general-discussion > fMRI masking problem (loss of brain regions)
Showing 1-2 of 2 posts
Results per page:
Mar 2, 2016  07:03 PM | Yashar Zeighami - McGill University
fMRI masking problem (loss of brain regions)

I am using for NIAK for a resting state dataset collected in young healthy subjects.
When using NIAK, we find a problem which I think is from making the fmri mask. The mask is smaller than the anatomical mask and loses hippocampi and the upper part of the cortex (motor/sensory areas).
I was wondering whether there is some parameters I can change to address this issue, or can we use the anatomical mask for fmri as well (is that a legitimate move?).
Thanks in advance.

Mar 3, 2016  04:03 AM | Pierre Bellec
RE: fMRI masking problem (loss of brain regions)
Dear Yashar,

Are you talking about individual or group masks? If individual, please check your functional volume. You may simply not cover these areas with the field of view. If group mask, there is a parameter in the preprocessing pipeline, which is not mentioned in the doc: 

opt.qc_coregister.thresh = 0.2;

To define a group mask, the average of all individual masks is first generated. The above threshold is then applied to get a binary group mask. In this example, any voxel which is in individual masks for 20% of subjects will be included in the group mask. This may seem really liberal (it's the default), but often times signal loss in ventral regions will create holes in the fMRI mask, although there is some signal to be analyzed here. FYI, old versions of niak had a much more stringent threshold. In the folder quality_control/group_coregister you have the average mask (with an hopefully sensible name). So you can load it and play around yourself with the threshold.

I hope this helps,