Cortical depth-based variation of population receptive field size in human V1, V2 and V3

2552 

Abstract Submission 

Maya Jastrzębowska¹, Nicolás Gravel¹, Polina Iamschchinina^1,2, Daniel Haenelt³, Nikolaus Weiskopf^3,4, Radoslaw Cichy¹

¹Freie Universität Berlin, Berlin, Berlin, ²Humboldt-Universität zu Berlin, Berlin, Germany, ³Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, ⁴Leipzig University, Leipzig, Germany

Maya Jastrzębowska  
Freie Universität Berlin
Berlin, Berlin

Nicolás Gravel  
Freie Universität Berlin
Berlin, Berlin

Polina Iamschchinina  
Freie Universität Berlin|Humboldt-Universität zu Berlin
Berlin, Berlin|Berlin, Germany

Daniel Haenelt  
Max Planck Institute for Human Cognitive and Brain Sciences
Leipzig, Germany

Nikolaus Weiskopf  
Max Planck Institute for Human Cognitive and Brain Sciences|Leipzig University
Leipzig, Germany|Leipzig, Germany

Radoslaw Cichy  
Freie Universität Berlin
Berlin, Berlin

The early visual cortex is governed by the well-established organizing principles of retinotopy and cortical magnification. Receptive field (RF) size varies systematically at the macroscopic scale, increasing with eccentricity from fovea to periphery and along the visual hierarchy. However, the mesoscopic organization of RFs is not yet well understood. While neurophysiology studies in rodents and non-human primates have shown that RF size varies across cortical laminae, evidence in humans remains sparse (Fracasso et al., 2016). Here, we capitalize on recent advances in laminar fMRI to map RF properties in vivo at the submillimeter scale.

We used 7 Tesla fMRI (0.8 mm isotropic resolution) and population receptive field (pRF) mapping to measure gradient-echo blood oxygenation level-dependent (BOLD) responses to a drifting bar stimulus in four human participants. We projected the functional data to eight equivolumetric cortical surfaces based on white matter and pial surface reconstructions (Waehnert et al., 2014). For each vertex at each cortical depth, we fit a Difference of Gaussians (DoG) pRF model to the BOLD time series and estimated the location in visual space and pRF size that best explained visual field selectivity. We used the polar angle and eccentricity maps (Fig. 1A) to define regions of interest (ROIs): early visual areas V1, V2 and V3. Given the known variation of pRF size with eccentricity and across visual areas, we constrained our further analysis to an isoeccentricity band centered on 2 degrees. Analysis was limited to those vertices for which the goodness-of-fit of the pRF model exceeded the 75th percentile at the single participant level. For each participant and ROI, we computed the average center pRF and surround RF (sRF) of the DoG model. We characterized the p/sRF profiles across cortical depth by fitting a fourth order polynomial. Finally we computed a suppression index at each cortical depth, which represented the ratio of pRF surround suppression to center excitation (Sceniak et al., 2001).

Addressing the need for validation in laminar fMRI, we first confirm that pRF size varies linearly with eccentricity across ROIs and is smallest in V1, followed by V2 and V3, as expected (Fig. 1B). We replicate previous findings of a U-shaped relation between pRF size and cortical depth in V1 (Fig. 1C; Fracasso et al., 2016), with larger pRF sizes in deep and superficial layers as compared to middle layers. We extend these findings by demonstrating depth-dependent patterns in V2 and V3. Similarly to V1, pRF sizes in V2 are largest in deep layers, followed by superficial and middle layers. Surround RF sizes also vary across cortical depth, though they tend to increase towards the pial surface. Nevertheless, the suppression index remains flat, consistent with previous reports, indicating that the ratio of surround suppression to center excitation does not change across cortical depth.

Our findings demonstrate the reliability of submillimeter fMRI in identifying RF properties at the scale of cortical laminar circuits. PRF size variation across cortical depth is robustly quantifiable in humans in vivo using the latest laminar fMRI processing tools, despite numerous differences in data processing compared to those previously reported (Fracasso et al., 2016). Our findings confirm and extend previous studies and lend support to future studies examining feedforward and feedback mechanisms of spatial vision.

BOLD fMRI ²

Perception: Visual ¹

Cortical Layers

FUNCTIONAL MRI

HIGH FIELD MR

Modeling

NORMAL HUMAN

Vision

Other - Population receptive field modeling