Re: [HCP-Users] MELODIC denoising vs. released ICA-FIX datasets

[Harms, Michael]

Hi Ely, Maybe I’m missing something, but if you are trying to see if you can duplicate the FIX-denoised file already provided by HCP, wouldn’t it be good to just run a modification of ‘hcp_fix’ itself and see how the results of that compare?  The modification would be to comment out the ‘melodic’ call, since melodic is not deterministic, and set up that script instead to use the filtered_func_data.ica that is provided as part of the FIX-extended packages.  I *think* (Steve can correct me if wrong) that everything else in FIX should be deterministic. cheers, -MH --  Michael Harms, Ph.D. ----------------------------------------------------------- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave. Tel: 314-747-6173 St. Louis, MO  63110 Email: mha...@wustl.edu From: <hcp-users-boun...@humanconnectome.org> on behalf of Stephen Smith <st...@fmrib.ox.ac.uk> Date: Tuesday, September 27, 2016 at 8:52 PM To: "Ely, Benjamin" <benjamin....@mssm.edu>, "HCP-Users@humanconnectome.org" <HCP-Users@humanconnectome.org>, "Glasser, Matthew" <glass...@wustl.edu> Subject: Re: [HCP-Users] MELODIC denoising vs. released ICA-FIX datasets Hi - it sounds like maybe it's working fine within the limits of slight differences in mathematical precision between the C++ vs matlab parts of the processing - so the main question would be - have you looked in a viewer at the difference image - e.g. are the voxels with large differences isolated or eg at the edge of the brain? Cheers. On 28 Sep 2016, at 02:49, Glasser, Matthew <glass...@wustl.edu> wrote: I think this is more of a question for the FSL list, but I don’t know fsl_glm well enough to say if what you are doing is equivalent or not. Peace, Matt. From: "Ely, Benjamin" <benjamin....@mssm.edu> Date: Tuesday, September 27, 2016 at 7:20 PM To: Matt Glasser <glass...@wustl.edu>, "Burgess, Gregory" <gburg...@wustl.edu>, "HCP-Users@humanconnectome.org" <HCP-Users@humanconnectome.org> Subject: Re: [HCP-Users] MELODIC denoising vs. released ICA-FIX datasets Hi Matt and Greg, Thanks for the feedback! I've looked at the various fix .m files from the current release; based on fix_3_clean.m, I tried the following for a single resting-state run: # highpass filter; sigma of 1000.08 = FWHM of 2355 per Smith et al 2013 NeuroImage, also consistent with comments in the fix_3_clean.m script fslmaths rfMRI_REST1_LR.nii.gz -bptf 1000.08 -1 REST1LR_bp # 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 # regress movement parameters out of timeseries and re-add mean fsl_glm -i REST1LR_bp.nii.gz -d Movement_Regressors_all.mat --out_res=REST1LR_bp_mc_demeaned.nii.gz --demean  fslmaths REST1LR_bp.nii.gz -Tmean REST1LR_bp_mean fslmaths REST1LR_bp_mc_demeaned.nii.gz -add REST1LR_bp_mean.nii.gz REST1LR_bp_mc # regress movement parameters out of melodic mix fsl_glm -i filtered_func_data.ica/melodic_mix -d Movement_Regressors_all.mat --out_res=melodic_mix_mc --demean # 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 Visual inspection and fslstats indicate reasonably good agreement between my denoised file and the HCP's denoised file; the mean difference is about 0.84 units (compared to a mean signal intensity of around 10,000), and the "robust" range of the difference is about +/- 72 units. More worryingly, though, the maximum difference is around 2000 units, and around 6000 voxels show differences greater than 500 units, so I'm not sure machine precision can account for the differences. 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.  Thanks again, -Ely   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 --------------------------------------------------------------------------- Stephen M. Smith, Professor of Biomedical Engineering Head of Analysis,  Oxford University FMRIB Centre FMRIB, JR Hospital, Headington, Oxford  OX3 9DU, UK +44 (0) 1865 222726  (fax 222717) st...@fmrib.ox.ac.uk    http://www.fmrib.ox.ac.uk/~steve --------------------------------------------------------------------------- Stop the cultural destruction of Tibet _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users   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