Population receptive fields in nonhuman primates from whole-brain fMRI and large-scale neurophysiology in visual cortex

  1. P Christiaan Klink  Is a corresponding author
  2. Xing Chen
  3. Wim Vanduffel
  4. Pieter R Roelfsema
  1. Netherlands Institute for Neuroscience, Royal Netherlands Academy of Arts and Sciences, Netherlands
  2. Psychiatry Department, Amsterdam UMC, Netherlands
  3. Laboratory for Neuro- and Psychophysiology, Department of Neurosciences, KU Leuven Medical School, Belgium
  4. Massachusetts General Hospital, Martinos Ctr. for Biomedical Imaging, United States
  5. Leuven Brain Institute, KU Leuven, Belgium
  6. Harvard Medical School, United States
  7. Department of Integrative Neurophysiology, Center for Neurogenomics and Cognitive Research, VU University, Netherlands
12 figures, 1 table and 2 additional files

Figures

Figure 1 with 1 supplement
Experimental setup and study design.

(A) Monkeys maintained fixation on a red dot while bars with high-contrast-moving checkerboards moved across the screen in eight different directions behind a virtual aperture (dashed line, not …

Figure 1—figure supplement 1
Comparison of hemodynamic response functions (HRFs).

(A) Population receptive field (pRF) models were fit to the blood-oxygen-level-dependent (BOLD) response with a monkey-specific (see Materials and methods) and a canonical HRF provided by the …

Figure 2 with 1 supplement
Population receptive field (pRF) model fits and retinotopic maps.

(A) R2 value map of the compressive spatial summation (CSS) pRF model projected on the surface rendering of the brains of two monkeys (M1, M2). The lower panel illustrates that the R2 value in the …

Figure 2—figure supplement 1
Proportion of voxels with R2 > 5% per region of interest (ROI).

For both animals (M1, M2) and all four population receptive field (pRF) models. This is supplement to Figure 2C that reports absolute numbers of voxels per area.

Figure 3 with 1 supplement
Retinotopy in the thalamus.

Thalamic population receptive fields (pRFs) in M1 (A) and M2 (B). The lateral geniculate nucleus (LGN, top rows) contained retinotopic maps of the contralateral visual field in both monkeys (M1: …

Figure 3—figure supplement 1
Striatal population receptive fields (pRFs).

(A) In M2, the head of the caudate nucleus in the striatum contained retinotopic maps of the lower contralateral visual field. Neurons in the head of the caudate have long been known to play a role …

Figure 4 with 2 supplements
Comparison of the four population receptive field (pRF) models.

(A) Comparison across pRF models. R2 data are in bins of 1% × 1%, and color indicates the number of voxels per bin. The compressive spatial summation (CSS) model fits the data best, while U-LIN and …

Figure 4—figure supplement 1
Comparison of fit performance across pRF models.

(A, B) Fit accuracy advantage of compressive spatial summation (CSS) and difference-of-Gaussians (DoG) models across brain areas. Both the CSS (A) and DoG (B) models had better fits (cross-validated …

Figure 4—figure supplement 2
Location of negative pRFs.

(A) Normalized amplitude of the suppressive surround Gaussian of difference-of-Gaussians (DoG) model fits (R2 > 5%). Values larger than 1 (blue tints) indicate that the amplitude of suppressive …

Figure 5 with 2 supplements
Population receptive field (pRF) size as a function of eccentricity according to the compressive spatial summation (CSS) model.

(A) Eccentricity-size relationship for early and mid-level visual areas. (B) Eccentricity-size relationship for areas in the temporal and parietal lobes. (C) Eccentricity-size relationship for …

Figure 5—figure supplement 1
Eccentricity-size relationship for all regions of interest (ROIs).

Linear fits (intercept and slope) to the eccentricity-size relationship per brain area. Shaded areas indicate the 95% confidence interval of the fit, n denotes the number of voxels (R2 > 5%). Linear …

Figure 5—figure supplement 2
Eccentricity-size relationship for all regions of interest (ROIs).

Linear fits (intercept and slope) to the eccentricity-size relationship per brain area. Shaded areas indicate the 95% confidence interval of the fit, n denotes the number of voxels (R2 > 3%). Linear …

Figure 6 with 1 supplement
Visual field coverage of population receptive fields (pRFs) with Utah arrays.

(A) In both monkeys (M3, M4), 14 Utah arrays were implanted on the left operculum that is partly V1. Different colors represent the center of multi-unit activity (MUA)-based pRFs for the individual …

Figure 6—figure supplement 1
Heatmaps of visual field coverage of the Utah arrays.

We reconstructed the multi-unit activity (MUA) population receptive fields (pRFs) with R2 > 50% in the compressive spatial summation (CSS) model in the visual field, normalized them to their peak …

Comparison of multi-unit activity (MUA) population receptive field (pRF) sizes with conventionally determined RF (cRF) sizes (moving bar stimulus).

Data points represent recording sites of individual animals (black: M3; blue: M4) and brain areas (closed circles: V1; open circles: V4). (A) pRF sizes estimated with the P-LIN model (X-axis in left …

Figure 8 with 3 supplements
Comparison of multi-unit activity (MUA)-based fit results from the four population receptive field (pRF) models.

Scatterplots compare R2 of pRF models. Each dot represents an electrode (black: V1; green: V4).

Figure 8—figure supplement 1
Comparison of local field potential (LFP) fits for the four population receptive field (pRF) models in V1.

Scatterplots compare R2 across pRF models and LFP frequency bands. Each dot represents an electrode. For a subset of electrodes, the pRF models that can capture negative responses …

Figure 8—figure supplement 2
Comparison of local field potential (LFP)-based fit results from the four population receptive field (pRF) models in V4.

Scatterplots comparing R2 across pRF models and LFP frequency bands. Each data point represents an electrode.

Figure 8—figure supplement 3
Comparison of population receptive field (pRF) fit accuracies for multi-unit activity (MUA) and local field potential (LFP) signals at the same recording sites.

(A) Comparison of R2 values from the compressive spatial summation (CSS) model across electrophysiological signals. Colors indicate the number of recording sites in 4 × 4% bins (logarithmic scale). …

V1 arrays with outlying population receptive field (pRF) sizes.

(A) Schematic representation of the location of the craniotomy made during surgery (dashed line) and the implanted electrode arrays (rectangles) depicted on the NMT standard brain. The color map …

Figure 10 with 2 supplements
Characteristics of local field potential (LFP)-α population receptive fields (pRFs) in V1 split by positive and negative gain values.

(A) Distribution of gain values for LFP-α pRFs of V1 electrodes estimated with the U-LIN model. Electrodes with positive gain pRFs are classified as α+, electrodes with negative gain pRFs as α-. (B, …

Figure 10—figure supplement 1
Characteristics of local field potential (LFP)-β population receptive fields (pRFs) in V1 split by positive and negative gain values.

(A) Distribution of gain values for LFP-β pRFs of V1 electrodes estimated with the U-LIN model. Recording sites with positive gain pRFs are classified as β+, sites with negative gain pRFs as β-. (B, …

Figure 10—figure supplement 2
Negative population receptive fields (pRFs) based on low-frequency local field potential (LFP) components are shifted toward the fixation point compared to the positive pRFs based on the high-frequency LFP at the same recording site.

(A) Relative locations of pRFs derived from LFP-α (yellow) and LFP-γl (blue) of the same recording sites. (B) Relative locations of pRFs derived from LFP-β (green) and LFP-γl (blue) of the same …

Comparison of population receptive field (pRF) location and size of different electrophysiological signals at the same electrode, estimated by the compressive spatial summation (CSS) model.

(A) Median distance between receptive field (RF) estimates. Electrodes were only included if R2 > 25% (multi-unit activity [MUA], local field potential [LFP]) or signal-to-noise ratio (SNR) > 3 …

Figure 12 with 1 supplement
Eccentricity-size relationship for population receptive fields (pRFs) across signal types.

(A) The pRF size-eccentricity relation for V1 electrodes (compressive spatial summation [CSS] model, R2 > 50%). Dots are individual electrodes, colored lines represent the slope of the …

Figure 12—figure supplement 1
Cross-signal comparisons of population receptive field (pRF) eccentricity-size relationship.

(A) The results of the cross-signal comparisons of compressive spatial summation (CSS)-pRF-based eccentricity-size slopes for different data inclusion criteria. For the top row, we only included V1 …

Tables

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Biological sample(Macaca mulatta)Rhesus macaque (Macaca mulatta), maleBiomedical Primate
Research Center,
the Netherlands
n/a-
OtherPhilips Ingenia 3.0T MR systemPhilipsn/aAt Spinoza Centre for Neuroimaging, Amsterdam, the Netherlands
Other8-channel phased array receive MR coil systemKU Leuvenn/aCustom-built
Other16-channel MR pre-amplifierMR Coils BVn/aCustom-built
OtherETL-200ISCANRRID:SCR_021044MR-compatible eye tracker
OtherE3X-NHOmronn/aFiber optic amplifiers
Other5-RLD-E1 Liquid Reward SystemCrist Instrument
Company, Inc
n/aJuice reward system
OtherBOLDscreen 32 LCD for fMRICambridge
Research
Systems
n/aMR-compatible display
OtherUtah array (electrodes)Blackrock
Microsystems
n/a-
Other128-channel CerePlex M head-stagesBlackrock
Microsystems
n/aData acquisition
Other128-channel CerePlex M head-stagesBlackrock
Microsystems
n/aData acquisition
Other128-channel Digital HubBlackrock
Microsystems
n/aData acquisition
Other128-channel Neural Signal Processor (NSP)Blackrock
Microsystems
n/aData acquisition
Software, algorithmBlackrock Central Software SuiteBlackrock
Microsystems
n/a-
OtherET-49CTomas
Recording
n/aEye tracker
Software, algorithmMATLABMathWorksRRID:SCR_001622-
OtherLISA clusterSURFsaran/aComputing cluster
Software, algorithmdcm2niixhttps://github.com/rordenlab/dcm2niixRRID:SCR_01409-
Software, algorithmNipypehttp://nipy.org/nipype/RRID:SCR_002502Used as the basis of the custom NHP-BIDS pipeline
Software, algorithmNHP-BIDSNetherlands
Institute for
Neuroscience
RRID:SCR_021813In-house developed, available via: https://github.com/VisionandCognition/NHP-BIDS
Software, algorithmFreeSurferhttp://surfer.nmr.mgh.harvard.edu/RRID:SCR_001847Used as the basis of the custom NHP-Freesurfer
Software, algorithmNHP-FreesurferNetherlands
Institute for
Neuroscience
RRID:SCR_021814In-house developed, available via: https://github.com/VisionandCognition/NHP-Freesurfer
Software, algorithmPycortexhttps://gallantlab.github.io/pycortex/n/aUsed as the basis of the customized NHP-Pycortex
Software, algorithmNHP-PycortexNetherlands
Institute for
Neuroscience
RRID:SCR_021815In-house developed, available via: https://github.com/VisionandCognition/NHP-pycortex
Software, algorithmanalyzePRFhttps://kendrickkay.net/analyzePRF/n/aToolbox was edited for this study and made available with the code and data
Software, algorithmJupyter Notebookhttps://jupyter.org/RRID:SCR_018315-
Software, algorithmFSLhttp://www.fmrib.ox.ac.uk/fsl/RRID:SCR_002823-
Software, algorithmTracker-MRI: Experiment control softwareNetherlands
Institute for
Neuroscience
RRID:SCR_021816In-house developed, available via: https://github.com/VisionandCognition/Tracker-MRI
OtherNMT v1.3NIH, AFNI
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/nonhuman/macaque/template_nmtv1.html#nmt-v1-3
n/aMacaque Brain Template and Atlas

Additional files

Supplementary file 1

Table with region of interest (ROI) abbreviations.

List of ROI abbreviations, color coded by where in the brain they are located.

https://cdn.elifesciences.org/articles/67304/elife-67304-supp1-v3.docx
Transparent reporting form
https://cdn.elifesciences.org/articles/67304/elife-67304-transrepform1-v3.pdf

Download links