I would strongly agree with Andreas's caution of using the FreeSurfer
labels, especially when the arcuate is fairly easy to identify entirely and
correctly in individual subjects (with 2 fiber probabilistic tractography,
that is), whereas the exact definition of Broca's and Wernicke's areas are
not even clear in the literature, let alone individual subjects. In my
publications I have tended to make the assumption that the terminations of
the arcuate approximate locations of language related frontal and temporal
cortex (and showed for a small meta analysis of temporal activations in
language fMRI studies that this assumption is not crazy), rather than
drawing ROIs where I think language cortex is and tracking the arcuate from
these. The arcuate in my opinion is a readily identifiable anatomical
structure with diffusion imaging like the thalamus is with structural
imaging, whereas the cortical areas involved in language cannot be directly
identified anatomically (yet).
I also agree that waytotal normalization might not be the way to go here,
given that you are interested in the asymmetries of FA.
To be able to comment more specifically and helpfully, it would be necessary
to know what questions you are trying to answer with this analysis.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Andreas Bartsch
Sent: Monday, January 18, 2010 2:38 PM
To: [log in to unmask]
Subject: [FSL] AW: [FSL] Publications normalizing tractography using
waytotal
Hi,
>I defined these anatomically using labels from Freesurfer
Oh, does Freesurfer distribute these? Or did you assume it's
Wernicke's/Broca's by cytoarchitectonic labels?
I guess I do not need to speel out the words of caution here (pertaining to
Brodmann labelling, correspondence of BA labels to "real" functionals, the
arcuate also connectingto Exner's etc etc).
>Here's what I did and what I calculated...
Sounds good - one possible way to do it.
>We did not do this
Well, that's the point. I guess I would have used Matt's method instead of
the waytotal normalization. Simply because a) you have no prior knowledge to
rely on that the arcuate really exists in the particular individual
(imagine, for example, that a clear angular gyrus is not identified in every
subject - maybe in just 66%, depending on the sample). and b) it sounds like
you want to detect asymmetries (and not enforce symmetry because, for
example, the tracking is adversely affected on one side due to some
pathology) - so I would have used Matt's method in your case.
Matt - any ideas?
Cheers-
Andreas
________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von
Lindgren, Kristen, Ann [[log in to unmask]]
Gesendet: Montag, 18. Januar 2010 18:17
An: [log in to unmask]
Betreff: Re: [FSL] AW: [FSL] Publications normalizing tractography using
waytotal
If I see you at a conference, I'll definitely buy you a beer :)
i) how did you define Broca's / Wernicke's?
I defined these anatomically using labels from Freesurfer
ii) what asymmetry quotients did you compare?
Here's what I did and what I calculated. Used probabilistic tractography
with multiple masks, identified the two masks, one corresponding to Broca's
area and the other to Wernicke's area, and checked off that seed space was
not diffusion. This was done in each hemisphere. I then normalized the
resulting fdt_paths for each subject and applied a constant threshold. I
then binarized each normalized and thresholded tract and multiplied it by
the participant's FA map. I then calculated average FA for the tract in each
hemisphere for each participant. I ensured that I was only looking at white
matter voxels by confining my calculations to those voxels that overlapped
with the participant's white matter label from Freesurfer.
For calculations, I looked at FA two ways. First I ran a MANOVA with FA as
the dependent variable, group as the between-subjects independent variable,
and hemisphere (left, right) as the within-subject independent variable.
For the second analysis, I calculated an asymmetry quotient for FA using the
formula (Left - Right)/[(Left + Right)/2]. These I compared with t-tests.
I did not do volumetric analyses or assess volumetric asymmetries.
iii) How did you ensure apriori that the arcuate actually exists?
We did not do this and I'm not really sure how you would, but we based our
assumptions that we would gather parts of this pathway based on previous
literature about this pathway and its connections.
I appreciate any thoughts you have. Thanks!
Kristen
-----Original Message-----
From: FSL - FMRIB's Software Library on behalf of Andreas Bartsch
Sent: Sat 1/16/2010 12:15 PM
To: [log in to unmask]
Subject: [FSL] AW: [FSL] Publications normalizing tractography using
waytotal
When / where is the party;)?
No - seriously, i) how did you define Broca's / Wernicke's? Anatomically,
functionally, by ESM?; ii) what asymmetry quotients did you compare? As
explained previously, the waytotal normalization enforces symmetry whereas
Matt's approach may be more sensitive to asymmetries. How did you ensure
apriori that the arcuate actually exists?
>pull out the FA value of each individual voxel
Sure - e.g. mask the FA volume by the (thresholded or unthresholded)
fdt_path and write out all values into an ascii file (fsl2ascii) and then do
whatever you want with it...
Cheers-
Andreas
________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von
Lindgren, Kristen, Ann [[log in to unmask]]
Gesendet: Samstag, 16. Januar 2010 17:15
An: [log in to unmask]
Betreff: Re: [FSL] AW: [FSL] Publications normalizing tractography using
waytotal
Matt and Andreas, thanks for the info. I used probabilistic tractography
between gray matter masks in Broca's and Wernicke's areas (so presumably
parts of the arcuate fasciculus) and used the resulting fdt_paths mask to
ascertain an average FA value across all of the voxels. I then compared the
average FA values between two groups on each side and also compared
asymmetry quotients. From earlier conversations, it sounded like waytotal
thresholding was the best way to do this, whereas controlling for ROI size
was better if I was looking at the volume of the resulting fdt_paths.
Also, just out of curiosity, is there anyway to pull out the FA value of
each individual voxel in each participant's fdt_paths instead of having to
calculate an average FA across all voxels?
Thanks again for all of your help! I seriously couldn't have finished my
thesis without you guys!
Kristen
-----Original Message-----
From: FSL - FMRIB's Software Library on behalf of Andreas Bartsch
Sent: Sat 1/16/2010 3:54 AM
To: [log in to unmask]
Subject: [FSL] AW: [FSL] Publications normalizing tractography using
waytotal
Hi,
yep - it's part of the book. The underlying Bayesian formulas were taken out
in the editing process, however. We have not made it - due to time
constraints - to seperately publish it in a journal but will hopefully do
this year.
Anyway - it's not such a big deal: as described previously in the list, the
waytotal i) depends on the prior that you know the tract is there (in the
particular individual examined) and ii) does not alter the spatial extent of
fdt_path itself but simply scales the values. Later, by thresholding you set
a cutoff that impacts the spatial extent of the tract.
So you have to ascertain prior to the tractography (!) and by other, e.g.
clinical means, that you can be sure that the tract is there and you have to
decide if you consider scaling by the waytotal appropriate prior to
thresholding.
Is it the spatial extent of the tract you are after, or some symmetry
analysis, or the probability values themselves?
Cheers-
Andreas
________________________________________
Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Matt
Glasser [[log in to unmask]]
Gesendet: Freitag, 15. Januar 2010 23:12
An: [log in to unmask]
Betreff: Re: [FSL] Publications normalizing tractography using waytotal
Hi Kristen,
I have not published any papers with this method yet, but Andreas Bartsch
may have. I know the method is described in Heidi and Tim's book, in
Andreas's chapter (19). The book (or at least part of it) is on Google
books:
http://books.google.com/books?hl=en&lr=lang_en&id=N20nnxByjVAC&oi=fnd&pg=PT9
&dq=Diffusion+MRI+from+quantitative+measurement+to+in+vivo+neuroanatomy&ots=
FTIbmZwxe3&sig=f8ySrMDRvk3gKbSvvVcZWX6ftik#v=onepage&q=&f=false
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Kristen Lindgren
Sent: Friday, January 15, 2010 4:00 PM
To: [log in to unmask]
Subject: [FSL] Publications normalizing tractography using waytotal
Hello. This question may be best answered by Matt Glasser, but I'm
wondering if anyone is aware of a publication using probabilistic
tractography and normalizing the results using the waytotal values. I'd be
especially interested in ones looking at the arcuate fasciculus. Way back
when when I was working on my thesis, Matt said that he was using this in
his studies, but I'm not sure if he or any others have published using this
method. I'm working on submitting my DTI paper and would love to have
something more than "personal communication" to site when describing my
method. Any thoughts or help is appreciated. Thanks in advance!
Kristen
|