i need to calculate SNR on specific ROis overlaid over unprocessed dmri data which are 4-d. Is it correct to assume noise distribution across all slices is uniform and thus interpolate through slices in order to apply SNR calculation and ROIs over 3-d images?
thank you,
Rosella
Hi Rosella,
The noise distribution is typically not perfectly uniform across the brain (not just across slices but in 3D), due to a considerable number of factors. The dwidenoise command has a -noise option that will provide a spatial map of the noise level that was estimated in each voxel.
I don’t quite follow some of the specifics of your question though. By definition, if you are trying to calculate SNR within a specific ROI, what you are going to obtain is some kind of aggregate from the volume of that ROI, regardless of how variable the noise is within that ROI, or whether it involves multiple slices. But that’s just a question of whether it is appropriate to represent the SNR of that ROI as a single scalar value rather than considering the whole distribution from all voxels within that ROI; it doesn’t actually involve “interpolation” in any way.
Rob
1 Like