When to use mean luminance block

[sunflowercyh]

"Inserting a mean-luminance block provides an important baseline condition. Without this base-line these large differences in pRF size cannot be distinguished."(Dumoulin & Wandell,2008)"the insertion of mean-luminance blocks (base-line) is crucial to measure large pRF sizes. " My questions are 1) Does areas with the large size mainly refer to LO and TO? Then how about V4（I noticed @JWinawer’s paper "Mapping hV4 and ventral occipital cortex" also used bars with blanks, was it mainly because of VO?）, if we want to measure pRF size of V1-V4, can we use bars without mean luminance.

2) Insertion of mean-luminance block is good to measure the large size, but what's the cost of incomplete sweep of leaning bars?

Thanks!

[JWinawer]

On Aug 28, 2013, at 6:12 PM, sunflowercyh wrote:

"Inserting a mean-luminance block provides an important baseline condition. Without this base-line these large differences in pRF size cannot be distinguished."(Dumoulin & Wandell,2008)"the insertion of mean-luminance blocks (base-line) is crucial to measure large pRF sizes. " My questions are 1) Does areas with the large size mainly refer to LO and TO? Then how about V4（I noticed @JWinawer’s paper on “Mapping hV4 and ventral occipital cortex” also used bars with blanks, was it mainly because of VO?）, if we want to measure pRF size of V1-V4, can we use bars without mean luminance.

Yes, you could do this without mean luminance blocks. The main advantage of the mean luminance blocks is that it allows you to estimate the baseline. If the pRF is quite large and has some response to stimuli everywhere in the receptive field, then you would overestimate the baseline without mean luminance blocks. Neurons (and voxels) in V1-V4 do not have receptive fields that fill the whole visual field.

2) Insertion of mean-luminance block is good to measure the large size, but what's the cost of incomplete sweep of leaning bars?

You do not need any leaning bars to estimate the population receptive fields. I believe that in principle, sweeps in any two directions would do, as long as the two directions are not the same. Beyond that, more stimuli give you more samples. Including full diagonal sweeps rather than part blanks would give you more samples at certain positions (for better SNR), but a poorer estimate of the baseline. You could alter the stimulus so that it includes 8 full sweeps and blanks between the sweeps rather than replacing part of the sweeps.

Thanks!

— Reply to this email directly or view it on GitHub.

[sunflowercyh]

Jon, Thanks for detailed explaination! I think it’s really a good suggestion that "includes 8 full sweeps and blanks between the sweeps rather than replacing part of the sweeps." SNR of our data is not good enough, maybe the whole sweep will help. BTW currently i use SPM to smooth fMRI image, I wonder whether Vistasoft have a similar function to set FWHM and do smoothing?

[JWinawer]

There is a smoothing option in the GUI, probably under Analysis/time series. Smoothing in the gray view smoothed along the cortical surface. Note that the grid fit in the prf code operates on spatially and temporally smoothed data.

On Aug 29, 2013, at 7:24 AM, sunflowercyh notifications@github.com wrote:

Jon, Thanks for detailed explaination! I think it’s really a good suggestion that "includes 8 full sweeps and blanks between the sweeps rather than replacing part of the sweeps." SNR of our data is not good enough, maybe the whole sweep will help. BTW currently i use SPM to smooth fMRI image, I wonder whether Vistasoft have a similar function to set FWHM and do smoothing?

— Reply to this email directly or view it on GitHubhttps://github.com/vistalab/vistasoft/issues/85#issuecomment-23484976 .

[sunflowercyh]

Thanks for reminding~ and I find “detrend” is similar as smooth in SPM, because both of them provide high-pass filtering. One more question, We used four scans of bars, I wonder to avoid participants feel boring, if it is necessary and feasible to present bars in a randomly way, for example in some scans following bars sweep from left to right, we may present bars sweep from bottom to top(picture 2) rather than in a fixed pattern like picture 1.

picture 1

picture2

[JWinawer]

Sure, you can vary the stylus sequence. You won't be able to average your time series do you variance explained will probably go down, but the accuracy of the fits should not go down ( for example as measured by cross validation)

Sent from my iPhone

On Aug 30, 2013, at 11:21 AM, sunflowercyh notifications@github.com wrote:

Thanks for reminding~ and I find “detrend” is similar as smooth in SPM, because both of them provide high-pass filtering. One more question, We used four scans of bars, I wonder to avoid participants feel boring, if it is necessary to present bars in a randomly way, for example in some scans following bars sweep from left to right, we may present bars sweep from bottom to top(picture 2) rather than in a fix pattern like picture 1.

picture 1 [image: 2013-08-31_000708]https://f.cloud.github.com/assets/4111883/1058905/f34b608a-118f-11e3-9ad7-9da459d305a8.png

picture2 [image: 2]https://f.cloud.github.com/assets/4111883/1058910/04f83a10-1190-11e3-907f-94e335fc3e61.png

— Reply to this email directly or view it on GitHubhttps://github.com/vistalab/vistasoft/issues/85#issuecomment-23572648 .

[sunflowercyh]

“ Neighboring voxels with poor pRF model fits (variance explained  30%) were removed from this computation” when determine CMF for each voxel（Harvey &Dumoulin，2011). And we also need to compute CMF, so I will still follow the stylus sequence in case variance explained would go down. Jon, we really appreciate all the patient help and informative suggestions you gave us.