 |
 |
Hi, On 9 Jun 2009, at 17:22, Nicole Pelot wrote: > Hi Steve, > > Thanks so much. I apologize for the questions with evident answers. > I was > actually being thrown by the fact that the time courses produced by > the Full > Model oscillate steadily every volume, whereas I am accustomed to > seeing my > 4 rest blocks interleaved with 3 active blocks, as is the result > from the > pre-subtraction (and is also the result I used to get when doing the > subtraction in Matlab). > > Hopefully my last questions on this issue: > > 1) I should indeed be using EV3 (zstat5) as my perfusion activation > maps & > time courses (just need re-thresholding). No - please read the documentation on the full modelling carefully - if you used the wizard then you want zstat1 (not 5), which tells you about the perfusion activation. > 2) If I have a 3ms TE, why is there such a strong "BOLD" component? I assume this wasn't 7T? But it probably was Spirals (as opposed to EPI) - is that right? In which case there's a lot of signal that comes in after the "3ms"? > 3) I normally use the time course of the mean z-score to compute % > signal > change, SNR and CNR (just in Excel), but I really cannot see how these > metrics fit in with the Full Model, since the timecourses no longer > manifest > my blocks. For example, I compute the percent signal change as the > difference between the average of the active blocks and the average > of the > rest blocks, all divided by the average of the rest blocks. You need to find out the peak-peak height of EV3 (look in design.mat); then take the value of the cope1 image at the voxel of interest - multiplying these together gives you the effect size - then you can normalise that by the mean signal level (look in mean_func). Cheers. > > Your help is really appreciated. > > Nicole > > > On Tue, 9 Jun 2009 08:54:30 +0100, Steve Smith > <[log in to unmask]> wrote: > >> Hi, >> >> Actually - your results look fine - and it's a nice example of the >> full model approach working slightly better than the pre-subtraction >> (which in theory it should be in an ideal world). >> >> Judging from the tsplot outputs, the peak Z is actually higher with >> full model, and there are slightly more supra-threshold voxels (I'm >> comparing zstat1 from each approach). Maybe you're just being thrown >> by the fact that (because the peak Z is lower with pre-subtraction) >> "yellow" corresponds to lower zstats with pre-subtraction - so you >> see >> more yellow.... >> >> Cheers, Steve. >> >> >> >> On 8 Jun 2009, at 13:49, Nicole Pelot wrote: >> >>> Thanks Steve. I have used the Model Wizard; I apologize for the >>> simple >>> question. But even now, when I run it again using the Wizard (rather >>> than >>> perfusion subtraction), it does not seem to give me reasonable >>> output. With >>> perfusion subtraction, I clearly get a very good fit; not the case >>> when >>> using the full model. I have a TE of 3ms, and therefore should have >>> negligible a BOLD component, but this does not seem to be the result >>> of the >>> full model. I then have to wonder whether: >>> >>> 1) I'm somehow misusing the Full Model, even with the wizard. >>> 2) I'm mis-interpreting the output. >>> 3) There's a more fundamental issue with my data. >>> >>> I've uploaded images (152682) to illustrate the issue... Is this >>> what a >>> problem with structured noise would look like (i.e. reason to >>> stick to >>> perfusion subtraction)? >>> >>> Thanks so much, >>> >>> Nikki >>> >>> On Sun, 7 Jun 2009 07:58:36 +0100, Steve Smith >>> <[log in to unmask]> wrote: >>> >>>> HI - have you tried using the model Wizard, as suggested in the >>>> manual? This will do the basic setup for you. >>>> Steve. >>>> >>>> >>>> On 5 Jun 2009, at 19:03, Nicole Pelot wrote: >>>> >>>>> Hi, >>>>> >>>>> I have read the FEAT Perfusion documentation quite a few times >>>>> now, >>>>> but I >>>>> still cannot figure out how to properly set up my EVs. I have a >>>>> block design >>>>> task: 56s/block, 7 blocks, alternating between rest and active >>>>> (start & >>>>> finish with rest). >>>>> >>>>> My main issue is with EV1. It clearly states that we should be >>>>> modeling >>>>> *rest* times, but that leads me to think of a block design with >>>>> 0's >>>>> for >>>>> active and 1's for rest. However, the sample design seems to show >>>>> all 1's >>>>> for EV1. Also, what convolution should be applied? None? >>>>> >>>>> I've structured EV2 as I always do to correspond to my block >>>>> design. >>>>> (On: >>>>> 56s; Off: 56s; Convolution: Double gamma). >>>>> >>>>> EV3, I chose "Interaction", and I made the mean of EV1 zero (but >>>>> the >>>>> min of >>>>> EV2). >>>>> >>>>> Thank you, >>>>> >>>>> Nicole >>>>> >>>> >>>> >>>> --------------------------------------------------------------------------- >>>> 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 >> --------------------------------------------------------------------------- > > > --------------------------------------------------------------------------- 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 ---------------------------------------------------------------------------