Dear Mark,
Thanks for the reply!
However, at what level should I threshold the native space seeds for having minimum contamination. Should I set the threshold at the maximum or is a threshold of 0.5 sufficient? The problem is that a visual inspection of the seeds in native space to assess the level of contamination from nearby voxels is rather difficult, given the difference in voxel size between MNI and native space images. Our seeds are amygdala subregions, so rather small seeds.
Any assistance would be greatly appreciated.
All the best,
Mark Olivetti
Hi Mark,
In practice those two methods will be very similar. I personally prefer the first method as it gives you greater control over what to do about the edges. If you want to be more conservative, and make sure that there is very little contamination from voxels outside of your ROI, then you can use a more strict threshold to achieve this. The second option doesn't really give you that option so much, as the border voxels will have leakage from other voxels already mixed in during the interpolation. However, it is really only make much of a difference for small masks and where you are wanting to be quite careful about this sort of contamination.
All the best,
Mark
On 7 Jan 2014, at 13:16, Mark Olivetti <[log in to unmask]> wrote:
> Correction
>
> Thanks for the link!
>
> I was wondering, however, what the best strategy is for extracting time series in a seed-based analysis, in which the seeds are created in MNI space:
>
> 1. Registering the seeds to native space using FLIRT with applyxfm option and the registration matrix file standard2example_func.mat, generated by FEAT during preprocessing. Subsequently, thresholding the native space seeds for better spatial specificity/accuracy. Finally, extracting time series using the newly created native space seeds.
>
> 2. Using the example_func2standard.mat file for registering functional data to MNI space, using MNI space seeds to extract the time series in MNI space, and conduct connectivity analysis in native space using those time series.
>
> Any assistance would be greatly appreciated.
>
> Regards,
> Mark
|