Dear Zara,
That function was written for a particular purpose and indeed it works
better for continuous maps than for thresholded ones. I don't think
there is an easy fix so perhaps you should use the other way of
visualization. I'll keep it in mind for future development.
Best,
Vladimir
2012/2/1 Zara Bergström <[log in to unmask]>:
> Dear list,
>
> I've found some strange output when using spm_eeg_img2maps for plotting
> thresholded spm scalp-time EEG results, and wondered if anyone has a
> suggestion for how to fix this?
>
> I computed group-level spms on EEG data at the sensor level, and then saved
> the thresholded spm as an image. I then used spm_eeg_img2maps to plot this F
> image as a scalp topography across time.
>
> However, using img2maps, the resulting topographies are distorted and don't
> really match the raw data topography. If I use the "results" window to
> overlay my spm results on the mask.img, it plots the largest F value is the
> same location where the raw EEG difference is maximal. However, using
> img2maps to plot the same F image, the F topography is different, even
> though I use the same raw data file for the channel locations. The
> topographies are most distorted around the edges on the scalp, where
> essentially everything seems to be wiped out. I've attached a document with
> examples for comparison.
>
> I've gone through spm_eeg_img2maps, and I think the problem stems from the
> way this function transforms the image pixel matrix into an interpolated
> scalp map. It seems to downsample the matrix into a channel vector by
> assigning the nearest pixel value to a channel (on line 93), and then uses
> spm_eeg_plotScalpData to interpolate those channels to create a scalp map.
>
> However, because I have thresholded the spm map to only include F values
> above a certain size and these are only sampled by 70 channels (quite
> typical channel number for EEG), the resulting interpolation doesn't work
> very well and the topography is really patchy and weirdly shaped, and some
> large clusters around the edges are almost absent because they were not
> directly under an electrode location. In fact, it looks like the electrodes
> around the edges are actually masked out of the plot.
>
> Do you think it might be possible to improve plotting with img2maps somehow,
> for example by increasing the number of pixels that are used for
> interpolation, or by widening the mask to include electrodes on the edges?
>
> Any input would be very much appreciated!
>
> Thanks for your time,
>
> Zara
|