help
help > RE: Loading conn files in SPM
Jan 29, 2015 01:01 AM | Alfonso Nieto-Castanon - Boston University
RE: Loading conn files in SPM
Hi Roger,
If you are already entering your "effect of X" for each condition/task as confounding effects during Denoising, then CONN is already computing those first-level analyses (effect of each task/condition controlling for all other confounding effects) during the Denoising step. If you check on the 'Create confound effect beta-maps' option in Setup.Options CONN will generate the corresponding Beta volumes (in conn_*/preprocessing/BETA_Subject#.nii) containing the estimated task effects for each subject, which you can then enter into SPM for additional second-level analyses. These BETA files will contain all of the estimated effects for each subject as a 4d volume, so you need to identify which ones are the effects corresponding to your tasks/conditions. The corresponding design matrix is block-diagonal, with one block per session, and within each session the effects are ordered as: 1) constant term; 2) all effects entered in the Confounding effects list, in the same order, and each expanded if necessary (e.g. when using higher order derivatives or multiple dimensions); 3) optional additional detrending term(s) (one for linear, two for quadratic, etc.). If you are unsure how to identify the proper terms please send me your conn_*.mat file and I will let you know which volumes within these BETA_Subject*.nii files correspond to your task conditions.
Alternatively, of course, you can also enter the preprocessed/denoised timeseries into SPM to perform these first-level analyses (but in that case please keep in mind that you should not have the "effect of X" entered as confounding effects during Denoising, otherwise that will remove exactly the same effects that you want to later estimate). That would still require you to define those first-level design matrices within SPM, since unfortunately there is no simple way to transfer that information back from CONN (in case it helps, you can find the task-related regressors that CONN uses in the files conn_*/data/COND_Subject#_Session#.mat; variable named 'data').
Hope this helps
Alfonso
Originally posted by Roger Beaty:
If you are already entering your "effect of X" for each condition/task as confounding effects during Denoising, then CONN is already computing those first-level analyses (effect of each task/condition controlling for all other confounding effects) during the Denoising step. If you check on the 'Create confound effect beta-maps' option in Setup.Options CONN will generate the corresponding Beta volumes (in conn_*/preprocessing/BETA_Subject#.nii) containing the estimated task effects for each subject, which you can then enter into SPM for additional second-level analyses. These BETA files will contain all of the estimated effects for each subject as a 4d volume, so you need to identify which ones are the effects corresponding to your tasks/conditions. The corresponding design matrix is block-diagonal, with one block per session, and within each session the effects are ordered as: 1) constant term; 2) all effects entered in the Confounding effects list, in the same order, and each expanded if necessary (e.g. when using higher order derivatives or multiple dimensions); 3) optional additional detrending term(s) (one for linear, two for quadratic, etc.). If you are unsure how to identify the proper terms please send me your conn_*.mat file and I will let you know which volumes within these BETA_Subject*.nii files correspond to your task conditions.
Alternatively, of course, you can also enter the preprocessed/denoised timeseries into SPM to perform these first-level analyses (but in that case please keep in mind that you should not have the "effect of X" entered as confounding effects during Denoising, otherwise that will remove exactly the same effects that you want to later estimate). That would still require you to define those first-level design matrices within SPM, since unfortunately there is no simple way to transfer that information back from CONN (in case it helps, you can find the task-related regressors that CONN uses in the files conn_*/data/COND_Subject#_Session#.mat; variable named 'data').
Hope this helps
Alfonso
Originally posted by Roger Beaty:
Hi
Alfonso,
Yes, I am trying to look at BOLD activation differences between conditions (i.e., standard SPM GLM analysis), not connectivity differences. The goal is basically to compare these univariate results from SPM with some multivariate connectivity results produced by CONN. If possible, I would like to use the preprocessed images from CONN. Any thoughts on how to do this?
Thanks,
Roger
Originally posted by Alfonso Nieto-Castanon:
Yes, I am trying to look at BOLD activation differences between conditions (i.e., standard SPM GLM analysis), not connectivity differences. The goal is basically to compare these univariate results from SPM with some multivariate connectivity results produced by CONN. If possible, I would like to use the preprocessed images from CONN. Any thoughts on how to do this?
Thanks,
Roger
Originally posted by Alfonso Nieto-Castanon:
Hi
Roger,
Sorry but CONN does not create first-level SPM.mat files that you could import into SPM. Could you please expand a bit on what exactly you are trying to do to see if there are some work-arounds that come to mind? (e.g. are you trying to use CONN preprocessed time-series to look at the "activation" -BOLD signal changes; not connectivity- associated with each condition; if that is the case you could have CONN generate those first-level estimates directly during the denoising step)
Thanks
Alfonso
Originally posted by Roger Beaty:
Sorry but CONN does not create first-level SPM.mat files that you could import into SPM. Could you please expand a bit on what exactly you are trying to do to see if there are some work-arounds that come to mind? (e.g. are you trying to use CONN preprocessed time-series to look at the "activation" -BOLD signal changes; not connectivity- associated with each condition; if that is the case you could have CONN generate those first-level estimates directly during the denoising step)
Thanks
Alfonso
Originally posted by Roger Beaty:
Hi Alfonso and
all,
Following up from my previous question, I'm now trying to load the preprocessed time series from Conn into SPM for first- and second-level analyses. The SPM batch editor requires an SPM.mat file with timing information and a time series in temporal order to specify a model. This information is already implied by the condition files produced by Conn. Is it possible to bypass these steps and simply upload the condition files from Conn to specify a model in SPM?
Thanks,
Roger
Originally posted by Alfonso Nieto-Castanon:
Following up from my previous question, I'm now trying to load the preprocessed time series from Conn into SPM for first- and second-level analyses. The SPM batch editor requires an SPM.mat file with timing information and a time series in temporal order to specify a model. This information is already implied by the condition files produced by Conn. Is it possible to bypass these steps and simply upload the condition files from Conn to specify a model in SPM?
Thanks,
Roger
Originally posted by Alfonso Nieto-Castanon:
Hi
Roger,
If you want to use in SPM the BOLD timeseries after removal of confounding effects, you may simply select in Setup->Options the checkbox labeled 'Create confound-corrected time-series'. and that will create as part of the Preprocessing step a series of .nii files with these BOLD timeseries which you may then enter into SPM for further analyses (these will appear in your conn_*/results/preprocessing/ folder and the files will be named niftiDATA_Subject###_Condition###.nii).
If you have already run your preprocessing step and want to avoid having to repeat it, you may also simply type in the command window:
and that will create these same files from the already computed preprocessing-step results.
Hope this helps
Alfonso
Originally posted by Roger Beaty:
If you want to use in SPM the BOLD timeseries after removal of confounding effects, you may simply select in Setup->Options the checkbox labeled 'Create confound-corrected time-series'. and that will create as part of the Preprocessing step a series of .nii files with these BOLD timeseries which you may then enter into SPM for further analyses (these will appear in your conn_*/results/preprocessing/ folder and the files will be named niftiDATA_Subject###_Condition###.nii).
If you have already run your preprocessing step and want to avoid having to repeat it, you may also simply type in the command window:
conn_matc2nii;
and that will create these same files from the already computed preprocessing-step results.
Hope this helps
Alfonso
Originally posted by Roger Beaty:
Hi everyone,
I'm trying to load .mat files that were preprocessed in CONN back to SPM to perform a GLM analysis. In other words, I ran the preprocessing pipeline using the CONN gui and would like to import the files produced by CONN back to SPM. My guess is that CONN formats .mat files specific to the toolbox, and that these files are not easily loadable into the SPM GUI. Is it possible to import CONN files back to SPM and run a traditional GLM?
Many thanks,
Roger
I'm trying to load .mat files that were preprocessed in CONN back to SPM to perform a GLM analysis. In other words, I ran the preprocessing pipeline using the CONN gui and would like to import the files produced by CONN back to SPM. My guess is that CONN formats .mat files specific to the toolbox, and that these files are not easily loadable into the SPM GUI. Is it possible to import CONN files back to SPM and run a traditional GLM?
Many thanks,
Roger
Threaded View
| Title | Author | Date |
|---|---|---|
| Roger Beaty | Jun 12, 2014 | |
| Leah Fleming | Nov 23, 2016 | |
| kito24 | Feb 7, 2017 | |
| Laila Franke | Dec 19, 2018 | |
| Pravesh Parekh | Dec 19, 2018 | |
| Leah Fleming | Feb 8, 2017 | |
| Alfonso Nieto-Castanon | Jun 13, 2014 | |
| Chaleece Sandberg | Jul 31, 2015 | |
| Alfonso Nieto-Castanon | Aug 3, 2015 | |
| Chaleece Sandberg | Aug 5, 2015 | |
| Nole Hiebert | Aug 23, 2018 | |
| Alfonso Nieto-Castanon | Aug 6, 2015 | |
| Chaleece Sandberg | Aug 6, 2015 | |
| Chaleece Sandberg | Aug 6, 2015 | |
| Alfonso Nieto-Castanon | Aug 6, 2015 | |
| laurentcohen | Sep 14, 2018 | |
| Chaleece Sandberg | Oct 2, 2015 | |
| Alfonso Nieto-Castanon | Oct 8, 2015 | |
| kito24 | Feb 9, 2017 | |
| Ines Del Cerro | Nov 9, 2017 | |
| Patrick McConnell | Nov 3, 2017 | |
| Roger Beaty | Jan 26, 2015 | |
| Alfonso Nieto-Castanon | Jan 27, 2015 | |
| Roger Beaty | Jan 27, 2015 | |
| Alfonso Nieto-Castanon | Jan 29, 2015 | |
| laurel morris | Mar 11, 2016 | |
| Alfonso Nieto-Castanon | Mar 14, 2016 | |
| Isabel Berwian | Nov 28, 2016 | |
| Anila D'Mello | Nov 10, 2015 | |
| Roger Beaty | Jun 14, 2014 | |