A question regarding the smoothing, Gaussian random field theory requires that you apply a lot of smoothing for the statistics to be valid. How do you calculate p-values corrected for multiple comparisons if you don't smooth the data? Do you use permutation tests?
- Anders 2015-03-14 23:01 GMT+01:00 Glasser, Matthew <[email protected]>: > If you don’t want ventricle activations in the task fMRI data, it would > be good to clean the data with something like ICA+FIX which will remove > spatially specific structured noise (such as whatever is causing the > ventricles to light up) prior to fitting the GLM. There are other stimulus > correlated artifacts in the task data (uncorrelated artifacts would tend to > get averaged out in the GLM analysis) such as strong deactivation in > orbitofrontal regions in the Tongue movement contrast. Use of ICA+FIX in > task analysis was looked at some inside the consortium, but I’m not sure if > there was ever a focus on seeing that stimulus correlated artifacts were > being removed (vs just seeing how Z-stats changed with cleanup, which they > don’t much since most of the variance in HCP fMRI timeseries is > unstructured). > > Also it makes me sad to see you aren’t using the CIFTI data, which are > substantially more accurately registered across subjects and don’t have the > unnecessary blurring with white matter and CSF signals (and in 3D across > sulci and gyri) induced by unconstrained volume-based smoothing as > misalignment between functional areas. The volume-based data simply don’t > allow you to take advantage of the high spatial resolution that the HCP > data were acquired with like the CIFTI data do, so you’re missing out on > all the cool new things you can see. > > Peace, > > Matt. > > From: vanessa sochat <[email protected]> > Date: Saturday, March 14, 2015 at 4:29 PM > To: "[email protected]" <[email protected]> > Cc: Russell Poldrack <[email protected]> > Subject: [HCP-Users] De-activations in "LANGUAGE" Task Contrast Maps > > Hi HCP Users, > > We recently generated group contrast maps using the 500 subjects release > data, and discovered strong deactivations in the ventricles for the "STORY" > contrast. We generated the group maps using fsl's randomise: > > randomise -i 4D -o OneSampT -1 -T > > > and the 4D file was concatenated cope1.nii.gz images from the single > subject directories, eg: > > >> /MNINonLinear/Results/tfMRI_LANGUAGE/tfMRI_LANGUAGE_hp200_s4_level2vol.feat/cope1.feat/stats/cope1.nii.gz > > > We have prepared a notebook that visualizes outliers > <http://nbviewer.ipython.org/github/vsoch/brainmeta/blob/master/image_comparison/notebook/outliers_in_hcp.ipynb#Visualizing-The-Outliers> > +/- 6 standard deviations from the mean for each of the 86 task/contrasts, > as well as counting the total number > <http://nbviewer.ipython.org/github/vsoch/brainmeta/blob/master/image_comparison/notebook/outliers_in_hcp.ipynb#Counting-Outliers-+/--6-Standard-Deviations> > . > > The outliers that are positive activations (which are beautiful, by the > way!) are not concerning. Would it be possible to talk about the outliers > in the ventricles for the following contrasts? > > > - tfMRI_LANGUAGE_STORY > - tfMRI_LANGUAGE_MATH > - tfMRI_LANGUAGE_neg_MATH > - tfMRI_LANGUAGE_neg_STORY > > We did a similar procedure > <http://nbviewer.ipython.org/github/vsoch/brainmeta/blob/master/image_comparison/notebook/hcp_language_explore.ipynb> > for a subset (N=46) of the single subject maps for the STORY contrast (the > same cope images described above) in case it is helpful. > > Thanks for your help with this. This dataset is amazing to have for > meta-analysis research! > > Best, > > Vanessa > > -- > Vanessa Villamia Sochat > Stanford University > (603) 321-0676 > > _______________________________________________ > HCP-Users mailing list > [email protected] > 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 > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > -- Anders Eklund, PhD _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
