Face pRF Paper Draft

From VISTA LAB WIKI

Jump to: navigation, search

This page is for an outline of my paper on measuring population receptive fields using face stimuli. I started this page in June 2009. At this point, I have many fMRI data sets, and some very compelling results. The challenge is summarizing these results to provide a clear interpretation of the data. It is likely that I will seek to include some additional behavioral analyses, and possibly some follow-up fMRI measurements. But I believe the bulk of the existing data is enough for a substantial journal article.

This page is protected and is specific for coauthors on the paper. If you are a coauthor, please be bold and edit the draft as you see fit. If you are not but are part of the VISTA group, please feel free to read this and send me comments. Otherwise, you shouldn't be at this page. Sayres 10:53, 3 June 2009 (PDT)


Contents

[edit] Collaborating using the wiki

There are a few reasons why I'm interested in using the wiki to edit this paper. First, it's useful to have a single up-to-date version of the paper that's readily accessible from anywhere. The draft below should represent the current version. Second, it preserves a linear sequence of edits from each author, so we can see e.g. what I changed and what you changed. Using .doc or other files has the problem that it's ambiguous which file was created when; in the past we've had several forks in a manuscript, as we've edited different drafts at different times. The wiki format prevents forking in this manner. Third, the dicussion page seems like a nice way to keep track of the various comments and questions that may come up as we edit it, rather than paging through the paper and checking where the comments are.

To use the wiki : create an account and log in to this papers wiki. (This is separate from the VPNL wiki, in that it uses a newer version of the wiki software, which allows certain things like better referencing.) Select the 'edit' tab on top to change parts you don't like. Use the 'minor edit' checkbox if the edits are mostly spelling, grammatical, or something like that.

If you think a change needs justification, flip over to the 'discussion' tab and enter your comment on why you made the change. (Sign the comment, just so we can keep track of who said what). If I have a problem and want to change back, I'll add a response beneath it. Please don't add the comments to the main text; the main text should just be the text itself that we would want to send out. All our discussions should happen in the discussion tab.

Citing papers : to place a citation in the main text of the manuscript, place the following text:

 {{ref|Author##}} 

The part after the "|" divider is the name of the reference; I use the first author's last name and the last two digits of the publication year, for conciseness. (e.g., {{ref|Brewer05}}) . Then, at the bottom of this page in the 'references' section, place the text:

 # {{note|Author##}} authors (yr) title journal vol(num):pgs  

Where the "Author##" is the same name as in the "ref" flag. You can use multiple "ref" flags to point to the same note at the bottom.

I'm using APA style for the citations right now; when we finalize the journal to submit, I'll reformat all the citations to make 'em nice.

One annoyance is that the footnote numbers in the text reflect the occurrence of that footnote in the page, rather than the occurrence of the citation in the references list. When we're close to a final version of the draft, I will convert this text into a submittable document (probably via TeX, so .tex, .ps, and .pdf) and will make the footnotes consistent. For now, if you click on the footnote, it will take you to the proper citation in the list.



[edit] Population Receptive Fields in Object-selective Cortex

[edit] Authors List

Rory Sayres, Kalanit Grill-Spector, Jonathan Winawer, Kevin Weiner, and Brian Wandell

1 Psychology Department, Stanford University, Stanford CA 94305 2 Neurosciences Program, Stanford University, Stanford CA 94305

[edit] Acknowledgements

[edit] Abstract

The retinal position of stimuli plays a central role in the functional organization of early cortical processing regions. Recent evidence indicates that retinal position continues to play a strong role at higher stages of the processing pathway, including regions which are highly selective for particular stimulus classes such as faces (Template:Ref, Template:Ref, Template:Ref). The nature of retinal position sensitivity in these regions has substantial implications for how they may mediate object recognition, as well as the underlying principles of functional organization in cortex. Previous attempts to quantitatively map these regions were hindered in part because standard mapping stimuli, such as drifting gratings masked behind moving apertures, do not elicit strong responses. We examined the retinotopic organization of face-selective and nearby cortical regions on the ventral surface, using stimuli containing intact and Fourier phase-scrambled face images masked behind moving bar apertures. We analyzed the responses using a recently-developed population receptive field (pRF) model Template:Ref. We find that simple Gaussian receptive fields provide a compact and powerful description of position sensitivity across ventral cortex. The size and eccentricity of pRFs varies smoothly across cortex, while the polar angle is somewhat less consistent. When we use constant-size faces, we see a pronounced foveal bias near face-selective regions. However, when we scale and space out faces in the periphery to make them more visible relative to the fovea, receptive fields become larger and more peripheral, reducing the foveal bias. Despite these shifts, we observe an underlying functional organization, in which each hemisphere has two distinct foveal representations along the fusiform gyrus, which appear more lateral than the previously-described ventral occipital (VO) fovea (Template:Ref), with the foveas separated by a representation of more peripheral retinal positions. Face-selective regions are located near these foveal representations, although their relative locations vary substantially across subjects. The eccentricity and size of estimated pRFs are reliable across data sets, although the polar angle representation is somewhat less so, precluding the identification of visual field maps anterior to hV4. We interpret these findings as reflecting distinct neural sensitivities underlying the BOLD response: an underlying retinal sensitivity which varies along the cortical surface and is stimulus-insensitive, and a sensitivity to the salience of stimuli, which is affected by the size, position, and crowding of face stimuli.

[edit] Introduction

Retinal position is central to functional organization in visual cortex.

Retinal position also strongly affects responses in higher order visual regions.

Existing measurements and models of functional organization in face-selective cortex are qualitative, not quantitative.

Population receptive fields offer a precise quantitative model of retinotopy.

Previous retinotopic mapping experiments did not use the optimal stimuli for category-selective cortex.

Our experiments sought to quantitatively measure retinotopic organization along the ventral stream.

File:Figure1Mockup.png
Figure 1. Experimental Design and Regions of Interest. We mapped population receptive fields (pRF) in visual cortex using images of intact and Fourier phase-scrambled faces masked behind moving bar apertures. For all experiments, the movement of the bar apertures along the visual field was identical; only the stimulus behind the aperture changed between conditions. A Examples of mapping stimuli. We used phase-scrambled faces (i); a field of intact faces which we not scaled with respect to the periphery and closely spaces (ii); faces which were scaled (doubling in size) in the periphery and more widely spaced (iii); and single full-field face images (iv). B Illustration of the order of the moving bar stimulus. Four times every scan the bar aperture was turned off, leaving a mean-luminance screen for 12 seconds, to measure the response to no visual stimulation as well as different retinal stimulus positions. C Location of face-selective regions of interest (ROIs). The ventral gray-matter surface of both hemispheres is represented, zoomed in on the posterior portion of the brain, and inflated to show sulci. (Inset: uninflated image of whole brain, showing the zoomed region.) The overlay shows the contrast of (face images) > (other object categories), thresholded at p < 0.001, uncorrected. Face selective ROIs in the fusiform gyrus are outlined in blue.

[edit] Results

[edit] Population Receptive Fields (pRFs) can be mapped with face stimuli

File:Figure2Mockup.png
Figure 2. Retinotopically sensitive responses to intact but not scrambled face stimuli. Ventral surface of cortex, with the proportion of each voxel's time series variance explained by the pRF model shown thresholded at 15% or greater, for phase-scrambled face images (A), unscaled faces (B), scaled faces (C) and full-screen faces (D). Face-selective regions are outlined in black, and are not strongly driven by phase-scrambled stimuli, resulting in a low variance explained, but are strongly driven by the three intact-face conditions.

[edit] The estimated pRF is sensitive to the mapping stimulus

File:Figure3Mockup.png
Figure 3. Estimated pRFs vary with the mapping stimulus in face-selective cortex, but not V1. Shown are time courses from a single representative voxel in V1 (A) and the posterior face-selective Fus region (pFus) (B). For each subplot, the black trace indicates the measured response to the sweeping bar stimulus in that voxel, the blue curve represents the predicted response from the best-fitting pRF for that voxel and stimulus, and the best-fitting pRF is plotted as a function of visual field to the right of the time series. Row i indicates the responses to phase-scrambled face stimuli; row ii indicates the responses to unscaled faces; row iii indicates the responses to scaled faces; and row iv indicates the responses to full-screen faces. The measured time series, and best-fitting pRFs, are very similar for all mapping stimuli in V1 (A), but vary substantially for face-selective cortex (B).
File:Figure4Mockup.png
Figure 4. Maps of pRF parameters for the intact-face mapping stimuli.


[edit] Modeling the effectiveness of stimuli in the fovea versus the periphery

[edit] Boundary extension for masked full face images

[edit] Discussion

[edit] Methods

[edit] Subjects

[edit] MR Data Acquisition

[edit] Anatomical Data Analysis

[edit] fMRI Data Analysis

[edit] ROI Selection

[edit] Population Receptive Field (pRF) Model

[edit] Behavioral measurements

[edit] References

Template:Note This is a test.

Personal tools