Pipeline Release and Morphometry Measurements

Questions about BrainVisa usage and installation

Moderators: denghien, riviere

Post Reply
User avatar
François Leroy
Posts: 30
Joined: Wed Feb 22, 2006 5:22 pm

Pipeline Release and Morphometry Measurements

Post by François Leroy »

Hi,

In my datasets, the morphologist pipelines 2007 and 2012 have been applied to dbA and dbB, respectively. When computing sulcal depth (with my own piece of software), I have larger sulcal depth in dbB (2012) than in dbA (2007). This could be due to many issues, including a change of software or hemispheric brain volume differences between subjects in dbB compared to subects in dbA.

In order to better understand these sulcal depth differences, I took one subject to which I applied both pipelines. Here are the differences I noticed:

1) despite identical AC-PC coordinates, the normalization transformation is surprisingly slightly different: in the common Talairach space, the MRI is more compressed along the X (sagittal) axis in 2012 than 2007, whereas it is more dilated along the Z (axial) axis in 2012 than 2007.

2) the hemispheric brain volume is 7.5% larger in 2012 than in 2007, which approximately matches a 0.5mm volume inflation (half a voxel) between 2007 and 2012.

3) According to this assumed volume inflation, the superior temporal sulcus is on average deeper in 2012 than in 2007 (using the Morphometry Statistics tool). The depth increase is of 0.9mm (almost one voxel).

My understanding is that I should not use two different pipelines to compute morphometric features. I probably have to re-run dbA with pipeline 2012 to have all measures with the same pipeline (2012). This is fine with me.

However, I deal with databases that have been processed by different labs so that I don't easily know which pipeline has been applied to it. How can I check the pipeline release that has been used? For example, is there a flag in the dataset that says which pipeline has been applied to it, eg pipeline 2007, 2011, 2012 or 2013?

Furthermore, I think this is important to know when and why measurements of hemispheric volume or sulcal features change in a software release. For example, we might have to re-run the new pipeline over an old dataset to comfirm a set of results. Is there some documentation about those changes?
In particular, the brainvisa sulci seem to get closer to the brain envelope/hull in the latest release than in the older ones. When loading together the hemispheric brain volume and the sulci graph, the sulci are more visible in the latest release than in the older ones. Do you confirm this feeling I have and if so, when did that occur?

Thanks so much for your help,
Francois
User avatar
riviere
Site Admin
Posts: 1361
Joined: Tue Jan 06, 2004 12:21 pm
Location: CEA NeuroSpin, Saint Aubin, France
Contact:

Re: Pipeline Release and Morphometry Measurements

Post by riviere »

Hi François,

There are indeed (and hopefully) differences in results from the pipelines 2012 and 2007: globally changes have been made in many steps to imporve the quality and robustness of the segmentations. I admit there is a lack of documentation on the expected effects. Maybe you should talk with Clara who has made all the improvements to get more details...

I only answer here the few questions I have elements right now:

1) First you have to make sure that the Talairach tranfsormation is done usibng the same method: in recent pipelines the default is to use a "real" normalization procedure (based on SPM or FSL), and do not use the AC/PC points by default. If they are used (in the non-default mode), the transform is based on these points, and also on the bounding box of the brain mask segmentation, which means that segmentation differences can have a (slight) impact on the Talairach transform.

2-3) These are probably "improvements" but I cannot answer without seeing the data and segmentations...

In any case, of course you should not use different pipelines to process different sub-parts of the data: each version produces different results, with probably different systematic biases.

There was no way to know a posteriori which pipeline version was used to process data before version 4.3. This is the job of "history management" which was only introduced in BV 4.3, released in 2012. Thus it is possible now, provided you have activated history management on your BV databases (which is still not activated by default since it will write many additional files in the database, it may be activated or disabled on each database, individually). It allows to get the full log (version, environment, process and parameters, plus execution information) for any data written by BrainVisa.

Denis
User avatar
François Leroy
Posts: 30
Joined: Wed Feb 22, 2006 5:22 pm

Re: Pipeline Release and Morphometry Measurements

Post by François Leroy »

Hi Denis,

I just spoke with Clara. There are indeed minor changes to the brain morphometry.

1) you're right: the slight brain reduction and dilation according to Z / X, respectively, probably come from different normalization procedures, ie non-linear (using SPM, 2012 pipeline) or linear (using standard AC-PC coordinates transformation, 2007 pipeline). Thus, it is not a bias that occurrs for each subject but is rather subject-specific. The dilation or reduction might be slightly different depending on the brain morphology, the tissue contrasts, etc.

2) There is probably a slight increase of the hemispheric pial volume in the 2012 pipeline compared to the 2007 one. However, this is to be confirmed.

3) Sulci are slightly deeper and larger since their junctions to the brain envelope have been improved (increased) for brains with open folds (typically old subjects). The sulcal mesh is also more accurate so that the sulcal area should be slightly larger for pipeline 2012 compared to pipeline 2007. As for sulcal depth, I compared both pipelines. I computed a 1 to 2mm homogeneous depth increase along the superior temporal sulcus.

I hope that the Brainvisa team will find time to include a short report on the brain morphometric changes occuring in the next pipeline updates.
w best wishes
Francois
Post Reply