Dice coef for warped anatomical and template brain?

AFNI Message Board
1

Sri Gurujis-

We really like the new QC options you’ve made available via afni_proc. Two questions emerge:

  1. Might we expect to see a correspondence between the qc_00_ve2a_epi2anat.???.jpgs and the “anat/EPI mask Dice” coefficient generously provided in the out.ss_review.*.txt file? I don’t see documentation pointing to what the Dice coef is but I’m assuming it’s something like the Sørensen–Dice coefficient?

  2. Is there a way to obtain something similar to the Sørensen–Dice coefficient but for indexing the similarity between the warped anatomical and the template to which it was warped?

I’m asking the last question because we’re doing QC on a large consortium-dataset of ~1500 braaaaiiiins! (happy Halloween) and it would be nice to have a handy summary index of the quality of the anat2temp warp for each subject.

Regards,

Paul

2

Hi Paul,

Yes, that is the Sorensen-Dice coefficient.

  1. The correspondence won’t be great, as the dice coefficient mostly shows whether the masks are approximately on top of each other. Since the coverage tends to differ, it mostly just shows whether there was a gross failure.

  2. Sure, I will look into adding it for the anat/template.
    Okay, that is in. Once we do a new build, the AP analysis should include a pair of results, like:

anat/EPI mask Dice coef : 0.840899
anat/templ mask Dice coef : 0.947241

Does that seem good?

Thanks,

  • rick
3

Hi Rick–

“1. The correspondence won’t be great, as the dice coefficient mostly shows whether the masks are approximately on top of each other. Since the coverage tends to differ, it mostly just shows whether there was a gross failure.”

Do you have a Dice coefficient criterion for gross failure?

“2. Sure, I will look into adding it for the anat/template. Okay, that is in. Once we do a new build, the AP analysis should include a pair of results, like:
anat/EPI mask Dice coef : 0.840899
anat/templ mask Dice coef : 0.947241”

That is excellent. Thank you!! We’ve run our 1500 or so datasets through the pipeline over the last two months so we probably won’t re-run the updated afni_proc on them. Do you mind replying with the updated code-chunk that operates on the warped anat and the template volume and I’ll insert it into a shell script?

Tusen tack from out here…

Paul

4

Hi Paul,

There is a “3ddot -dodice” command in your proc script now, that gets the coefficient between the EPI and anat masks. So this one would compute it between the anat and group (template) masks, as in:

3ddot -dodice mask_anat.$subj+tlrc mask_group+tlrc \
      |& tee out.mask_at_dice.txt

That is what the command will look like in an updated script. The “at” is short for anat/template.

Does that seem reasonable?

  • rick
5

Just right, Rick! Many thanks! I’ll apply generously to the data…