[Mrtrix-discussion] FA and ADC values for a tract

Wed Mar 26 14:59:52 PDT 2014

Hi All,
Yes, HMOA is identical to AFD. Flavio came up the idea at the same time as
us. In fact he independently coined the term AFD, however unfortunately the
reviewers would not let him use it since our publication came out first
(even though he submitted his paper earlier).

Luis, yes as long as the DWI is normalised then this is fine. We normally
perform a global normalisation based on the b=0 white matter (or CSF if
your population/resolution allows a partial volume free estimate), followed
by a bias field correction. This ensures we don't have issues in voxels
containing partial volume with CSF.  You could also perform a voxel-wise
normalisation to the b=0, as long as you are happy to accept that the AFD
in voxels with some CSF with be artificially smaller (something that most
people seem to accept with other measures such as FA and ADC). If you are
doing a voxel-wise normalisation ideally you need a number of b=0's,
otherwise noise in the b=0 heavily influences the AFD.

Luis, I'm guessing when you say tract-specific, you are averaging the AFD
within a tract-of-interest? Since AFD is proportional to the intra-axonal
volume, ideally you need to be summing the AFD to account for differences
in fibre bundle width (as the modulation does in voxel-wise analysis).
However, one issue with summing the AFD is that your measure is then
sensitive to differences in fibre bundle length - something that should not
really be related to the connectivity of that bundle. One way to account
for this is to normalise the summed AFD by the mean streamline length.
However, this is also not perfect since your tract-of-interest may contain
voxels that have contributions from other (unrelated) fibre bundles (eg one
bundle may merge with another at some point), and therefore your AFD
measure is not purely related to your tract-of-interest. This is where
SIFT-ed streamline counts should more appropriate. SIFT relates the
steamlines to the underlying AFD in a global fashion, and therefore allows
you tease out the contributions from specific tracts even where tracts
merge together (sorry it's a bit hard to explain without a diagram).

Since Thijs is so good at plugging AFD (keep going and we'll have to give
you a job ;)), I might as well shamelessly plug Rob's SIFT ISMRM abstract
from last year (see attached).

The next major release of MRtrix will contain SIFT and some AFD tools.
However, FOD registration probably won't be included until a few months
after the initial release.

Cheers!
Dave

On 27 March 2014 08:06, Luis Concha <lconcha at unam.mx> wrote:

> Hi Dorian and fellow mrtrixers.
>
> Yes, it is very much like HMOA, which is actuallly just like AFD (*,**) .
> There is, however, one difference: In the approach I mentioned, the tracts
> are first computed using the tractography algorithms supplied by
> streamtrack, and in the paper by Dell'Acqua, HMOA is used to drive the
> tractography algorithm itself, as well as to select tract-specific HMOA a
> posteriori (Figure 9 in Dell'Acqua).
>
> My main concern is that I use find_SH_peaks in order to simplify the
> identification of the corresponding (parallel) FOD peak to the tract
> segment. Given that find_SH_peaks output can be affected by the number of
> directions used as seeds for  peak finding, is there any min number that
> should be used? (I used 120).
>
>
> I would be happy to share my scripts to obtain AFD/HMOA, if anyone is
> interested.
>
>
> (*) Marco Catani, during the ISMRM Diffusion Workshop last year, said that
> it was one of the rewiewers who asked them to change AFD to HMOA.
> (**) One key point in Raffelt's paper on AFD is the implementation of a
> voxel-based framework for the analysis of this measurement for group
> analyses, which I am hoping will be available in mrtrix soon.
>
> Luis
>
>
>
>
>
> On Wed, Mar 26, 2014 at 2:17 PM, Dorian P. <alb.net at gmail.com> wrote:
>
>> Hello everybody, hello Dr. Concha.
>>
>> As we are on the topic, it seems to me the idea of Dr. Concha is similar
>> to the HMOA measure of Dell'Acqua et al.
>> http://www.ncbi.nlm.nih.gov/pubmed/22488973
>>
>> Am I correct thinking they are similar, if not identical measures?
>>
>> I am looking for ways to obtain this measure in mrtrix. Anyone can give
>> some hints or share the steps to do this? Will mrtrix plan to output the
>> amplitude of FODs for tracts in next releases?
>>
>>
>> Thank you
>> Dorian
>> Thomas Jefferson University
>>
>>
>>
>>
>>
>> On Wed, Mar 26, 2014 at 3:58 PM, Luis Concha <lconcha at unam.mx> wrote:
>>
>>>
>>> Hi Thijs.
>>>
>>> It is great to know that the new release of mrtrix will have tools for
>>> AFD. In the meantime, I have played around with some scripts to obtain such
>>> a measurement in a tract-specific way and I would appreciate your input and
>>> that of others in this list. The scripts do this:
>>>
>>> 1. Perform tractography and select a set of streamlines for analysis.
>>> 2. For each streamline segment (i.e., the straight line between two
>>> adjacent steps), find the FOD lobe that is most parallel (dot product
>>> between streamline segment and FOD peaks).
>>> 3. Once the FOD lobe has been identified, extract its amplitude
>>> (tract-specific AFD).
>>>
>>> I am assuming that if DWI signal is normalized (as you mentioned), then
>>> FOD amplitudes should be comparable between subjects. Also, since there is
>>> no spatial normalization of images, there is no need to modulate the AFD to
>>> compensate for the spatial transformation.
>>>
>>> If you find any flaws with this, please let me know. Oh, and if you
>>> could give us a teaser of what new features we will see in the new release
>>> of mrtrix, that would be fantastic!
>>>
>>>
>>>
>>>
>>> Dr. Luis Concha
>>> Instituto de Neurobiología
>>> Laboratorio C-13
>>> UNAM, Campus Juriquilla
>>> Boulervard Juriquilla 3001
>>> Juriquilla, Querétaro.
>>> C.P. 76230
>>> México
>>> Tel (442) 2 38 10 54
>>> Fax (442) 2 38 10 46
>>> http://personal.inb.unam.mx/lconcha/
>>>
>>>
>>> On Wed, Mar 26, 2014 at 1:47 PM, Thijs Dhollander <
>>> thijs.dhollander at gmail.com> wrote:
>>>
>>>> Hi Helen,
>>>>
>>>> Opinions on FA/ADC quantification can be wildly varying... :-)  There's
>>>> no problem calculating and reporting their values, but interpretation is a
>>>> big issue: they can be quite sensitive in many cases, but they are always
>>>> far from specific.  To avoid starting a long rant on how we should abandon
>>>> these measures as quickly as possible, I'd rather refer to this nice
>>>> article:
>>>> http://www.sciencedirect.com/science/article/pii/S1053811912007306 ;
>>>> it contains some nice and clear passages explaining some of the relevant
>>>> issues to be wary of, as well as further references that are worth a good
>>>> read.
>>>> So the question remains: what's better then?  There's quite some
>>>> alternatives, but I personally particularly like apparent fibre density
>>>> (AFD).  A nice and in-depth article on it can be found here:
>>>> http://www.sciencedirect.com/science/article/pii/S1053811911012092 .
>>>> Last year's ISMRM had an abstract on some further improvements on analysing
>>>> AFD: http://cds.ismrm.org/protected/13MPresentations/abstracts/0841.pdf(if you or anyone in your lab attended, they can access it;  the
>>>> accompanying oral presentation can be found at
>>>> http://cds.ismrm.org/protected/13MPresentations/0841/ ).  The next big
>>>> version of MRtrix (expected to be released somewhere around this year's
>>>> ISMRM in May... ;-)) will contain much more tools that provide you with the
>>>> means to perform such analyses.
>>>> In the mean time, you can already easily obtain an (orientationally
>>>> averaged) AFD map by simply calculating the average of your (normalized)
>>>> DWI images from a single shell acquisition.  The same contrast (up to a
>>>> fixed constant factor) can also be obtained by opening up the result from
>>>> an SH fit of the signal (or even a CSD outcome) in mrview, and looking at
>>>> the first coefficient (i.e. the 0th order).
>>>>
>>>> Hope this already helps somewhat!
>>>>
>>>> Cheers,
>>>> Thijs
>>>>
>>>> PS: now that I've mentioned the upcoming ISMRM; I'd like to encourage
>>>> everyone on this list that still has many questions (or not) to surely
>>>> attend and look for each other -- discussions on these new and better ways
>>>> of quantifying are always much more lively in person :-)
>>>>
>>>>
>>>>
>>>> *Thijs Dhollanderthijs.dhollander at gmail.com
>>>> <thijs.dhollander at gmail.com> Tel. +32 475 36 44 27
>>>> <%2B32%20475%2036%2044%2027>*
>>>> Medical Image Computing (MIC), ESAT-PSI, Department of Electrical
>>>> Engineering, KU Leuven
>>>>
>>>>
>>>> On Wed, Mar 26, 2014 at 4:52 PM, Helen Carlson <
>>>> Helen.Carlson at albertahealthservices.ca> wrote:
>>>>
>>>>> Hello Thijs
>>>>>
>>>>> Thank-you for your reply to my question. I would like to quantify
>>>>> changes in white matter over time, specifically pre/post treatment. Is
>>>>> there a better way to do this than using the mean FA and ADC? Or is this
>>>>> measurement appropriate?
>>>>>
>>>>> Thanks again
>>>>>
>>>>> Helen.
>>>>>
>>>>>
>>>>>
>>>>> *From:* mrtrix-discussion-bounces at www.nitrc.org [mailto:
>>>>> mrtrix-discussion-bounces at www.nitrc.org] *On Behalf Of *Thijs
>>>>> Dhollander
>>>>> *Sent:* March 23, 2014 9:47
>>>>> *To:* mrtrix-discussion at www.nitrc.org
>>>>> *Subject:* Re: [Mrtrix-discussion] FA and ADC values for a tract
>>>>>
>>>>>
>>>>>
>>>>> Hi Helen,
>>>>>
>>>>> What you're seeing is (probably) correct: FAs larger than 1 and ADCs
>>>>> below 0 can appear due to the tensor estimation in MRtrix being a simple
>>>>> (unconstrained) linear least squares one.  The tensors are not explicitely
>>>>> constrained to be positive definite, and thus negative eigenvalues might
>>>>> appear.  When one or more eigenvalues of a given tensor are negative, the
>>>>> FA can go beyond 1, while the ADC can go below 0.
>>>>>
>>>>> Cheers,
>>>>> Thijs
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> *Thijs Dhollander thijs.dhollander at gmail.com
>>>>> <thijs.dhollander at gmail.com>Tel. +32 475 36 44 27
>>>>> <%2B32%20475%2036%2044%2027>*
>>>>> Medical Image Computing (MIC), ESAT-PSI, Department of Electrical
>>>>> Engineering, KU Leuven
>>>>>
>>>>>
>>>>>
>>>>> On Fri, Mar 21, 2014 at 10:10 PM, Helen Carlson <
>>>>> Helen.Carlson at albertahealthservices.ca> wrote:
>>>>>
>>>>> Hello all,
>>>>>
>>>>> I would like to determine the average FA and ADC values for a given
>>>>> track (in this case, the arcuate). So I followed previous directions posted
>>>>> here and have used the following commands on my arcuate which successfully
>>>>> creates a probability mask for FA and ADC:
>>>>>
>>>>>
>>>>>
>>>>> tracks2prob -template fa.mif Arcuate.tck FA_mask.mif
>>>>>
>>>>> tracks2prob -template adc.mif Arcuate.tck ADC_mask.mif
>>>>>
>>>>>
>>>>>
>>>>> I then used mrstats to get the average, max, min etc of the tracks by
>>>>> using these commands:
>>>>>
>>>>>
>>>>>
>>>>> mrstats fa.mif -mask FA_mask.mif
>>>>>
>>>>> mrstats adc.mif -mask ADC_mask.mif
>>>>>
>>>>>
>>>>>
>>>>> Which also seemed to work except that I was expecting my FA to range
>>>>> from 0.1 to 1.0 (approximately). Instead, the FA values have a min of 0 and
>>>>> a max of 1.13. In addition, my ADC values have a negative minimum
>>>>> (-0.0024). It seems these values are out of range and do not make sense but
>>>>> I do not know what I am doing wrong.
>>>>>
>>>>>
>>>>>
>>>>> Thanks in advance for any help
>>>>>
>>>>> Helen
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> *Helen Carlson, Ph.D*
>>>>>
>>>>> Neuroimaging Research Associate
>>>>>
>>>>> Calgary Pediatric Stroke Program & Neuropsychology
>>>>>
>>>>> Alberta Children's Hospital
>>>>>
>>>>>
>>>>> ------------------------------
>>>>>
>>>>> This message and any attached documents are only for the use of the
>>>>> intended recipient(s), are confidential and may contain privileged
>>>>> information. Any unauthorized review, use, retransmission, or other
>>>>> disclosure is strictly prohibited. If you have received this message in
>>>>> error, please notify the sender immediately, and then delete the original
>>>>> message. Thank you.
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Mrtrix-discussion mailing list
>>>>> Mrtrix-discussion at www.nitrc.org
>>>>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Mrtrix-discussion mailing list
>>>>> Mrtrix-discussion at www.nitrc.org
>>>>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>>>>
>>>>>
>>>>
>>>> _______________________________________________
>>>> Mrtrix-discussion mailing list
>>>> Mrtrix-discussion at www.nitrc.org
>>>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>>>
>>>>
>>>
>>> _______________________________________________
>>> Mrtrix-discussion mailing list
>>> Mrtrix-discussion at www.nitrc.org
>>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>>
>>>
>>
>> _______________________________________________
>> Mrtrix-discussion mailing list
>> Mrtrix-discussion at www.nitrc.org
>> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>>
>>
>
> _______________________________________________
> Mrtrix-discussion mailing list
> Mrtrix-discussion at www.nitrc.org
> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>
>

-- 
*David Raffelt (PhD)*
Post Doctoral Fellow

The Florey Institute of Neuroscience and Mental Health
Melbourne Brain Centre - Austin Campus
245 Burgundy Street
Heidelberg Vic 3084
Ph: +61 3 9035 7024
www.florey.edu.au
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20140327/6fbf3a9f/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 2135.PDF
Type: application/pdf
Size: 1093118 bytes
Desc: not available
URL: <http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20140327/6fbf3a9f/attachment-0001.pdf>