Hi Mark,
Before I had received your message, I had tried a slightly different way than previously described. My brains seem to align well with each other (i.e. each subject's 3 runs) and with the standard brain when everything is in standard space. Is this a good indication that it worked? Here are the steps I ended up using:
1) transform each subject's betted mag images and phase maps into standard space, average each
2) erode the edges of the mag image a bit (a recommendation from someone working on brainstem matters)
3) create an averaged radian image
4) translate the average rad and mag images back into subjects' functional space (note that I go back to functaional space where I have an average of the functional runs, which I use to pre-align my BOLD runs before using a single transformation from "bold mean" space to standard per subject, since the standard methods doesn't seem to be accurate enough for fine structures)
5) I checked the values in the averaged rads as compared with other subject's rad images, and the values are similar (+/-400 or so). They look quite similar though of course a bit more blurry.
6) I then use the b0-unwarping option and enter the averaged images when I put in my functional runs.
7) to check alignment, I transform my unwarped runs to standard space and see that structures of interest are well aligned and there are no other funny distortions.
Since it seems to work, why I like this method is it doesn't require me to work out a threshold (since to be honest I'm not really sure what this step is doing or how to select it!). Should I try to run things like this or is there something fundamentally wrong with it?
If I should try it your way, may I ask for clarification on a couple of things? This is what I've understood as the steps, based on your previous message:
1) transform all the rads image into standard space, average them
2) transform the average standard rad image into native space
3) threshold the rads image based on looking at the values in the others - what does this do exactly? What if the values are similar?
4) create a mask (i.e. binarise) your mean native thresholded rad image using fslmaths (?) - is this something that happens automatically when you enter the mags and rads image in the b0 unwarping option of fsl stats? I'm not sure what this is doing
5) use this mask to mask a copy of T1 which will replace the magnitude brain-extracted image into the prestats tab
My apologies for the long question.
Best regards,
Emily
|