also...
the matrix doesn't actually have to be symmetric. consider trains: most people leaving long island are going into new york city (perhaps 3500 out of 5000) but only a small fraction of those leaving new york are going to long island (perhaps 100/5000).
-dh
On Tue, 2 Sep 2014, David R. Haynor wrote:
> hi jiyang,
>
> i don't know about the command, but most of the asymmetry in the matrix you
> included is just poisson noise. to see this, compute M = (a(i,j) + a(j,
> i)/2), and then look at the range [M - 2*sqrt(M), M + 2*sqrt(M)]. most of
> the time you'll see that a(i,j) and a(j,i) are both in that range. the rest
> of the asymmetry probably arises from the fact that the random walk may not
> be perfectly reversible (roundoff, discretization, etc.), although it's
> close.
>
> -dh
>
> On Wed, 3 Sep 2014, Jiyang Jiang wrote:
>
>> Dear FSL list,
>>
>> We aim to generate n-by-n connectivity matrix using probtrackx2 (build
>> 507), where n is the number of ROIs segmented using AAL atlas. We came
>> across some problems in understanding the parameters and results. Your help
>> is very much appreciated!
>>
>> Currently, we are running the following command:
>>
>>
>> probtrackx2 --network \
>> -x ~/ptx2_nw/seedlist.txt \
>> -l \
>> --onewaycondition \
>> -c 0.2 \
>> -S 2000 \
>> --steplength=0.5 \
>> -P 300 \
>> --fibthresh=0.01 \
>> --distthresh=0.0 \
>> --sampvox=0.0 \
>> --forcedir \
>> --opd \
>> -s ~/ptx2_nw/11004.bedpostX/merged \
>> -m ~/ptx2_nw/11004.bedpostX/nodif_brain_mask \
>> --dir=~/ptx2_nw/probtrackx2_try6
>>
>>
>> Our questions are:
>>
>> 1. Is this command correct for generating our ROI*ROI matrix?
>> 2. We have tried to run the command a). with --omatrix1, b). with
>> --targetmasks=~/ptx2_nw/seedlist.txt, and c). with neither of these two
>> flags. The outcomes (fdt_network_matrix) were the same. In Saad’s response
>> to one of the previous enquiry
>> (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1202&L=fsl&P=R65946&1=fsl&9=A&I=-3&J=on&X=A9C4254AE89C6902A5&Y=neuroimageemaillist%40gmail.com&d=No+Match%3BMatch%3BMatches&z=4),
>> if I understand it correctly, --targetmasks seems to record any streamline
>> arriving at the first target, regardless whether the streamline continues
>> to reach other ROI’s (track no further), whereas for --omatrix1 will record
>> the real destination of a fiber tract. For example, in the situation of a
>> fiber tract from A, passing through B, to C (A → B → C), --targetmasks
>> will record this tract as A → B, whereas --omatrix1 will record A→C. If
>> that is the case, I suppose the results of using --omatrix1 and
>> --targetmasks should be different, unless there is not any streamline from
>> the seed ROI to the target ROI passing through any of the other ROI’s. Is
>> that correct?
>> 3. In the same thread mentioned above
>> (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1202&L=fsl&P=R65946&1=fsl&9=A&I=-3&J=on&X=A9C4254AE89C6902A5&Y=neuroimageemaillist%40gmail.com&d=No+Match%3BMatch%3BMatches&z=4),
>> it seems --network will do the same thing as --omatix1, and --stop has the
>> same effects as --targetmasks. Can I use --network and --targetmasks at the
>> same time to consider each mask as a seed in turn and all other masks as
>> stop masks?
>> 4. Why the fdt_network_matrix is not symmetric (see below)?
>> 5. Does the number in the matrix below represent the number of
>> probabilistic streamlines in consideration of the probability of diffusion
>> distribution at each voxel on the route? If that is the case, I suppose at
>> lease some decimals are expected?
>>
>>
>> 0 2297 52 198 129 9485 1752 76
>> 3821 0 102 78 83 1293 4746 189
>> 32 43 0 4 4435 393 406 1013
>> 232 51 5 0 241 3649 461 34
>> 95 76 4380 383 0 1022 1704 1102
>> 7286 821 231 4542 619 0 4849 238
>> 1757 3395 959 394 1376 6632 0 767
>> 93 228 2947 74 1612 953 614 0
>>
>> Sorry for so many questions, and thank you very much for your time!
>>
>> Kind regards,
>> Jiyang
>>
>
|