 |
 |
Hi - yes, this is a reasonable approach. Cheers, Steve. On 15 Feb 2007, at 16:13, Clare Kelly wrote: > Thanks Steve.. > > I don't mean to question the FSL's ability to deal with motion - > it's just that I've never seen such a strong artifact before, > appearing consistently across several subjects (which was also why > I thought it might not be motion, but some sort of flow or weird > scanner artifact) - it's showing up on a random-effects group > analysis with 16 subjects!! > > What I did with that particular subject was: correlated the > timeseries for the rim ICA component with the > prefiltered_func_data_mcf_abs > .rms to ensure it was motion-related, then entered it as a nuisance > covariance in my correlation analysis. This worked really well, > almost completely removing the artifact from the stat maps. > > I intend to do this for each subject who shows such rim-type > artifacts (i.e., as long as the component correlates significantly > with motion). Does this seem like a reasonable approach?? > > Thanks again, > Clare > > > On 2/15/07, Steve Smith < [log in to unmask]> wrote:On 14 Feb > 2007, at 19:04, Clare Kelly wrote: > > > > > Thanks Daniel and Steve, > > > > You're completely right, when I look at the graphs and a movie > > there is a large rightward rotation of the head. > > > > So how come this artifact appears, despite motion correction and > > the inclusion of motion parameters in the model? > > Normally because the motion is so large that there will be residual > effects (due to spin history, interpolation, nonlinear distortion > etc) even after perfect motion correction and inclusion of linear > motion regressors. > > Cheers, Steve. > > > > > > > I've attached the prefiltered_func_data_mcf_rel.rms graph. > > > > Thanks so much for all your help! > > Clare > > > > > > > > > > On 2/14/07, Daniel Wolf <[log in to unmask]> wrote: Hi Clare, > > I'm not an FSL expert, but wanted to add my 2 cents anyway. > > Motion artifact is most likely explanation, commonly affects > edges of > > brain more than other regions, and produces ICA components with low > > frequencies. > > The fsl motion summary parameters, MAD and MRD, are mean values. > So, > > even tho you see mean relative displacement of .08, that doesn't > mean > > there wasn't some volume-to-volume movement big enough to produce > > artifacts (it only takes about .2mm abrupt motion, sometimes > less, to > > produce an artifact that cannot be removed by standard motion > > correction algorithms). If you look in the motion correction > > directory "mc" you will see a textfile > > prefiltered_func_data_mcf_rel.rms...this has the n-1 values for > > volume-to-volume displacement (the 6 motion parameters are combined > > into a single measure). If you plot these numbers, or just look at > > them, you'll see whether any are unusually high...look especially > for > > >.5mm jumps... plotting them will tell you roughly what timepoint > they > > happened at (this will probably match up with the scans showing big > > changes in the ICA component timecourse), then you can make sure to > > check these timepoints during the movie Steve proposed. (also, to > > make the movie, you can do: avwmaths filtered_func_data -sub > > mean_func filtered_func_demeaned. removing the mean makes it much > > easier to see the artifacts.) > > Dan > > > > > > On 2/14/07, Clare Kelly < [log in to unmask]> wrote: > > > Thanks Steve I will try that - sorry about the image, it was to > > get the > > > email under the 50K limit! > > > > > > Best, > > > Clare > > > > > > > > > On 2/14/07, Steve Smith < [log in to unmask] > wrote: > > > > It's hrad to tell as the image you sent is very low res, but > this > > > > looks like a motion artefact. View the data as a movie before > and > > > > after motion correction and see if this makes sense.... > > > > Cheers. > > > > > > > > > > > > On 13 Feb 2007, at 14:42, Clare Kelly wrote: > > > > > > > > > > > > > > Hello all, > > > > > > > > > > I've been running some connectivity analyses with resting > > data and > > > > > see some fairly large rim artifact for several seeds in many > > of my > > > > > subjects. > > > > > > > > > > I re-ran my preprocessing with melodic and I see a component > > such > > > > > as the one in the attached figure in several subjects. For > this > > > > > subject (who probably showed the most severe artifact), > this was > > > > > the second component, explaining 8.16% of the explained > variance > > > > > (7.6% of the total variance). > > > > > > > > > > I'm planning to include the timeseries for this component as a > > > > > nuisance covariate in my analysis, but I'm wondering if > > anyone has > > > > > any idea what it is, and if it's something I can prevent in > the > > > > > future. I didn't think it could be movement as the total > > (absolute) > > > > > movement in this subject was only 0.23mm, relative movement > was > > > > > 0.08mm. > > > > > > > > > > > > > <sub8019movement.jpg> > > > ---------------------------------------------------------------------- > -- > --- > 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) > [log in to unmask] http://www.fmrib.ox.ac.uk/~steve > ---------------------------------------------------------------------- > -- > --- > ------------------------------------------------------------------------ --- 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) [log in to unmask] http://www.fmrib.ox.ac.uk/~steve ------------------------------------------------------------------------ ---