Two small comments: 1) As noted in Barch et al. 2013, the EMOTION and LANGUAGE tasks do not include a fixation baseline condition. Therefore, activation estimates for an individual condition (e.g., Fearful faces) will not be valid. The only valid estimates are between conditions (Fearful faces vs. neutral shapes).
2) I mis-spoke below. Feat automatically demeans the timeseries data and the design matrix (rather than adding a B0 intercept term to the GLM). Either way, you should not demean the task data before running it through Feat. --Greg ____________________________________________________________________ Greg Burgess, Ph.D. Staff Scientist, Human Connectome Project Washington University School of Medicine Department of Neuroscience Phone: 314-362-7864 Email: [email protected] > On Dec 18, 2015, at 11:06 AM, Greg Burgess <[email protected]> wrote: > > Hello Nico, > >> On Dec 18, 2015, at 9:39 AM, ZUO, Nico <[email protected]> wrote: >> >> >> Dear HCP Colleagues, >> >> Here I want to obtain an group-average areas for a task activation, for >> example for the Emotion task, and I want to get the activation areas either >> from FEAR or NEUT task by constrast to the fixation duration. The HCP >> website said that the Q1 release contained such kind of (mybe) data, >> >> http://www.humanconnectome.org/documentation/Q1/data-in-this-release.html#group_average >> , called HCP_Q1_GroupAvgUnrelated20.zip, however I did’nt find the entrance >> for downloading from the HCP database. I am not sure it is not available any >> more. > > > If at all possible, you should move to the latest releases of task fMRI data > (500 subjects and 900subjects data releases). The later releases include > subject-level (aka Level 2) FEAT outputs, making it unnecessary to run FEAT > on the minimally-preprocessed images at all! > > >> Additionally, if I want to find these activation areas from the minimally >> processed data (the path in the S500 release (based on volume data) is like >> “…100307/MNINonLinear/Results/tfMRI_EMOTION_LR…”), but I have a little >> confusion about the adhered .fsf file, tfMRI_EMOTION_LR_hp200_s4_level1.fsf >> . The processed task fMRI volume data, for example, >> tfMRI_EMOTION_LR.nii.gz, has been detrened, but not been regressed out the >> head motion (although it contains the transformation file, >> Movement_Regressors.txt ) and white matter/CSF. And it has not been >> filtered to removing the noise. I am not sure my understanding is right ? >> However, it seems that the confiuration file, >> tfMRI_EMOTION_LR_hp200_s4_level1.fsf, is directly acting on the unpolished >> data, tfMRI_EMOTION_LR.nii.gz. I am not sure it is right for calling >> FSL/Feat ? > > > The level1 fsf files are only included for reference or for modification. If > it is your goal to include nuisance confound regressors, then it may be > worthwhile. Otherwise, stop and use the publicly available analyzed feat > outputs. > > Additional information about the processing is also available in the follow > two papers: > Barch, D. M., Burgess, G. C., Harms, M. P., Petersen, S. E., Schlaggar, B. > L., Corbetta, M., et al. (2013). Function in the human connectome: Task-fMRI > and individual differences in behavior. NeuroImage, 80, 169–189. > http://doi.org/10.1016/j.neuroimage.2013.05.033 > Glasser, M. F., Sotiropoulos, S. N., Wilson, J. A., Coalson, T. S., Fischl, > B., Andersson, J. L., et al. (2013). The minimal preprocessing pipelines for > the Human Connectome Project. NeuroImage, 80, 105–124. > http://doi.org/10.1016/j.neuroimage.2013.04.127 > > I will try to answer some of your questions here. The minimally-preprocessed > image files (for example, tfMRI_EMOTION_LR.nii.gz) have NOT been demeaned or > detrended. The minimal preprocessing for NIFTI volume files essentially > performs distortion corrections, rigid-body motion correction, and > registration to a group-template (see Glasser et al Figure 19). The CIFTI > grayordinate files map voxels to surface and subcortical gray matter, which > involves some resampling and minimal spatial smoothing (Glasser et al Figure > 20). > > The Feat lower-level analysis subsequently performed spatial smoothing (4mm > FWHM in volume, or several different levels of smoothing in CIFTI > grayordinates), highpass temporal filtering (using fslmaths -bptf with 200 > second cutoff), and prewhitening using FILM. There are no additional confound > regressors removed from the data (e.g., no regression of motion parameters, > CSF or WM, motion spike regressors, FIX noise regressors). Be forewarned that > these tend to overlap with the task design regressors. More details are > included in Barch et al. 2013. > > >> Another question is for processing the filtered/regressed/demean-ed >> tfMRI_EMOTION_LR.nii.gz (I assumed my above understanding is right so I >> continue to try…). The demean-ed task fMRI data is floating around 0, for >> example [-310, 320]. Unfortunately the FSL/Feat is not able to generate the >> mask during the “Prestats” step, the log file is as follows, > > > It is not correct to run analyses using Feat on demeaned data. Feat will > remove the mean for you by including a B0 intercept term in the GLM. > > It is possible to run Feat analyses using the FSL Feat_Gui. However, there > are optimizations in the process, especially the analysis of surface data in > CIFTI format, if you use the Task fMRI HCP Pipelines available at > https://github.com/Washington-University/Pipelines > > > --Greg > > ____________________________________________________________________ > Greg Burgess, Ph.D. > Staff Scientist, Human Connectome Project > Washington University School of Medicine > Department of Neuroscience > Phone: 314-362-7864 > Email: [email protected] > > >> >> <log> >> >> …… >> >> Prestats >> >> /mnt/software/fsl5.0/bin/fslstats prefiltered_func_data -p 2 -p 98 >> >> -114.3 125.35 >> >> /mnt/software/fsl5.0/bin/fslmaths prefiltered_func_data -thr -108.2 -Tmin >> -bin mask -odt char >> >> …… >> >> </log> >> >> Certainly here the function “fslmaths” is definitely not able to generate a >> meaningful mask by the “-108.2 -Tmin” parameters since the data distribute >> on the both sides of 0. Since the HC data has already been stripped, so the >> mask.nii.gz is already there (Actually the “-Tmin” doesn’t make any sense >> here). Maybe this question should be posted to the FSL forum, but it is >> related to the HCP data so I posted it here too. >> >> Any reponse would be Appreciated. Thank you. >> >> Nico Zuo >> >> -- >> Institute of Automation >> Chinese Academy of Sciences >> Beijing 100190, China >> >> >> >> >> _______________________________________________ >> HCP-Users mailing list >> [email protected] >> http://lists.humanconnectome.org/mailman/listinfo/hcp-users >> > > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
