BIDS and NonlinearGradientCorrection

1

BIDS has the NonlinearGradientCorrection key to indicate whether the the gradient nonlinearity correction has been applied at the MRI scanner or not. If I want to save both the uncorrected and corrected images in the same anat folder, is there any recommended way of naming the two files ? I am thinking of using acq- to distinguish what would otherwise be identical filenames.

2

Hi,

Good question!
One aspect is to be sure that the metadata: NonlinearGradientCorrection is present is the accompanying json file. (at the moment it is RECOMMENDED by BIDS specification but may be REQUIRED if PET data is present).
Regarding the saving of two versions of the same image, depending if the gradient non linear correction filter is turned on or off, I would personally suggest to use use the rec- label, as it is not a parameter affecting the acquisition per se, more the reconstruction of the image.
You may use for example rec-NLGC for example but this is really up to you.

3

Thanks so much @jsein . rec- makes sense. I have a related question. For the T1w, the two json files differ by contents of ImageType as saved by dcm2niix – the uncorrected one having ND , whereas the corrected one having DIS2D and DIS3D . _ND also appears in the SeriesDescription, as I have seen in other Siemens datasets as well. The thing is: the other modalities in the same sessions which DO NOT have two versions saved also have ND in ImageType. Does this mean they are not corrected for gradient nonlinearities ?

4

The thing is: the other modalities in the same sessions which DO NOT have two versions saved also have ND in ImageType . Does this mean they are not corrected for gradient nonlinearities ?

I think you are totally right: ND in Image Type means that the Non linearity Gradient Correction filter was not applied.

Side note: all this thread is specific to Siemens terminology, I don’t know if the other constructors label their images in this way regarding wether the Non Linear Gradient Correction was applied or not.

Side note 2: if you wanted to apply Non Linear Gradient Correction to images that were not corrected, it is still possible: either at the console of the scanner where the images were acquired, or with the gradunwarp program, for which you need to provide the constructor’s gradient non linearity file.

5

Thanks again. Why wouldn’t one want the correction to be applied to ALL modalities, I wonder.

Yes, I have tried using gradunwarp before, but in this case it might not be easy to get access to the coefficients. Oh well.

6

On which scanner were these images acquired on? Depending on the scanner, the brain in the Field-of-View may very well be in the Linear part of the Gradient and the correction would not change anything. You can assess that by comparing your two anatomical images, the one with and the one without Non Linearity Gradient Correction.

7

Thanks for the tip! It’s a Prisma_fit.