Re: [Marsbar-users] [SPM] error code in marsbar

[Matthew B. <mat...@gm...>]

Hi,

> i am sorry to place my question in the spm list, but the relating marsbar
> list did not respond. so maybe one of the matlab experts in spm could give
> me an idea what went wrong?

Sorry to be slow to reply, I've been away for a while.

> in my contrast (2 sample t-test) 4 significant cluster showed up. with help
> of marsbar i created 4 rois. i tried to get signal information from the raw
> data with these 4 rois (mat-files). with 3 of them it worked perfectly
> well, the remaining last file did not work at all. even after redoing every
> step, the same error occured. the error code says:
>
> MarsBaR analysis functions prepended to path
> Loaded MarsBaR defaults from base defaults
> Fetching data                           :                         1/1
> Warning: No valid data for roi 1 (F-contrast_-12_58_34)

This is almost invariably becaue you are sampling from SPM results
images, that have NaNs at the edges of the brain.  Marsbar uses linear
resampling by default to get the data from the images, so voxels at
the edge of the brain disappear due to resampling with NaN values.
The fix is to change the ROI resampling to nearest neighbour using
something like:

roi_filename = 'my_roi.mat';
my_roi = maroi(roi_filename);
my_roi = spm_hold(my_roi, 0); % set NN resampling
saveroi(my_roi, roi_filename)

and then rerun the data extraction...

Best,

Matthew