NITRC Spatially Constrained Parcellation Forum: open-discussion

How to find the centroids of the clusters?

Sara Eb — Sat, 12 Sep 2020 2:52:30 GMT

Hi I have knee structural MRIs with the ground truth label files (of different compartments) and the mask created from the template (constructed from 20 subjects). Is it possible to use this tool for structural MRIs for the spatial parcellation?
I have attached the knee MRI, label file and the mask

2) my second question is about the code, I ran the test example code with the test subjects, Is it possible to find the clusters' centroids in this technique? If yes, where can I find the centroids?

3) How to create the parcellation within all subjects? I have a template and 243 test subjects. Once we create the parcellation based on template, how the same parcellation can be applied on test subjects?

Your expert opinion is really appreciated

Sara

RE: error running test script

Rebecca Rebi — Fri, 22 May 2020 14:20:08 GMT

I ran into the same issue. I found that the original code was made with an older version of numpy than the one I have. The solution was to replace "rank" with "ndim".
https://kite.com/python/docs/numpy.rank

error running test script

skyflyer fun — Thu, 16 Apr 2020 23:49:25 GMT

Hi, I am getting the following error when running "pyClusterROI_test.py",
NameError: name 'rank' is not defined
from line 70 of "make_local_connectivity_ones.py"
if(rank(idx) == 1):

I'm running nmpy 1.18.1

thank you in advance for any help or suggestions

Err running tcorr and scorr

Steve Da — Thu, 08 Jun 2017 3:04:15 GMT

Hi all,
I don't have any problem to run the sample data. However when I started using my own rs fmri data, for some voxels, I have the following errors and cannot successfully finish the parcellation.

[b]For scorr[/b]

Traceback (most recent call last):
File "scorr_test.py", line 115, in
make_local_connectivity_scorr( in_file, maskname, outname, 0.5 )
File "/data/fnl/Computing/xda/Projects/DepInf/craddock_2011/cluster_roi_test_rs/make_local_connectivity_scorr.py", line 190, in make_local_connectivity_scorr
R[isnan(R)]=0
TypeError: 'numpy.float64' object does not support item assignment

[b]For tcorr[/b]
Traceback (most recent call last):
File "tcorr_test.py", line 114, in
make_local_connectivity_tcorr( in_file, maskname, outname, 0.5 )
File "/data/fnl/Computing/xda/Projects/DepInf/craddock_2011/cluster_roi_test_rs/make_local_connectivity_tcorr.py", line 176, in make_local_connectivity_tcorr
R=R[nndx,:].flatten()
IndexError: invalid index to scalar variable.

Do you have any idea to solve these problems?

Thanks and best regards,

Steve Da

Using FSL for cluster labelling via PyCluster

S Mody — Sun, 30 Aug 2015 18:43:14 GMT

Hello all,

1. Has anybody used PyCluster for automatic labelling as described on the PyCluster "Usage" page? :
[url=http://ccraddock.github.io/cluster_roi/usage.html]http://ccraddock.github.io/cluster_roi/usage.html[/url]
See the bottom of the page.

2. The auto labelling requires the FMRIB, FSL software to be installed. Has anyone installed FSL on openSUSE?

FSL is a huge download (1.7 GB). Moreover, It is not clear from the FMRIB FSL page whether it can be installed on Linux distributions other than Debian or CentOS.
I'm using openSUSE and I don't want to go to the trouble of downloading all of it and then finding out that it doesn't work on my distribution.
Installing on windows requires a installation of a virtual environment - so that is even less convenient.

Is there some smaller part of FSL that can be downloaded just for the purpose of cluster labelling.

Thank you,
Regards
SK Mody

RE: number of clusters

Daniel Lurie — Mon, 17 Aug 2015 9:00:47 GMT

Hi again,

This script should work until Cameron releases an official fix: https://github.com/danlurie/cluster_roi/blob/master/nifti_gen_fix.py

The problem was essentially an overflow error; the data type from the mask image (to which the parcellation values are written) is uint8 by default, which can't handle values over 255. My fix converts everything to float64, which is overkill, but works.

Let me know if there are any issues.

Dan

RE: number of clusters

Daniel Lurie — Mon, 17 Aug 2015 7:49:35 GMT

Hi all,

I'm also experiencing this issue (212 subjects, 4mm voxels, HO 25% GM mask, group average parcellation).

It looks like the actual parcel data (in the .npy files) contains the right number of parcels. You can check this on your own results with the following python code:[code]import numpy as np[/code][code]parcel_data = np.load('/path/to/parcellation_k.npy')[/code][code]np.unique(parcel_data, return_counts=True)[/code]
This will return two arrays of equal length. The first array shows every unique value in the data, so if your k=100, it should have *around* 100 entries (as Cameron mentions in the paper, NCUT can sometimes produce clustering with empty clusters, so your actual k might end up somewhat smaller than your desired k). The second array is the number of voxels with each value at the same array index in the first array.

I think the issue is with the scripts that convert the NumPy parcellation data into a NIfTI volume. Working on a fix now, will report back if/when I'm successful.

Dan

RE: number of clusters

S Mody — Sat, 15 Aug 2015 10:46:10 GMT

@Shenwei, perhaps the reason you get so few clusters is that the correlation matrix is so sparse that only 255 of the eigenvectors are numerically significant. Did you consider this possibility?

Regards,
Sandeep Mody.

RE: number of clusters

S Mody — Sat, 15 Aug 2015 10:32:23 GMT

[i]Originally posted by Adam Steel:[/i][quote]Hi Shangwei -

I'm having the same issue at the moment. Was this issue ever resolved?

Thanks

Adam[/quote]
[color=#000000]It appears that (at least some of the) code may not complete in a practical amount of time for higher resolutions. I used preprocessed fMRI data from the Human Connectome Project which has a voxel resolution of (2 x 2 x 2) mm. I used a cluster count of 1000. The function make_local_connectivity_tcorr() with a grey mask having about 255,000 voxels runs successfully to create the correlation matrix. The correlation matrix has around 1,000,000 non-zero entries. However the next step binfile_parcellate() runs forever - as the eigenvalue decomposition of the Laplacian of the correlation matrix does not converge (after 24+ hours of run time).[/color]

There is a comment in the ncut() function of the toolbox about regularizing the matrix for better stability. So it appears that the author himself may have had problems with the eigenvalue decomposition - which is of course not at all unexpected. Trying to find the eigenvectors of a large sparse matrix can be hell. Hopefully the author reads this and comments.

[color=#000000]My next step is to resample the data and grey-mask to (4 x 4 x 4) and rerun the code.[/color]

[color=#000000]Wondering whether anyone else has had a similar experience.[/color]

[color=#000000]Regards,[/color]
[color=#000000]Sandeep Mody.
[/color]

RE: number of clusters

Shengwei Zhang — Thu, 13 Aug 2015 21:21:28 GMT

Unfortunately not, Adam.

Shengwei