Hi everyone,
I am currently using the noise maps of dwidenoise for my SNR calculations of GRAPPA multichannel coil ss/rs-EPI DWI images (because other ways like the subtraction method, etc. yielded very inconsistend noise values). I am very happy with the results because I finally have consistent noise levels across my repetitions with the same parameters. But since my images have a voxel size of 0.6x0.6x3mm (slice thickness of 3mm) I wonder if I can use the MPPCA algorithm with a kernel size of 5x5x1. Does anybody know if MPPCA can be used in the “2D-plane” and still have accurate noise estimation?
I already tried a kernel size of 5x5x5 on my images and the noise levels where about the same as 5x5x1. Sadly I cannot find a source for this and for my article I may have to justify the smaller kernel size.
All comments are much appreciated.