Stochastic output

[silvia-oxc]

Hi, I was checking some tool based on greedy and I encountered that if I repeat 2 times the registration of the same image (brain MRI) the output is pretty similar but not exactly the same, i.e. if you subtract the 2 output the sum is not zero. This may generate some variation in the determination of the boundaries (and volumes) of regions due to variability of pixel intensities on the borders. Is this an intrinsic characteristic of the greedy tool or it can be an issue of the version we have? There is any way to remove this random factor by fixing a seed somewhere and to still guarantee that the registration performance is good?

Thank you a lot in advance for your help!
best regards
Silvia

[roohy466]

Facing the same Issue, Is there any way to fixed the random seed for replication reason?

[pyushkevich]

Hi This is something I've been meaning to fix, there needs to be an option added to provide a random seed. Will put it on my to do list, but if anyone would like to make a pull request they are most welcome! Paul

…

On Wed, Mar 19, 2025 at 5:09 AM Rouhi ***@***.***> wrote: Facing the same Issue, Is there any way to fixed the random seed if is there any for replication reason? — Reply to this email directly, view it on GitHub <#51 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAJPEW7O4A5YNZNOHGWIUTT2VEX5TAVCNFSM6AAAAABVVIAGHGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZVHAZDINZWGE> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> [image: roohy466]*roohy466* left a comment (pyushkevich/greedy#51) <#51 (comment)> Facing the same Issue, Is there any way to fixed the random seed if is there any for replication reason? — Reply to this email directly, view it on GitHub <#51 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAJPEW7O4A5YNZNOHGWIUTT2VEX5TAVCNFSM6AAAAABVVIAGHGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZVHAZDINZWGE> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[roohy466]

My question is if it is enough to add the seed "randy(12345)" to any "vnl_random"? or it is more complicated than that?

[pyushkevich]

It seems that there is something else going on, beyond the vnl_random. Looking into it now.

…

On Wed, Mar 19, 2025 at 6:11 AM Rouhi ***@***.***> wrote: My question is if it is enough to add the seed "randy(12345)" to any "vnl_random"? or it is more complicated than that? — Reply to this email directly, view it on GitHub <#51 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAJPEWYDCMGVYD5RTOP3I532VE7GNAVCNFSM6AAAAABVVIAGHGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZWGAZTGNJQGY> . You are receiving this because you commented.Message ID: ***@***.***> [image: roohy466]*roohy466* left a comment (pyushkevich/greedy#51) <#51 (comment)> My question is if it is enough to add the seed "randy(12345)" to any "vnl_random"? or it is more complicated than that? — Reply to this email directly, view it on GitHub <#51 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAJPEWYDCMGVYD5RTOP3I532VE7GNAVCNFSM6AAAAABVVIAGHGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZWGAZTGNJQGY> . You are receiving this because you commented.Message ID: ***@***.***>

[pyushkevich]

This is not completed because I need to propagate the '-seed' parameter to all the code that uses the random number generator, but here is a start: 0b590d0 This code should give the same output for the same input, regardless of the number of threads on a machine. Paul. On Wed, Mar 19, 2025 at 12:26 PM Paul Yushkevich ***@***.***> wrote:

…

It seems that there is something else going on, beyond the vnl_random. Looking into it now. On Wed, Mar 19, 2025 at 6:11 AM Rouhi ***@***.***> wrote: > My question is if it is enough to add the seed "randy(12345)" to any > "vnl_random"? or it is more complicated than that? > > — > Reply to this email directly, view it on GitHub > <#51 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAJPEWYDCMGVYD5RTOP3I532VE7GNAVCNFSM6AAAAABVVIAGHGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZWGAZTGNJQGY> > . > You are receiving this because you commented.Message ID: > ***@***.***> > [image: roohy466]*roohy466* left a comment (pyushkevich/greedy#51) > <#51 (comment)> > > My question is if it is enough to add the seed "randy(12345)" to any > "vnl_random"? or it is more complicated than that? > > — > Reply to this email directly, view it on GitHub > <#51 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAJPEWYDCMGVYD5RTOP3I532VE7GNAVCNFSM6AAAAABVVIAGHGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZWGAZTGNJQGY> > . > You are receiving this because you commented.Message ID: > ***@***.***> >

[roohy466]

Wow! That is quick! I ll try to compile it as I used the ITK 4.0.1. Thanks

[roohy466]

worked like a charm! Thank you very much

greedy -d 3 -a -i source.nii base.nii -ia-image-centers -dof 6 -o test.mat -seed 1000282992 -threads 1

run 1:

  Level   0  Final RAS Transform:
   0.9975  -0.0704  -0.0011  -5.6828
   0.0704   0.9964   0.0482  18.6293
  -0.0023  -0.0482   0.9988  -3.2946
   0.0000   0.0000   0.0000   1.0000

run 2:

 Level   0  Final RAS Transform:
 0.9975  -0.0704  -0.0011  -5.6828
 0.0704   0.9964   0.0482  18.6293
-0.0023  -0.0482   0.9988  -3.2946
 0.0000   0.0000   0.0000   1.0000

[roohy466]

OK! I had many issues installing Greedy from the source but this is the solution for Ubuntu22.04

apt-get install -y --no-install-recommends \
    libinsighttoolkit5-dev \
    libcamitk5 \
    libvtkgdcm-cil \
    fftw3 \
    fftw3-dev \
    python3-vtkgdcm \
    libvtkgdcm-java \
    libgdcm-tools \
    pkg-config && \
    cd /tmp  && \ 
    #------------------------------------------------------------------------------------------- 
    # installing VTK from the source
    git clone --recursive https://gitlab.kitware.com/vtk/vtk.git && \
    mkdir vtk_build && \
    cd vtk_build && \
    cmake ../vtk && \
    make -j$(nproc) && \
    make install && \
    export VTK_DIR=/tmp/vtk_build && \
    #------------------------------------------------------------------------------------------- 
    #copying the VTK libraries from /usr/local/lib to /usr/lib
    ln -s /usr/local/lib/libvtk* /usr/lib && \ 
    #------------------------------------------------------------------------------------------- 
    # installing Greedy from the source
    cd /tmp  && \ 
    git clone https://github.com/pyushkevich/greedy && \
    mkdir greedy_build && \
    cd greedy_build && \
    cmake ../greedy && \
    make -j$(nproc) && \
    make install