I think there are two problems, either of which could certainly cause large differences in the outputs.
> # format movement parameters (manually corrected header after paste step, not > shown) > Text2Vest Movement_Regressors.txt Movement_Regressors.mat > Text2Vest Movement_Regressors.txt Movement_Regressors_dt.mat > paste Movement_Regressors.mat Movement_Regressors_dt.mat > > Movement_Regressors_all.mat It appears that you can’t rely entirely on fix_3_clean to understand what has happened to HCP FIX :-) The processing you’re doing to the Movement_Regressors differs from “functionmotionconfounds.m”, and appears incorrect to me. FIX creates 24 motion regressors to remove from the data. Those 24 regressors are a) the parameters from the motion correction (6 regressors), b) backward derivative of those parameters (6 regressors), c) the square of the parameters and backward derivatives (12 regressors). The Movement_Regressors_all.mat that you've created will not contain those regressors. The Movement_Regressors.txt and Movement_Regressors_dt.txt effectively contain the same information, except the "dt" file is detrended and (I believe) demeaned. However, I never use it because it's using a simple linear detrend. You want to detrend the regressors with the 2000s high pass filter. This is also done somewhere in the code, to provide a file called "mc/prefiltered_func_data_mcf_conf_hp", but I can’t find where that occurs at this moment. > # regress unique variance from bad components (taken from .fix file) out of > timeseries > > > fsl_regfilt -i REST1LR_bp_mc.nii.gz -d melodic_mix_mc -o > REST1LR_bp_mc_softICA -f "1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, > 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 28, 29, 30, 31, 32, 33, 34, 36, > 38, 39, 40, 41, 42, 43, 44, 45, 46, 49, 50, 51, 52, 53, 54, 55, 56, 57, 61, > 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 81, 82, 83, 84” > # compare against HCP's released FIX-denoised file > > fslmaths rfMRI_REST1_LR_hp2000_clean -sub REST1LR_bp_mc_softICA > diff_REST1LR_bp_mc_softICA This doesn’t seem to do what you claim. I believe this is regressing ALL of the variance in the noise ICs, hence it is an aggressive regression. To do the soft regression, you need to remove from the noise variance all of the variance in the motion parameters AND the variance in the signal components. That yields the “unique variance” to regress from the timeseries data. I don’t believe you are doing that here, therefore, these will not be consistent with the HCP FIX pipeline. > Does the above denoising scheme seem consistent with what FIX is doing? I > plan to use FIX going forward, rather than trying to replicate it using the > FSL command-line, but I'd like to understand any discrepancies between the > two. > Hopefully this is been a useful pedagogical exercise! The FIX scripts run in octave and matlab, so there shouldn’t be reason to ‘rewrite’ the FIX code in fsl commands once you feel that you understand the code. --Greg ____________________________________________________________________ Greg Burgess, Ph.D. Staff Scientist, Human Connectome Project Washington University School of Medicine Department of Psychiatry Phone: 314-362-7864 Email: gburg...@wustl.edu ________________________________ The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail. _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users