Print

Print


ok!
On Jan 24, 2012, at 11:17 AM, Michael Harms wrote:

> Also, one other thing to at least consider is that, even if you have the
> exact same set of 16 gradient directions available as part of your 32
> direction acquisition, the 32 direction scan itself was presumably twice
> as long, so there is the possibility that movement related problems may
> be more severe in your 32 direction data sets.  Or, artifacts arising
> from table vibration (a known problem for Siemen's Trio scanners) may
> well be different between the two sets of acquisitions.  So, in my
> opinion, even if TR, TE, voxel size, head coil, etc were identical, it
> still wouldn't be completely sufficient to simply use 16 of the 32
> directions without first demonstrating that the data from those 16
> directions was of comparable average quality across the two sets of
> acquisitions.
> 
> cheers,
> -MH
> 
> On Tue, 2012-01-24 at 13:50 -0500, David Gutman wrote:
>> Just to check... is EVERYTHING else in your scans the same?    I.e.
>> TR, TE, voxel size, and head coil?
>> 
>> Again I think we discussed this recently in the group listserve... but
>> basically if there are any systematic differences between group 1 and
>> group 2 in terms of acquisition parameters,  it's really impossible to
>> draw any firm conclusions between differences in the group, unless you
>> also model the effect of scanner.     Adding 5-10 subjects that are
>> scanned with both a 16 and 32 gradient protocol is one possibility.
>> Or you should add some more controls with the 32 gradient protocol  so
>> you can account for the effects of scan parameters  in your model.
>> 
>> 
>> I've really wanted to do the same thing in the past where we take old
>> "control" subjects  or old patient populations scanned with sequence X
>> at time X and compare them to time Y with sequence Y to do some
>> exploratory analysis.... but since there are so many known (and
>> unknown things) that seem to effect the exact FA value I have never
>> felt comfortable doing it for the reasons discussed above.
>> 
>> 
>> On Tue, Jan 24, 2012 at 12:57 PM, GwenaŽlle DOUAUD
>> <[log in to unmask]> wrote:
>>> Dear Angela,
>>> 
>>> considering only the 16 directions that are shared by the two groups sounds
>>> like a reasonable option. You will need to change your bvecs and bvals files
>>> accordingly, and use fslroi to get rid of the 16 non-corresponding
>>> directions in your 32 gradients dataset before using dtifit etc. to create
>>> the FA (and other) maps.
>>> 
>>> Cheers,
>>> Gwenaelle
>>> 
>>> 
>>> ________________________________
>>> De : Christine Zakrzewski <[log in to unmask]>
>>> ņ : [log in to unmask]
>>> Envoyť le : Mardi 24 janvier 2012 17h51
>>> Objet : Re: [FSL] TBSS some questions
>>> 
>>> I doubt it.  The images with 32 diffusion gradients will provide a more
>>> accurate measure of the dependent measure (Fractional Anisotropy,
>>> Radial/Mean Diffusivity, etc).  There is probably no way to limit the number
>>> of gradients used to make these calculations. Wait for a more certain answer
>>> from one of the experts.
>>> Christine
>>> 
>>>> Date: Tue, 24 Jan 2012 18:43:20 +0100
>>>> From: [log in to unmask]
>>>> Subject: Re: [FSL] TBSS some questions
>>>> To: [log in to unmask]
>>>> 
>>>> I intend after scanning. Can I consider only the 16 directions that are
>>>> shared by the two groups?
>>>> Thank you
>>>> 
>>>>> 
>>>>> The technician, or physicist at your scanner can probably best answer
>>>>> this
>>>>> question.Christine
>>>>>> Date: Tue, 24 Jan 2012 17:16:32 +0100
>>>>>> From: [log in to unmask]
>>>>>> Subject: Re: [FSL] TBSS some questions
>>>>>> To: [log in to unmask]
>>>>>> 
>>>>>> Is there a way to 'reduce' the number of directions from 32 to 16 (same
>>>>>> scanner) to be able to compare groups?
>>>>>> 
>>>>>> thank you!
>>>>>> Angela
>>>>>> 
>>>>>> 
>>>>>>> the answer is no..... especiallu if group one was scanned with low
>>>>>>> directions and group two was scanned with more directoons
>>>>>>> On Jan 24, 2012 10:32 AM, "Angela Favaro" <[log in to unmask]>
>>>>>> wrote:
>>>>>>> 
>>>>>>>> Thank you
>>>>>>>> But my question is: can I (after performing TBSS scripts) compare
>>>>>>>> subjects
>>>>>>>> with a different number of directions in their DTI? If the quality
>>>>>>>> is
>>>>>>>> different - as you are saying - I think the answer is no.
>>>>>>>> 
>>>>>>>> Angela
>>>>>>>> 
>>>>>>>> 
>>>>>>>>> 
>>>>>>>>> I can provide the following insight:1. The more diffusion
>>>>>> directions,
>>>>>>>> the
>>>>>>>>> better the dependent measure will be.2. You will have perform
>>>>>>>> across-group
>>>>>>>>> analyses with FEAT.3. You should probably resample your original
>>>>>> image
>>>>>>>> to
>>>>>>>>> 1x1x1mm resolution before spatially normailizing to the MNI
>>>>>>>> templae.Hope
>>>>>>>>> this helps.Christine
>>>>>>>>>> Date: Tue, 24 Jan 2012 12:40:50 +0100
>>>>>>>>>> From: [log in to unmask]
>>>>>>>>>> Subject: [FSL] TBSS some questions
>>>>>>>>>> To: [log in to unmask]
>>>>>>>>>> 
>>>>>>>>>> Dear FSL experts,
>>>>>>>>>> I have some questions about TBSS for which I have not found
>>>>>> answers
>>>>>>>> in
>>>>>>>>>> the
>>>>>>>>>> mailing list.
>>>>>>>>>> 
>>>>>>>>>> 1. Which is the influence of the number of directions on the
>>>>>> quality
>>>>>>>> of
>>>>>>>>>> TBSS?
>>>>>>>>>> 
>>>>>>>>>> I have a set of DTI images with B0=800 and number of
>>>>>>>>>> directions=16
>>>>>>>> and
>>>>>>>>>> another set with the same B0 and 32 directions.
>>>>>>>>>> I guess that probabilistic tracking is not reliable with only 16
>>>>>>>>>> directions, but what about TBSS? Can images of the two groups be
>>>>>>>>>> compared
>>>>>>>>>> with TBSS in some way?
>>>>>>>>>> 
>>>>>>>>>> 2. During the TBSS procedure there is a registration to the
>>>>>> standard
>>>>>>>>>> image
>>>>>>>>>> with resultion 1mm*1mm*1mm. This result in a very long time when
>>>>>>>>>> performing statistics with randomise. My original scanning
>>>>>> resolution
>>>>>>>> is
>>>>>>>>>> 2mm. Do you think that registration of the skeletonised images
>>>>>>>>>> and
>>>>>>>> mask
>>>>>>>>>> to
>>>>>>>>>> a 2mm resolution would impair the statistical analyses? Or is it
>>>>>>>>>> a
>>>>>>>>>> feasable approach?
>>>>>>>>>> 
>>>>>>>>>> Thank you for your help!
>>>>>>>>>> 
>>>>>>>>>> Angela
>>>>>>>>> 
>>>>>>>> 
>>>>>>> 
>>>>> 
>>> 
>>> 
>> 
>> 
>>