help
help > RE: HCP and Conn toolbox
Sep 27, 2015 05:09 PM | Noah Philip - Center for Neurorestoration and Neurotechnology, Brown University, USA
RE: HCP and Conn toolbox
Thank you Alfonso. I would be
very appreciative of your batch scripts and parallel processing as
otherwise this will be a herculean task.
Noah_Philip at Brown.edu
Noah
Originally posted by Alfonso Nieto-Castanon:
Noah_Philip at Brown.edu
Noah
Originally posted by Alfonso Nieto-Castanon:
Hi Noah,
The HCP data is very clean but in my experience there is still some amount of residual bias and inter-subject variability in the observed voxel-to-voxel correlation distributions that is improved by performing additional denoising steps. Since the degrees of freedom of the BOLD timeseries in this dataset are really very large there should be no theoretical "downside" to performing additional denoising steps so I would still recommend adding ART and aCompCor as part of denoising. If you want I will send you my scripts for batch processing the HCP dataset (these are tested only on the Q6 release but they should -likely- work on other releases as well; also these make use of the newer parallelization options in CONN, if you are analyzing the entire dataset it is highly recommended to run these analyses on a cluster to keep running time to reasonable levels).
Best
Alfonso
Originally posted by Noah Philip:
The HCP data is very clean but in my experience there is still some amount of residual bias and inter-subject variability in the observed voxel-to-voxel correlation distributions that is improved by performing additional denoising steps. Since the degrees of freedom of the BOLD timeseries in this dataset are really very large there should be no theoretical "downside" to performing additional denoising steps so I would still recommend adding ART and aCompCor as part of denoising. If you want I will send you my scripts for batch processing the HCP dataset (these are tested only on the Q6 release but they should -likely- work on other releases as well; also these make use of the newer parallelization options in CONN, if you are analyzing the entire dataset it is highly recommended to run these analyses on a cluster to keep running time to reasonable levels).
Best
Alfonso
Originally posted by Noah Philip:
Good morning - a quick question on using Conn on
the HCP data
what are your thoughts on using the FIX-ICA denoised data as opposed to CONN's CompCor? Is there a good compelling reason to use one over the other? Also, if using the denoised data, how would you use it practically within the toolbox? Would there be a way to skip" directly into the first level analyses?
Thank you for the guidance!
Noah
what are your thoughts on using the FIX-ICA denoised data as opposed to CONN's CompCor? Is there a good compelling reason to use one over the other? Also, if using the denoised data, how would you use it practically within the toolbox? Would there be a way to skip" directly into the first level analyses?
Thank you for the guidance!
Noah
Threaded View
| Title | Author | Date |
|---|---|---|
| Noah Philip | Sep 25, 2015 | |
| Alfonso Nieto-Castanon | Sep 26, 2015 | |
| Noah Philip | Sep 27, 2015 | |