Hi Saad, many thanks for the clarification - I see now why my
suggestion makes little sense.
Regarding my study - I tried the termination masks but realized this
is not what I want to do. So here is what I want to do:
I have Region A, B and C. I want to show connections between
A and B,
A and C,
B and C (which I hope are not there according to the hypothesis)
I'd like to also compare the connection strengths/probabilities for
these connections in a number of subjects.
Network-mode will give me Connections that pass through the regions,
but I want them to end there.
Termination masks will give me all connections from one region,
including those that terminate in the target region - I want only
those that terminate.
Waypoint masks again will give connections that pass through.
Exclusion masks is not what I want.
So I'm a bit lost and start to wonder if that's what I want is a
reasonable thing to do, or not? If it is reasonable, which I still
hope - how would you suggest to do that?
Thanks, Michael
On 24-May-09, at 12:36 PM, Saad Jbabdi wrote:
> Hi Michael,
>
> On 24 May 2009, at 19:07, Michael Scheel wrote:
>
>> Hi Saad, thanks for the answer. Some questions just to make sure
>> that I understood correctly.
>>
>> I've often come across the symmetric mode (---mask2) in the
>> discussions and I was assuming that this is something like:
>>
>> Track from RegionA to RegionB AND from RegionB to RegionA - only
>> those tracts survive that fullfill both conditions. From your answer
>> I understand that they don't necessarily have to take the same way.
>> However - wouldn't that be a strong confirmation of the certainty of
>> the tract travelling this route? I guess depending on seed and
>> target chances are sometimes low that one sample from AtoB has a
>> counterpart BtoA and one would need a huge sample number to find
>> those. One would have to store for every sample of seed region A the
>> "voxelway" and then compare this with the table of voxelways from
>> regionB. I guess the problem is how to visualize the results of this
>> - hower just an idea.
>
>
> There is no theoretical reason for accepting only sample tracts that
> take the exact same route both ways. Remember: in probtrackx, we are
> building a (spatial) histogram of connections using sampling. Each
> sample is a random draw from that distribution. So, even if the
> distribution that represents A->B is the exact same distribution as B-
>> A, there is no reason why the samples should be the same..
>
>>
>>
>> My question really is for the symmetric mode. Does this show (as in
>> network-mode) connections from A that pass through B AND connections
>> from B that pass through A. Or is it connections from A that end in
>> B AND from B that end in A.
>
> Network mode is exactly the same thing as symmetric mode, but extended
> to more than two regions. I.e. if you use two regions in network mode,
> you get the same answer as two masks symmetric.
> The samples that are kept are the ones that pass through the target
> regions (as opposed to "end in").
>
>
>>
>>
>> Another question re the network mode:
>> You said:
>>> It will show you an overlay of the following three distributions:
>>> . connections from A that pass through either B or C
>>> . connections from B that pass through either A or C
>>> . connections from C that pass through either A or B
>>
>> Is there an option of tracking:
>> . connections from A that end in either B or C
>> . connections from B that end in either A or C
>> . connections from C that end in either A or B
>> I guess I could use the target mode for every combination and
>> overlay the results - correct? Is there already an option or script
>> to automate this?
>
> You are probably thinking of using termination masks. Unfortunately,
> you cannot use the seeds as terminations in network mode. You will
> need to run probtrackx several times.
>
>>
>>
>>
>> Thanks a lot, Michael
>>
>
> Cheers,
> Saad.
>
>>
>> On 24-May-09, at 2:45 AM, Saad Jbabdi wrote:
>>
>>> Hi Michael,
>>>
>>> On 23 May 2009, at 21:56, Michael Scheel wrote:
>>>
>>>> Hi fsl-gurus,
>>>>
>>>> I have 2 questions re the probtrackx network mode:
>>>>
>>>> Say I have a Region A, B and C defined as masks and enter them as
>>>> Multiple seed masks.
>>>> What connections will be shown to me:
>>>> 1) Any connections between A and B, A and C, B and C or 2) just
>>>> connections that connect all of them, i.e. A-B-C, A-C-B and B-A-C.
>>>
>>> neither exactly 1) nor exactly 2).
>>> It will show you an overlay of the following three distributions:
>>> . connections from A that pass through either B or C
>>> . connections from B that pass through either A or C
>>> . connections from C that pass through either A or B
>>>
>>>
>>>>
>>>>
>>>> Assuming it is (1), and looking only at connections between A and
>>>> B.
>>>> As I understand from previous discussions, tracking samples are
>>>> send
>>>> from A to B and B to A and only those that reach their target in
>>>> both conditions 'survive'.
>>>> What I'm wondering about is what are the constraints regarding
>>>> that:
>>>> Does the sample has to go the exact same voxelway to be found valid
>>>> or is it rather an overlay where voxels survive that were passed
>>>> from either direction, i.e. at least one sample passed from AtoB
>>>> AND
>>>> BtoA.
>>>
>>> It is the latter. A tract survives if it reaches its target,
>>> regardless of the route it has taken.
>>>
>>> Cheers,
>>> Saad.
>>>
>>>
>>>>
>>>>
>>>> Thanks, Michael
>>>>
>>>
>>> --
>>> Saad Jbabdi
>>> Oxford University FMRIB Centre
>>>
>>> JR Hospital, Headington, OX3 9DU, UK
>>> +44 (0) 1865 222545 (fax 717)
>>> www.fmrib.ox.ac.uk/~saad
>>
>
> --
> Saad Jbabdi
> Oxford University FMRIB Centre
>
> JR Hospital, Headington, OX3 9DU, UK
> +44 (0) 1865 222545 (fax 717)
> www.fmrib.ox.ac.uk/~saad
|