I don't think using the max value of the fdt_paths is a good idea. That
will vary depending on how closely packed the samples are at the narrowest
point of the tract. For example if you have a 100000 sample pathway that at
its narrowest point goes through a single voxel, that voxel's value would be
100000, and that would be the max of the fdt_paths. If you had a separate
100000 sample pathway that at its narrowest point was divided evenly among 4
voxels, the maximum value of the fdt_paths would be 25000. Normalizing
based on this number would give you very different probability values across
the entire pathway, even if the two were otherwise identical.
Peace,
Matt.
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Cherif Sahyoun
Sent: Monday, December 01, 2008 4:10 PM
To: [log in to unmask]
Subject: Re: [FSL] question on old and new waytotals
Hi Saad,
Can you talk about the implications of using something other than the
waytotal for normalizing? I'll get you started :)
- using waytotal would give the conditional probability of going
through a given voxel, given that there is a path.
- using the ROI_size*samples would give the absolute probability of
going through a voxel (on the path)
What to you think of using the max value of the fdt_paths? That should
give a normalized conditional probability similar to using waytotal
(though obviously we important differences since now we are forcing
the most likely voxels to have a probability of 1, which was not the
case using waytotal)...
If what one wants is just to normalize across subjects to be able to
compare mean p of a tract, I guess that would work?
Best,
Cherif
----------------------------------------------------------------------------
--------------
Cherif P. Sahyoun HST-MEMP
Developmental Neuroimaging of Cognitive Functions
C: 617 688 8048
H: 617 424 6956
[log in to unmask]
"Live as if this were your last day. Learn as if you'll live forever"
Gandhi
----------------------------------------------------------------------------
---------------
On Mon, Dec 1, 2008 at 4:54 PM, Saad Jbabdi <[log in to unmask]> wrote:
> Hi Markus,
> When you say 50 papers, you must be thinking of the seed_to_target values,
> not waytotal, am I right? The output of seed_to_target is NOT
> underestimated, it is only waytotal.
> the waytotal file is a recent output in probtrackx, so I would be
surprised
> if there were 50 articles published that are using it!
> To answer your question, I think it is quite hard to predict the behaviour
> of waytotal as it is now, so I would recommend re-running your analysis
with
> the patch to come if you are planning to use waytotal.
> Cheers,
> Saad.
>
> On 1 Dec 2008, at 17:20, Markus Gschwind wrote:
>
> That is very good news! Thank you so much!
>
> However, I am really curious if there is an officially recommended way of
> dealing with this underestimation. Roughly guessed, there are about 50
> publications using those "old" waytotals and if I contribute another one,
> now that it is known that those values are not always true...
> Should all the people who are still working with FSL 4.0.x really restart
> the whole analyis in FSL 4.1?
>
> Would there be another way?
>
> Many regards,
> Markus
>
>
>
>
>
> 2008/12/1 Saad Jbabdi <[log in to unmask]>
>>
>> Hi All,
>>
>> The patch -should be- available tomorrow :-)
>>
>> Thank you all for pointing this out!
>>
>> Cheers,
>> Saad.
>>
>>
>> On 28 Nov 2008, at 20:23, Martin Kavec wrote:
>>
>>> Thanks a lot Saad,
>>>
>>> at least we helped to point out the problem. Could you please let us
>>> know,
>>> when we could expect the patch?
>>>
>>> Thanks,
>>>
>>> Martin
>>>
>>> On Friday 28 November 2008 19:50:43 Saad Jbabdi wrote:
>>>>
>>>> Hi Markus (and Yan Liu),
>>>>
>>>> I am terribly sorry, but just realised that I haven't included that
>>>> fix to the released FSL!!! You will need to wait for the next patch
>>>> now...
>>>>
>>>> And to answer your question, in principle it should underestimate
>>>> waytotal by 50% on average if you set the option "--randfib".
>>>> Otherwise it is difficult to predict by how much it will underestimate
>>>> it for each data set..
>>>>
>>>> Again, I am sorry for any inconvenience.
>>>>
>>>> Cheers,
>>>> Saad.
>>>
>>>
>>> --
>>> **********************************
>>> Senior Clinical Research Associate
>>> MRI Unit of the Department of Radiology
>>> Erasme Hospital
>>> Lennik Street 808
>>> B-1070 Brussels
>>> BELGIUM
>>>
>>> tel: +32-2-555-4325
>>> fax: +32-2-555-3994
>>> email: [log in to unmask]
>>> **********************************
>>>
>>> -----------------------------------------------
>>> Find a way, or make one!
>>>
>>
>> Saad Jbabdi
>> Oxford University FMRIB Centre
>>
>> JR Hospital, Headington, OX3 9DU, UK
>> +44 (0) 1865 222545 (fax 717)
>> www.fmrib.ox.ac.uk/~saad
>
>
>
> --
> Dr. med. Markus Gschwind, M.D.
> Laboratory for Neurology and Imaging of Cognition
> Dept of Neurosciences
> University Medical Center (CMU)
> 1 Michel-Servet - 1211 GENEVA - CH
>
> Tel 0041 (0) 22 379 5324
> Fax 0041 (0) 22 379 5402
> email: [log in to unmask]
> http://labnic.unige.ch
>
> Saad Jbabdi
> Oxford University FMRIB Centre
> JR Hospital, Headington, OX3 9DU, UK
> +44 (0) 1865 222545 (fax 717)
> www.fmrib.ox.ac.uk/~saad
>
>
>
>
>
>
|