Hi > On 12 Apr 2015, at 17:15, Nomi, Jason <[email protected]> wrote: > > Thank you Steve! > > Yes - I was speaking about the group-ICA spatial maps. I have noticed that I > can threshold the group-ICA spatial maps for each component at a much higher > level than other ICAs that I have done on non-HCP data. > > Your explanation about strong CNR makes sense. > > I am still a little unclear about the relationship of instantiating a strong > threshold on the group-ICA spatial maps relative to the time series.
there is no thresholding in the dual-regression approach - it's a multiple spatial (or temporal depending on the stage) regression. Eg in stage 1, the multiple unthresholded spatial maps are the multiple regressors - with all voxels used, but the higher-valued ones "used more". Cheers > > For example, if I threshold a component's spatial map at a lower level, more > areas of activation will naturally show up. Does the time series represent > all voxels in the spatial map when there is no thresholding? Or, does the > time series represent only the strongest voxels of activation? > > Thus, when I apply a strong threshold for image presentation to "clean up" > the image a little, does the time series also include those voxels that are > not visible due to high thresholding? > > Thanks again! > > Jason > > > From: Stephen Smith <[email protected]> > Sent: Saturday, April 11, 2015 3:29 AM > To: Nomi, Jason > Cc: [email protected] > Subject: Re: [HCP-Users] Parcellated Connectome > > Hi - there are many factors that affect overall scaling - more below: > > >> On 10 Apr 2015, at 14:22, Nomi, Jason <[email protected] >> <mailto:[email protected]>> wrote: >> >> Dear Experts, >> >> I have noticed that the time-series for individual subjects from the dual >> regression output in the parcellated connectome (100 comp ICA) has a much >> larger range than I am used to seeing. >> >> The range for time series values are approximately -800 to 800 while dual >> regression outputs that I have conducted myself are usually around -5 to 5. >> >> I also notice that I can set the threshold much higher for the independent >> components when isolating activation compared to dual regression analyses >> that I have done myself. This "cleans up" the component representation >> substantially. >> >> My questions are: >> >> 1) Is there a particular reason for this large increase in ranges? >> > > In this case most likely because we set the max of the group maps used in > dualreg stage 1 to be 1. This causes output timeseries to have larger scaling > - but the overall scaling is arbitrary anyway. > >> >> 2) Does the larger threshold for component activation have any influence on >> the time series that is being produced? Does the time series from the dual >> regression output only represent the areas from the independent component >> with the most intense activation? I would like to ensure that my >> presentation of component images using a much higher threshold is actually >> representative of the time series that I am analyzing. >> > > Do you mean the group-ICA spatial maps or maps output by diualreg stage 2? > > The group-ICA maps have high peaks (compared with the background scaling) for > a couple of reasons: a) because there are so many subjects being combined > that the ICA components are strong, and b) the group-PCA reduction has > removed a lot of unstructured noise before the PCA+ICA step. But despite the > maps having strong "CNR", they are still valid maps. > > Cheers, Steve. > > > >> >> Thanks! >> >> Jason >> >> >> >> _______________________________________________ >> HCP-Users mailing list >> [email protected] <mailto:[email protected]> >> http://lists.humanconnectome.org/mailman/listinfo/hcp-users > > > --------------------------------------------------------------------------- > Stephen M. Smith, Professor of Biomedical Engineering > Associate Director, Oxford University FMRIB Centre > > FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK > +44 (0) 1865 222726 (fax 222717) > [email protected] <mailto:[email protected]> > http://www.fmrib.ox.ac.uk/~steve <http://www.fmrib.ox.ac.uk/~steve> > --------------------------------------------------------------------------- > > Stop the cultural destruction of Tibet <http://smithinks.net/> > > > > --------------------------------------------------------------------------- Stephen M. Smith, Professor of Biomedical Engineering Associate Director, Oxford University FMRIB Centre FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK +44 (0) 1865 222726 (fax 222717) [email protected] http://www.fmrib.ox.ac.uk/~steve <http://www.fmrib.ox.ac.uk/~steve> --------------------------------------------------------------------------- Stop the cultural destruction of Tibet <http://smithinks.net/> _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
