 |
 |
Dear Joon-Soo, >Dear SPMers, > >I am just a biginner in the field of SPM. >I recently tried to use SPM for the analysis of fMRI time-course data, >but I have been faced to some questions. > >First, I gave it a try to realign the images along the time-course. >It seemed to result good, when the option 'Coregister only' is selected. >However, when I select 'reslice' option, I got the new images named r*.img. >I think the resliced images are a little distorted while SPM is reslicing. I take it here that the "distortion" refers to the intensity changes alluded to below, rather than any actual shape distortions. >I don't know the details of reslice algorithm, but if it changes the intensity >of voxels, I could lose the original time-course data. > >So, I did simulation making simple phantom images. > > 1 1 1 1 1 2 2 2 2 2 1 1 1 1 1 (total 15 images) > >The order of numbers( 1 1 ...2 2 ...) represents time-course. >1 respresents a simple circle and 2 same-size circle including another >circle inside. >The small circle inside the large circle has different value which means >activation area >whithin a brain. These images have no change in their position ever. > >After I performed 'reslce', I got the totally same images for all 15 images. >I think this algorithm recognizes the changes of intensity as motion movement. >Am I right or wrong? > I think you may well be right. It depends on wether you have used the "Adjustment for movements" or not. The default option is a 2nd order adjustment so I guess you might have. What the adjustment for movement does is that, in addition to reslicing according to the estimated movement parameters, it will regress out all variance that is correlated with the movement parameters. Obviously your simulated data contains no movement, and the registration of your 1 1 1 1 1 2 2 2 2 2 1 1 1 1 1 time series is likely to result in finding no movement at all within the "blocks", but is likely to find some (extremly small) "movements" between blocks. Now, since there are no actual movements the estimated movement parameters will be colinear with your "experimental" design, and the variance will be regressed out. I suggest you try redoing your simulation, this time using "No adjustment for movement". I expect that now you will find your "activations" again. The problem of "activations" posing as movements is extremly unlikely to occurr on real data where there are noise and actual movements present, so I don't think that this should completely scare you off "adjustment for movement". The only time when this becomes a problem on real data is when movements are actually correlated to (caused by) the experimental paradigm. >Anybody can kindly answer me? >Please any comment will be a great help to me. > > > Joon-Soo Kim Good luck Jesper %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%