help
help > RE: Setup
Jun 14, 2016 03:06 PM | Alfonso Nieto-Castanon - Boston University
RE: Setup
Hi Bob,
Yes, there is one additional subtlety when entering data that has been preprocessed externally. Typically the functional volumes that you enter in CONN are normalized to MNI space and spatially smoothed. Yet, when extracting the BOLD signal within your ROIs it is typically recommended to extract the BOLD signal from functional volumes which have not been smoothed (extracting the ROI signals from spatially smoothed functional volumes instead would effectively include a small portion of the BOLD signal from outside the ROI boundaries due to the original smoothing step). This is admitedly a somewhat subtle/small effect, yet in order to accomodate this (as well as to accomodate ROIs that may have been defined in a different non-MNI space and other more convoluted cases) CONN allows you to define multiple functional "datasets" (in the Setup.Functional tab) and then specify, for each ROI, from which "dataset" CONN should extract the corresponding ROI-level BOLD signal (in the Setup.ROIs tab). The defaults in CONN are set so that, if you have run all preprocessing steps either within CONN or from SPM, then "dataset-0" (the data used for voxel-level analyses) is pointing to the MNI-space smoothed functional data, while "dataset-1" (the data used for ROI-level BOLD signal extraction) is pointing to the MNI-space unsmoothed functional data. If you have run your preprocessing steps on a different software package then, in addition to entering the functional/structural/ROIs/covariates data into CONN, it is also recommended to go to Setup.functional and change the default settings there so that "dataset-1" now points to the location of your unsmoothed functional data (that will typically have been generated as part of your preprocessing procedure, but the naming and location of these files will vary depending on the software used). If your data has been preprocessed externally and you do not have access to the unsmoothed functional volumes (e.g. some already preprocessed public datasets), that is fine, then simply either leave the default settings (and CONN will give you that warning message saying that it is failing to find "dataset-1" and it is using "dataset-0" for ROI-level extraction instead) or change in Setup.functional the "dataset-1" option to "same as dataset-0" to avoid that warning. The only penalty in this case will be that ROI-level data will be extracted from already smoothed functional volumes (with the associated small inclussion of beyond-boundaries effects) but that should be a relatively small/minor issue in any way (perhaps in this case it might be advisable to increase the number of erosion steps used in the WhiteMatter and CSF ROIs definitions in order to minimize the potential for inclusion of gray matter voxel effects there).
Let me know if this helps clarify
Best
Alfonso
Originally posted by Bob Kraft:
Yes, there is one additional subtlety when entering data that has been preprocessed externally. Typically the functional volumes that you enter in CONN are normalized to MNI space and spatially smoothed. Yet, when extracting the BOLD signal within your ROIs it is typically recommended to extract the BOLD signal from functional volumes which have not been smoothed (extracting the ROI signals from spatially smoothed functional volumes instead would effectively include a small portion of the BOLD signal from outside the ROI boundaries due to the original smoothing step). This is admitedly a somewhat subtle/small effect, yet in order to accomodate this (as well as to accomodate ROIs that may have been defined in a different non-MNI space and other more convoluted cases) CONN allows you to define multiple functional "datasets" (in the Setup.Functional tab) and then specify, for each ROI, from which "dataset" CONN should extract the corresponding ROI-level BOLD signal (in the Setup.ROIs tab). The defaults in CONN are set so that, if you have run all preprocessing steps either within CONN or from SPM, then "dataset-0" (the data used for voxel-level analyses) is pointing to the MNI-space smoothed functional data, while "dataset-1" (the data used for ROI-level BOLD signal extraction) is pointing to the MNI-space unsmoothed functional data. If you have run your preprocessing steps on a different software package then, in addition to entering the functional/structural/ROIs/covariates data into CONN, it is also recommended to go to Setup.functional and change the default settings there so that "dataset-1" now points to the location of your unsmoothed functional data (that will typically have been generated as part of your preprocessing procedure, but the naming and location of these files will vary depending on the software used). If your data has been preprocessed externally and you do not have access to the unsmoothed functional volumes (e.g. some already preprocessed public datasets), that is fine, then simply either leave the default settings (and CONN will give you that warning message saying that it is failing to find "dataset-1" and it is using "dataset-0" for ROI-level extraction instead) or change in Setup.functional the "dataset-1" option to "same as dataset-0" to avoid that warning. The only penalty in this case will be that ROI-level data will be extracted from already smoothed functional volumes (with the associated small inclussion of beyond-boundaries effects) but that should be a relatively small/minor issue in any way (perhaps in this case it might be advisable to increase the number of erosion steps used in the WhiteMatter and CSF ROIs definitions in order to minimize the potential for inclusion of gray matter voxel effects there).
Let me know if this helps clarify
Best
Alfonso
Originally posted by Bob Kraft:
Once again I am confused. You are talking
about smoothed and smoothed data. If I wanted to do all the
preprocess outside of Conn (normalization, slice timing correction,
distortion correction, etc) what would I need to enter into CONN
for the Setup. My current understanding is that I would need
to enter in the normalized fMRI data, the structural image, and the
tissue masks (gm, wm, and csf). I would also need to include the
motion parameters and the scrubbing matrix.
Is there anything I am missing?
Bob
Is there anything I am missing?
Bob
Threaded View
| Title | Author | Date |
|---|---|---|
| Bob Kraft | May 11, 2016 | |
| Alfonso Nieto-Castanon | May 13, 2016 | |
| Bob Kraft | Jun 5, 2016 | |
| Alfonso Nieto-Castanon | Jun 6, 2016 | |
| Bob Kraft | Jun 14, 2016 | |
| Alfonso Nieto-Castanon | Jun 14, 2016 | |
| Bob Kraft | Jun 29, 2016 | |
| Alfonso Nieto-Castanon | Jul 1, 2016 | |