I’m running a VBM analysis with the cat12 toolbox. After preprocessing for most images, the cat12 automated report has poor ratings for the bias under “Image and Processing Quality.” The ratings are typically around 65-70% with a grade of D+, which shows up in an ominous red color.
But I can’t find any guidance on 1) how this rating is determined 2) any specific steps I can take to improve it 3) whether the results are still valid
I’m new to neuroimaging in general, so I may be missing something very basic.
Any help would be greatly appreciated!
The bias of the image is related to the hardware (B1+ and B1- field, emitting and receiving coils) and also the filters applied to your images by the scanner (for instance with Siemens: use of normalize or Prescan Normalize filters). So to better understand your question, could you feed us with the magnetic field you are working at, the type of scanner and coils that were used to acquire those images?
Ok, I will have to check with our data, we use a 3T Prisma with 64channel coil and I think both 32ch and 64ch have similar receive field with a pretty strong receive bias. Using the Prescan Normalize filter would give you images with a better score. Of note post-precessing bias correction steps such as N4 (Ants) but also FSL, Freesurfer work very well also. CAT12, is using a bias correction step also to estimate the bias of the images and to correct for it during its segmentation process). Such post processing steps would work very well also and you don’t necessarily need to have a bias corrected image as input of standard post-processing pipelines as their bias correction routines work well. But again, I will check if we get a D score also for our raw images with no Prescan normalize filter.
I just checked on our data: the bias score of CAT12 (with prescan normalize) on our anatomical images with the 64CH at 3T range from B to C. I can totally understand that you would get D+ with no filter activated on a 32 CH coil. But despite from the strong bias, this 32CH is very good because it gives you a strong signal and many capabilities for acceleration. So,to get a better bias score you could :
use the prescan normalize filter
use other postprecessing bias corrective steps before evaluation
use an other sequence such as MP2RAGE for instance.
Internally, about how the rating is determined, my guess would be that it is something related to the variance measured in the bias field extracted by the CAT12 procedure.
To me it seems like the results are consistent, especially with no filter applied to your images.
Thanks for checking on this!
It sounds like this bias rating is not such a big issue. As you suggested, it’s based on the pre-bias-corrected image and the cat12 bias correction is quite good, so it should not be a problem
I am conducting a study using the myelin maps obtained with the HCP structural preprocessing pipeline using as input T1w MPRAGE and T2w images.To conduct group comparisons between myelin maps is necessary perform a correction of the B1+ transmit field (Glasser et al., 2022, Neuroimage). Unfortunately, our MRI protocol does not include a specific acquisition of B1+ maps.
I would like to know if there is a way for me to estimate B1(+) inhomogeneities from my T1w MPRAGE images in order to get B1+ maps. CAT12 conducts bias correction before segmentation, do you know if I can save the B1+ transmit field maps estimated by CAT12?
May I suggest you to post a new question to Neurostars as your question is not directly related to the initial topic? The bias field estimated by CAT12 is the receive field bias, and to my knowledge, CAT12 has no information on the transmit field bias just using the T1w and T2w image alone.