 |
 |
Dear Mark, thank you very much for your reply. I guess I just need to work my way around it. I wish SPM had a way to implicitly include those regressors in the design matrix as in FSL. All the best, Elias On 23 Apr 2015, at 10:55, Mark Jenkinson <[log in to unmask]<mailto:[log in to unmask]>> wrote: Hi, It is not a good idea to regress these effects out of the data before running a GLM for two reasons: (i) the degrees of freedom will be inflated in the subsequent GLM as it doesn’t know about this processing; and (ii) the same filtering process really needs to also be applied to the regressors (EVs) in the GLM or otherwise there can be problems with shared variance not being accounted for correctly. Therefore I would not recommend this style of processing if you can avoid it. If you cannot insert these regressors into your GLM then you need to figure out if there is a way that you can both filter the regressors in the GLM appropriately and also tell it to adjust its DOF estimate. All the best, Mark From: Elias <[log in to unmask]<mailto:[log in to unmask]>> Reply-To: FSL - FMRIB's Software Library <[log in to unmask]<mailto:[log in to unmask]>> Date: Wednesday, 22 April 2015 12:26 To: "[log in to unmask]<mailto:[log in to unmask]>" <[log in to unmask]<mailto:[log in to unmask]>> Subject: [FSL] fsl_regfilt on motion parameters and outliers Hello, I want to use fsl_regfilt to remove volumes generated from the confound matrix file from the fsl_motion_outliers (which I find to be the most robust method identifying outlier volumes). I understand that I can add the file Feat if analysing in FSL, but due to consistency with previous data analysis I need to analyse in SPM, with 6 sessions for each subject, within a 2x2x2 design (each session represents a factorial level in a phfMRI experiment). As you might understand the design matrix becomes quite complicated in SPM if I include regressors for the motion parameters and outlier scans in the SPM design matrix. I thought it would be simpler and neater to regress out motion and outlier scans using fsl_regfilt and then run those filtered images in SPM first/second level with no need to add the motion and outlier regressors. Would that be possible? For the motion parameters I assume it is simple as instead of the .par file I can input the rp_ file from SPM. e.g. fsl_regfilt -i 4D -d rp_file.txt -f "1,2,3,4,5,6" -o 4D_movcor I am a bit confused what would be the best way to regress out the outlier scans from fls_motion_outliers. the option -f can either be component number or filter threshold. So if i set -f 1 for the confound matrix would it work (as all other values are zeros? eg fsl_regfilt -i 4D_movcor -d counfound_matrix.txt -f 1 -o 4D_filtered or do I add each column of the matrix as a component e.g.: fsl_regfilt -i 4D_movcor -d counfound_matrix.txt -f "1,2,3,4.... (number of columns for each matrix)" -o 4D_filtered would it be preferred or maybe also filter our the metric values for each volume that fsl_motion_outliers generates? I hope this makes sense and many thanks in advance, Elias