Diffusion Bundles Transformation question

Narly · Post by **Narly** » Mon Jul 02, 2007 11:51 am

Bonjour a tous!
I have done some probabilisitic tracking (developed by Muriel some time ago now) using command line scripts, and am now trying to use Diffusion Bundles Transmormation to quantify connectivity between 2 ROIs that I have defined (in a *arg image) and resampled from normalised T1 space into each subject's original T2 space. I would like to quantify connectivity between 2 ROIs, and I would like the output to contain values (eg; mean FA, etc) along the selected bundles between but also WITHIN the ROIs of interest (ie; I want the FA etc values from within the parts of the bundles that are within the ROI to also be included).
I have a few questions:
a) I want to compare quantified values across individuals, therefore I would like to apply a T2--> normalisedT1 transformation when necessary. Should this be done at the Diffusion Bundles Transformation stage (and if so, should I specify the T22nanat.trm in 'bundles_transformation' or in 'selection_regions_transformation' or in 'split regions transformation', or in all 3?), or at the 'Diffusion Bundles Analysis' stage, or at both?
b) For the 'Diffusion Bundles Transformation' step, I have selected the *bundle of interest, as well as the 'selected regions' and the *brules file containing the selection rules. Also, for the 'split regions' option, I selected the same *arg image containing my 2 ROIs (since I want to examine connectivity ONLY between these two regions) - is this correct?
c) when I run the process, I get the following error message: 'argument t2_diffusion is mandatory'. I don't quite understand - is it not possible to run 'Diffusion Bundles Transformation' independantly (since I have already done tracking on my data)? I have of course UNselected the 'ROI drawing' and the 'DTI Simple Tracking' options before clicking 'Run'.
d) in the next step, 'Diffusion Bundles Analysis', should I put the *transformed* bundle (output from the last step, once it works) for the 1st 'Bundles' option? And again, if so, do I need to specify the T2-->nanatT1 transformation here again, given that if I enter it previously, the transformation is probably already taken into account when generating the transformed bundle?
thanks in advance for your help,
Narly.

denghien · Post by **denghien** » Wed Jul 04, 2007 4:51 pm

Hi Narly,

Before I answer to questions, just to be certain I understand well: you draw ROI
on T1 then resample ROI in T2 space.

Question a:
If you just want to compare value, then only 'Diffusion Bundles Analysis' is necessary. 'Diffusion Bundles Transformation' is necessary to preprocess your bundles, this step is not mandatory, but usefull to select with more accuracy and/or to isolate into ROI (like you want to do).
To summarize:
"Fiber tracking": to track fiber from diffusion data and ROI: you get bundles
'Diffusion Bundles Transformation': to select with more accuracy bundles from bundles: you get new bundles
'Diffusion Bundles Analysis' : to extract value (FA ....) from bundles

You want to apply the T22nanat.trm to bundles in 'Diffusion Bundles ransformation', you indicate in 'bundles_transformation' AND in 'selection_regions_transformation' AND in 'split regions transformation'. But n 'Diffusion Bundles Analysis', bundles are projected into maps (T2 space) so bundles must be in the same space and the transformation will be anat2T2.trm. I need to confirm this with Yann because I don't never try.

Question b:
It seems OK. The "selected region" and "split_region" mode are independent and complementary.

Question c:
You're right but you use a pipeline and this case you must fill in pipeline arameters. If you use the "Diffusion -> Tracking -> Diffusion Bundles Transformation" this parameter doesn't exist.

Question d:
Now it seems that your bundles are in normalized T1 space and maps are n T2 space. So the right transformation will be nanat2T2.trm. As I said above, I need to confirm this with Yann.

Note: to have the inverse of a transformation, you can use AimsInvertTransformation.

Isa

Narly · Post by **Narly** » Fri Jul 06, 2007 3:51 pm

Hi Isa,
Thanks for your answers. Yes, I have drawn my ROI in normalised T1 space, and resample it into T2 space.
Your suggestions seem to work. I would still like to know whether or not to apply the T22nanat.trm again at the Diffusion Bundles Analysis stage or not. Maybe it would be better not to since I don't actually want to project back into T2 space - the whole point of tranforming into nanat space in the first place is so that I can make inter-subject comparisons. Is this 2nd step (Diffusion Bundles Analysis) always projected back into T2 space? I assume not since it would have to know which transformation to use in order to so so. I hope that by NOT specifying a transformation at this second stage, I will be able to derive analysis values based on the input at this stage - ie; based on the *transformed* bundles in NANAT space.
thanks in advance for your answer,
best,
Narly.

Narly · Post by **Narly** » Fri Jul 06, 2007 4:40 pm

Hi again,
I think that I just answered my own questions, but please confirm that this is correct. I think that I SHOULDN'T specify transformations at ANY point in Diffusion Bundles Transformation nor in Analysis stages (only if I want to visualise my transformed bundles onto nanat space), since given that I selected the *same* ROI across subjects in NANAT space and then resampled it into T2 space, I should be able to make comparisons on the stats obtained the above steps (*features file) across subjects. Otherwise, I would have to specify the T22nanat.trm at the bundle selection step, but then I would also have to specify the nanat2T2.trm at the bundle analysis step, since this latter step uses FA, ADC, VR, SD, etc images, all of which exist in T2 space.
thanks again,
Narly.