We’re currently working on an open-source DWI dataset using the FBA pipeline for single-shelled data. When doing some basic QCing of the images (i.e. visual checking for Venetian Blinding, signal dropouts etc), we noticed that most scans contain signal hyperintensities - most often around WM/CC, or CSF (see below image):
I’m not sure what you’re referring to exactly, but the region you’re highlighting here will probably have high intensity due to the presence of CSF and/or strong signal pile-up due to susceptibility-induced distortions around the orbitofrontal cortex. In general, you’d expect high signal in the CSF in the b=0 images since diffusion MRI is typically strongly T2-weighted (long TE), and CSF has a long T2. I’m not sure I see anything unusual about the data you’re showing here.
As regards the influence of these hyperintensities on the response function estimation: most of our approaches handle these issues without any obvious problem – but yes, under the ‘right’ conditions, this can sometimes affect the process. It would certainly be worth double-checking the response functions you get if you have concerns here.
Thanks for the helpful response @jdtournier! Given that we are seeing signal hyperintensities in regions where there is typically a strong signals pile-up (i.e. orbitofrontal cortex, WM-CSF), it is good to hear that there isn’t anything that we should too be worried about at present.
We will definitely visually check the response functions to see if there are any unexpected deviations due to the influence of these hyperintensities. Thanks again
On that note, it’s worth adding the -voxels option to dwi2response to inspect the location of where the response was estimated from. That way you can verify whether these problematic regions were used or not.
Sorry for the delayed response to your post @jdtournier, I only just saw this now. Yep, that is a good point in terms of using the the -voxels option and the associated voxels.mif files. Thanks again for your help on this!