sdm-help-list > failure to execute "combine image"
Showing 1-6 of 6 posts
Jul 11, 2017 08:07 PM | Lars Schulze
failure to execute "combine image"
Dear Joaquim,
thank you very much for providing and developing SDM.
I noticed there is an option to "combine images" from the same sample of participants. Unfortunately, this option will not execute with my version of SDM (5.141). I am able to input the images and the correlation (I tried different values here), and I get the notification regarding the variance reduction file, but the moment SDM should start to calculate nothing is happening. But SDM does not freeze. I just does not execute this command. Hopefully, you have an idea what went wrong.
However, this provides me the chance to ask some additional questions regarding the correlational calculations. I tried to combine contrasts from two different tasks but assessed in the same sample. Since I do not have access to the raw data, I extracted the respective coordinates and effect sizes. The first contrast comprised the left superior frontal gyrus only. The second contrast comprises a region in the left cerebellum and the left inferior frontal gyrus. Thus, there is only minimal overlap between both contrasts. Afterwards I calculated the preprocessing of these files in a separate analysis. Then, I imported the t-maps using Matlab and calculated the correlation between both maps. Should one use all elements of these matrices (r=-.0029) or should one use only the elements that are non-zero (r=-.6605)? In my case, the overwhelming majority of the map are Zeros (more than 99%).
As stated above, the correlation coefficients are negative in my case. Is it at all adequate to combine these images? And finally, you state that there might be a warning "if the expected increase in effect size is more than 100%". What would you recommend to do in these situations?
Thank you very much in advance,
Lars
thank you very much for providing and developing SDM.
I noticed there is an option to "combine images" from the same sample of participants. Unfortunately, this option will not execute with my version of SDM (5.141). I am able to input the images and the correlation (I tried different values here), and I get the notification regarding the variance reduction file, but the moment SDM should start to calculate nothing is happening. But SDM does not freeze. I just does not execute this command. Hopefully, you have an idea what went wrong.
However, this provides me the chance to ask some additional questions regarding the correlational calculations. I tried to combine contrasts from two different tasks but assessed in the same sample. Since I do not have access to the raw data, I extracted the respective coordinates and effect sizes. The first contrast comprised the left superior frontal gyrus only. The second contrast comprises a region in the left cerebellum and the left inferior frontal gyrus. Thus, there is only minimal overlap between both contrasts. Afterwards I calculated the preprocessing of these files in a separate analysis. Then, I imported the t-maps using Matlab and calculated the correlation between both maps. Should one use all elements of these matrices (r=-.0029) or should one use only the elements that are non-zero (r=-.6605)? In my case, the overwhelming majority of the map are Zeros (more than 99%).
As stated above, the correlation coefficients are negative in my case. Is it at all adequate to combine these images? And finally, you state that there might be a warning "if the expected increase in effect size is more than 100%". What would you recommend to do in these situations?
Thank you very much in advance,
Lars
Jul 24, 2017 10:07 AM | Joaquim Radua
RE: failure to execute "combine image"
Dear Lars,
thank you for this message.
The correlation used in these estimations refers to the correlation between individual values in one task and individual values in another task within one voxel. Thus, the correlation is for one voxel, rather than accross voxels. That said, for simplicity and due to the lack of data, the correlation coefficient is assumed to be the same in all voxels, and that's why a single correlation coefficient is introduced.
Regarding the bug, can you send us the "execution log"? (go to Tools -> View Execution Log)
Hope this helps, and thank you again,
Joaquim
thank you for this message.
The correlation used in these estimations refers to the correlation between individual values in one task and individual values in another task within one voxel. Thus, the correlation is for one voxel, rather than accross voxels. That said, for simplicity and due to the lack of data, the correlation coefficient is assumed to be the same in all voxels, and that's why a single correlation coefficient is introduced.
Regarding the bug, can you send us the "execution log"? (go to Tools -> View Execution Log)
Hope this helps, and thank you again,
Joaquim
Jul 24, 2017 10:07 AM | Lars Schulze
RE: failure to execute "combine image"
Dear Joaquim,
thank you very much for your answer and the clarification of the correlational procedure.
With regard to the bug: The execution log is empty. This seems to be in line with my perception that the command is simply not executed. I did attach attach you an image of the input commands (see figure 1).
Best regards,
Lars
thank you very much for your answer and the clarification of the correlational procedure.
With regard to the bug: The execution log is empty. This seems to be in line with my perception that the command is simply not executed. I did attach attach you an image of the input commands (see figure 1).
Best regards,
Lars
Jul 24, 2017 12:07 PM | Anton Albajes-Eizagirre - FIDMAG - Germanes Hospitalaries
RE: failure to execute "combine image"
Dear Lars,
I couldn't test this on windows yet, but it may be possible that the comas and spaces in the paths are the cause of the issue (together with an incapacity of SDM to deal with them).
Could you try to run the same command but with the data in a folder without comas or spaces?
Thanks,
Anton
Originally posted by Lars Schulze:
With regard to the bug: The execution log is empty. This seems to be in line with my perception that the command is simply not executed. I did attach attach you an image of the input commands (see figure 1).
I couldn't test this on windows yet, but it may be possible that the comas and spaces in the paths are the cause of the issue (together with an incapacity of SDM to deal with them).
Could you try to run the same command but with the data in a folder without comas or spaces?
Thanks,
Anton
Originally posted by Lars Schulze:
With regard to the bug: The execution log is empty. This seems to be in line with my perception that the command is simply not executed. I did attach attach you an image of the input commands (see figure 1).
Jul 24, 2017 12:07 PM | Lars Schulze
RE: failure to execute "combine image"
Dear Anton,
I did move the images to a different folder (without spaces and commas).
Launching Images Combination...
Launched: sdm MyCombination = combine 0.3, C:/Users/Lars/Desktop/test/b_preproc/pp_HC_3back_Landre_2012.nii.gz, C:/Users/Lars/Desktop/test/b_preproc/pp_HC_ID_Landre_2012.nii.gz
Unfortunately, "combine images" still doesn't execute. And the "execution log" is still empty.
Thanks for your help,
Lars
I did move the images to a different folder (without spaces and commas).
Launching Images Combination...
Launched: sdm MyCombination = combine 0.3, C:/Users/Lars/Desktop/test/b_preproc/pp_HC_3back_Landre_2012.nii.gz, C:/Users/Lars/Desktop/test/b_preproc/pp_HC_ID_Landre_2012.nii.gz
Unfortunately, "combine images" still doesn't execute. And the "execution log" is still empty.
Thanks for your help,
Lars
Aug 17, 2017 11:08 AM | Anton Albajes-Eizagirre - FIDMAG - Germanes Hospitalaries
RE: failure to execute "combine image"
Dear Lars,
There is a bug on the core binaries of the Windows version. Attached I am posting a zip file with the three binaries that should replace the binaries of your sdm package (folder bin/windows64 of the Sdm package). These binaries are for the 64 bits version of Sdm. Please let me know if you need the 32 bits versions.
Originally posted by Lars Schulze:
Unfortunately, "combine images" still doesn't execute. And the "execution log" is still empty.
Best,
Anton
There is a bug on the core binaries of the Windows version. Attached I am posting a zip file with the three binaries that should replace the binaries of your sdm package (folder bin/windows64 of the Sdm package). These binaries are for the 64 bits version of Sdm. Please let me know if you need the 32 bits versions.
Originally posted by Lars Schulze:
Unfortunately, "combine images" still doesn't execute. And the "execution log" is still empty.
Best,
Anton