help
help > RE: Motionartifacts / ART-detection toolbox
Oct 11, 2012 02:10 AM | Alfonso Nieto-Castanon - Boston University
RE: Motionartifacts / ART-detection toolbox
Dear Diana,
Yes, in the art toolbox select 'save SPM regressors' and that will create a covariate file (one per session) that you can directly enter as a first-level covariate in the conn toolbox (effectively removing outlier scans from consideration when computing functional connectivity measures). In my experience the default threshold values/combination should work just fine, but you are right that these default values have been obtained in the context of optimizing the resulting sensitivity of standard functional analyses; we have not yet explored the effect of different thresholding options in the context of functional connectivity analyses. My impression is that perhaps a more conservative set of thresholds might be preferable in the context of functional connectivity analyses given that the 'cost' associated with false negatives -failing to identify outlier scans- is higher in connectivity analyses (where outlier scans can act as confounding effects, increasing type I errors) than in standard functional analyses (where outlier scans mostly affect sensitivity, typically increasing type II errors), but until we perform some specific analyses I cannot give you a more definite answer regarding 'optimal' movement threshold values.
Best
Alfonso
Originally posted by Diana Wotruba:
Yes, in the art toolbox select 'save SPM regressors' and that will create a covariate file (one per session) that you can directly enter as a first-level covariate in the conn toolbox (effectively removing outlier scans from consideration when computing functional connectivity measures). In my experience the default threshold values/combination should work just fine, but you are right that these default values have been obtained in the context of optimizing the resulting sensitivity of standard functional analyses; we have not yet explored the effect of different thresholding options in the context of functional connectivity analyses. My impression is that perhaps a more conservative set of thresholds might be preferable in the context of functional connectivity analyses given that the 'cost' associated with false negatives -failing to identify outlier scans- is higher in connectivity analyses (where outlier scans can act as confounding effects, increasing type I errors) than in standard functional analyses (where outlier scans mostly affect sensitivity, typically increasing type II errors), but until we perform some specific analyses I cannot give you a more definite answer regarding 'optimal' movement threshold values.
Best
Alfonso
Originally posted by Diana Wotruba:
Dear all
I am comparing the connectivity of a patient-group with a group of healthy controls. Therefore I'd like to to use the ART-detection toolbox (http://www.nitrc.org/projects/artifact_d...) to control for the movement-artifacts. I understand that i can do this by loading the single subjects motion/artifacts.mat file I get from the ARTtoolbox as a covariate in conn. is this correct?
Furthermore, in ART there are different possibilities for thresholds-usage (diff-global / diff-motion/ comp motion). I would be very interested in what experiences you have with using different thresholds in resting-state data. Which movement- (and/or z-) tresholds do you suggest me to use?
Most of the subjects didn't move for more than 1,5 mm. In some of the subjects there are single peaks of more than 2mm movement, though. Do you suggest removing those subjects from the analysis or is it sufficient to control for those motion-peaks using ART?
Any advice is much appreciated
Best,
Diana
I am comparing the connectivity of a patient-group with a group of healthy controls. Therefore I'd like to to use the ART-detection toolbox (http://www.nitrc.org/projects/artifact_d...) to control for the movement-artifacts. I understand that i can do this by loading the single subjects motion/artifacts.mat file I get from the ARTtoolbox as a covariate in conn. is this correct?
Furthermore, in ART there are different possibilities for thresholds-usage (diff-global / diff-motion/ comp motion). I would be very interested in what experiences you have with using different thresholds in resting-state data. Which movement- (and/or z-) tresholds do you suggest me to use?
Most of the subjects didn't move for more than 1,5 mm. In some of the subjects there are single peaks of more than 2mm movement, though. Do you suggest removing those subjects from the analysis or is it sufficient to control for those motion-peaks using ART?
Any advice is much appreciated
Best,
Diana
Threaded View
| Title | Author | Date |
|---|---|---|
| Diana Wotruba | Aug 23, 2012 | |
| Alfonso Nieto-Castanon | Oct 11, 2012 | |