontal connections within the visual cortex (horizontal connections [indicated by orange arrows] between excitatory [open circles] and inhibitory [filled circles] neurons, and (3) feedback processing (green arrows) between V1 and higher areas (i.e., V2, V3, V4, and IPS) based on known anatomical circuits.

…”

LayNii: A software suite for layer-fMRI

“…The linear offset model has been explicitly discussed in the context of linear correlation analyses (Fracasso et al 2018) and in the context of layer-dependent shape parameters (Gau et al 2020). This linear model is furthermore implicitly used for in countless task contrast subtraction analyses of previous layer-fMRI studies. The scaling model (Guidi et al 2020; Kashyap et al 2018; Kazan et al 2017; Lawrence et al 2019) is based on the assumption that the layer-specific bias in GE-BOLD is driven by layer-dependent variations of vein density. As such, it is assumed that the superficial signal is larger than the signal in deeper layers, simply because the amount of venous blood volume is higher in superficial layers compared to deeper layers.…”

“…Thus, it is assumed that a simple multiplicative (or divisory) normalization can correct for the macrovascular signal bias. This layer-dependent signal normalization has been proposed as part of several layer-fMRI analysis strategies, including: scaling the layer-dependent signal with estimates of layer-dependent venous CBV (Guidi et al 2020; Kazan et al 2017; 2016), normalizing the signal difference between different task responses by the mean signal response of all involved task responses (Kashyap et al 2018), refraining to infer neuroscience conclusions solely based on task response differences in favor of focusing on task response ratios that are normalized by the presumably vascularly driven signal fluctuations (Lawrence et al 2019). The leakage model (Markuerkiaga et al 2016) is based on the assumption that the BOLD signal in each layer constitutes a signal mixture of activity originating in multiple layers.…”