Print

Print


Hi Walid,

Please see below:

On Tuesday, 1 August 2017, Walid Yassin <[log in to unmask]> wrote:

> Hello Anderson,
>
>
> Thank you very much for your answers. Please see my replies and
> clarifications below.
>
>
> On 2017 Aug 1, at 22:51, Anderson M. Winkler <[log in to unmask]>
> wrote:
>
> Hi Walid,
>
> Please see below:
>
>
> On 31 July 2017 at 04:20, Walid Yassin <[log in to unmask]> wrote:
>
>> Hello,
>>
>> I would like to report the coordinates of the regions found in my
>> clusters.
>>
>>
>> This is what I did so far.
>>
>> 1- TBSS, followed by randomize (All is well till here).
>> 2- I got my clusters using the autoaq command and the results are 4
>> clusters. Each cluster has several regions in it.
>> I have the voxel# of the clusters and the coordinates of the clusters.
>> I want to get the coordinate and voxel number of each of the sub clusters.
>> So to get the coordinates I wrote the cluster command with --olmax.
>> Now to name the clusters, I went to atlasquery and used -c for each of
>> the coordinates.
>> Until now I have all the sub cluster coordinates and names, however I
>> don’t know the cluster size of each sub cluster region.
>>
>> Can someone help me know the cluster size of each of the sub cluster
>> region?
>>
>
> Could you describe what is meant by subcluster? I'm not sure I follow...
> Each peak (local maximum) within a cluster doesn't define a new cluster
> within that cluster. Same for regions spanned by the cluster, but maybe I'm
> not following...
>
>
> All I’m asking here is about the size of the regions inside each cluster.
> I wanted to know how many voxels are possibly found inside of each of those
> reigns.
>
> Example:
>
> Cluster 3  (2000voxcels)
> -SLF (how many?)
> -ILF (how many?)
>
>
> Sorry if the use of “subcluster” is confusing..
>
>
The way as atlasquery works is by using the probabilities that a voxel
belongs to a region of the atlas. To obtain what you want perhaps the
simplest is to go to the atlases' directory, i.e., ${FSLDIR}/data/atlases/,
then use fslmaths to define a mask for each of the regions of the atlas,
then use fslstats to see the volume of the intersection.


>
>
>> Also, I have another question.
>> Since in one of my clusters, for e.g., I saw that the superior
>> longitudinal fasciculus (SLF), exists twice. Each having separate
>> coordinates.
>> Also the SLF not only exists twice in the same cluster, but also is a sub
>> region in other clusters
>>
>> Example
>>
>> Cluster 1
>>
>> 1-SLF                   coordinates
>> 2- Some region  coordinates
>> 3- SLF                  coordinates
>>
>>
>> Cluster 2
>>
>> 1- Some region
>> 2- SLF
>> 3-Some regions  coordinates
>>
>
> Could you copy/paste the actual output from the command?
>
>
> Yes! Here it is..
>
> ClusterIndex Voxels MAX MAX X (mm) MAX Y (mm) MAX Z (mm) COG X (mm) COG Y
> (mm) COG Z (mm)
> 4 4689 0.97 45 -41 32 37.2 -20.4 32.8
> 3 924 0.96 15 33 0 20.2 36 15.2
> 2 345 0.956 51 -48 5 48.9 -46 -2.57
> 1 18 0.951 46 -10 26 46.2 -10.3 26.8
> ------------------------------------------
> Structures to which each center of mass belongs to:
> 4,37.2,-20.4,32.8,JHU White-Matter Tractography Atlas,79% Superior
> longitudinal fasciculus R, 42% Superior longitudinal fasciculus (temporal
> part) R
> 3,20.2,36,15.2,JHU White-Matter Tractography Atlas,39% Forceps minor, 18%
> Anterior thalamic radiation R
> 2,48.9,-46,-2.57,JHU White-Matter Tractography Atlas,24% Superior
> longitudinal fasciculus R, 23% Superior longitudinal fasciculus (temporal
> part) R
> 1,46.2,-10.3,26.8,JHU White-Matter Tractography Atlas,26% Superior
> longitudinal fasciculus R, 6% Superior longitudinal fasciculus (temporal
> part) R
> ------------------------------------------
> Structures to which each cluster belongs to:
>
> Cluster #4
> Anterior thalamic radiation R:0.0230326
> Corticospinal tract R:1.8859
> Cingulum (cingulate gyrus) R:0.00703775
> Inferior fronto-occipital fasciculus R:0.319898
> Inferior longitudinal fasciculus R:1.39966
> Superior longitudinal fasciculus R:8.39241
> Superior longitudinal fasciculus (temporal part) R:2.46534
>
> Cluster #3
> Anterior thalamic radiation R:10.7403
> Cingulum (cingulate gyrus) R:0.0844156
> Forceps minor:25.2511
> Inferior fronto-occipital fasciculus R:6.5368
> Superior longitudinal fasciculus R:0.0974026
> Uncinate fasciculus L:0.012987
> Uncinate fasciculus R:1
> Superior longitudinal fasciculus (temporal part) R:0.0162338
>
> Cluster #2
> Inferior fronto-occipital fasciculus R:0.00869565
> Inferior longitudinal fasciculus R:1.24348
> Superior longitudinal fasciculus R:7.90435
> Superior longitudinal fasciculus (temporal part) R:8.37681
>
> Cluster #1
> Superior longitudinal fasciculus R:31
> Superior longitudinal fasciculus (temporal part) R:7.55556
>
>
>
The above is fine. The SLF is large and multiple clusters end up
overlapping with it.


>
>
>>
>>
>> Is it proper to make a ROI (SLF) and try to check how many significant
>> voxels are in that ROI instead of what I did above?
>> If that is okay, how can I do that?
>>
>> E.g. SLF is the ROI
>>
>> To get the # voxels in ROI
>>
>> -Go to FSL view, load the JHU tact atlas SLF, save it as make and
>> binarize it, then how do I get the # of voxels inside that ROI?
>>
>
>
> Yes, it's possible but then how could we know whether the result is
> significant? That is, whether however many voxels are found within the ROI
> is something worth reporting or not? Instead, if the objective is to
> investigate only the SLF, consider supplying it as a mask to randomise.
>
>
> Maybe I didn’t clarify enough what I’m asking.
> I meant when I already got the results of randomize between the 2 groups
> and I found some significant voxels and made a sig. voxel mask. If I
> overlay the SLF ROI from an atlas over the significant voxel mask, can I
> know how many voxels that are in the sig. voxel mask are also in the ROI?
> That way I could know how many sig. voxels are in the SLF (ROI)… I'm not
> sure if this is possible which is why I’m asking (The objective is not only
> SLF).
>

Yes, you should be able to do as above. However, doing so is still a
problem, because this number -- voxel count within the ROI -- is on its own
right a test statistic, which should be interpreted with some measure of
uncertainty, which on its turn is given by the p-value and that you'd
obtain from randomise, using the ROI as mask.

Hope this helps!

All the best,

Anderson



>
>
>
>>
>> For coordinates of this ROI, should I just report the local max of SLF
>> itself? If not, how do I get that as well
>>
>
> We should report what is sufficient to characterise the effect.
> Coordinates alone probably aren't sufficient, but are useful in
> meta-analyses. The full region is probably more useful.
>
>
>
> Thank you very much for all the above..
>
> Best,
>
> WY
>
>
> All the best,
>
> Anderson
>
>
>
>>
>> Sorry for asking several Qs..
>>
>>
>> Thank you in advance,
>>
>> WY
>
>
>