 |
 |
Thank you very much Donald and Luis, my confusion vanished :) greetings David 2014-07-15 0:35 GMT+02:00 MCLAREN, Donald <[log in to unmask]>: > > > > On Mon, Jul 14, 2014 at 12:04 PM, David Hofmann <[log in to unmask]> > wrote: > >> Hi Donald and all, >> >> thank you, I understand better now and think I got something wrong in my >> preprocessing. >> >> I'll describe my preprocessing steps in more detail, since I'm a still a >> little bit confused right now :): >> >> My data consists of 34 testsubjects, each with 3 sessions (34 x 3). >> > > With first level models, you have 1 subject with 3 sessions. Each session > has N scans, so you will have 3*N scans. > > >> >> *1.* I slice-timed the three sessions per subject so I got slice-timed files >> for every session per subject (3 x 34) >> > > You will have 3*N scans that are slice-time corrected. > > >> >> *2.* I used those files to do my realignment and got realigned files for >> every session per subject (3 x 34) and* one mean image for all session >> of a subject* (34 mean images, one per subject). In the spm manual it >> writes that the sessions are first realigned to each other, by aligning the >> first >> scan from each session to the first scan of the first session and then the >> images within each session are aligned to the first image of the session. >> > > You should have 3*N realigned images per subject plus the mean image per > subject. > > >> >> *3.* Then I used this mean image to coregister with the structural scan >> > > I would register the structural to the fMRI mean image. > > >> >> *4. *I used the coregistered image for segmentation >> >> *5.* Then I used Dartel to warp grey and white matter images and after >> that created the normalized smoothed images >> > >> In the end I got *34 normalized, smoothed* images to use for the GLM. >> > > You need to warp the 3*N images to MNI space, then smooth them. The mean > images cannot be used to create the first level models as they have no > information about the task related changes in the BOLD response. > > >> >> I think you meant that I need to have *34 x 3 normalized smoothed images* >> (34 scans for each session of a subject) to include in the first level >> model, right? >> > > No. You need 3*N normalized and smoothed images per subject. If your run > only has 34 images, I would be very surprised as this would be very short. > >> >> This makes sense, since I want to include the movement parameters for >> every session and also am confused about how to define my conditions and >> onsets for each session when I have only 34 scans per subject, but >> condition onsets for three sessions. >> > > You should have 3*N scans per subject, you have N rows in the 3 rp motion > parameter files that match the number of scans. You will define the onsets > separately for each session in the first level model. > > >> I hope this is more clear. >> >> greetings >> >> David >> >> >> 2014-07-14 9:11 GMT+02:00 MCLAREN, Donald <[log in to unmask]>: >> >> See below. >>> >>> On Sun, Jul 13, 2014 at 12:42 PM, David Hofmann <[log in to unmask] >>> > wrote: >>> >>>> Hi Donald, >>>> >>>> thanks for the answer and sorry for the late reply. In deed >>>> between-session effects are not of interest. >>>> >>>> I'm not sure if I understood you correctly: Do I need to pre-process >>>> every session for every subject seperately? >>>> >>> >>> Slice-timing is performed separately for each session. >>> Motion correction should be performed across all sessions to make sure >>> all the images are aligned. >>> >>> >>>> In other words, do I need to have three different mean functional >>>> images (one per session) for every subject for the first level model? >>>> >>> >>> The first level model does not use mean functional images. You want to >>> use the smoothed warped images. In the first level model for each subject, >>> you will setup 3 sessions. Each session will include the ALL the >>> smoothed/warped images for that session (note: its common not to use the >>> first few images of the scan to allow for magnetization equilibrium to be >>> reached - the images are removed before pre-processing). At the level of >>> generating the contrasts, you will form a contrast across the three >>> sessions. Thus, you will get one con_ image per subject for each contrast >>> you are interested in investigating. >>> >>> If your contrast weights in each run are divided by 3, then the contrast >>> will be the average of the contrast across sessions. Otherwise it will be >>> the sum of the contrasts. As long as all subjects have all sessions, both >>> methods will give you the identical second-level results. If subjects are >>> missing sessions, then the results will differ and you will want to make >>> sure your contrast represents the average across sessions. >>> >>> >>> >>> >>>> >>>> Because up to now I created *one* mean image over all sessions per >>>> subject by using three session tabs per subject in the slice-timing and >>>> realignment. >>>> >>> >>> I'm not sure how you got to "one" mean image per over all sessions, I >>> think my above answer will help clarify any issues. >>> >>> >>>> >>>> greetings >>>> >>>> David >>>> >>>> >>>> 2014-07-03 15:21 GMT+02:00 MCLAREN, Donald <[log in to unmask]>: >>>> >>>> Put all 3 sessions into your first level model, create a single >>>>> contrast across the three sessions, and then use this contrast in the >>>>> 2nd-level model. This assumes that you don't care about any session effects >>>>> or that you plan to ignore the session effect. This may or may not be good >>>>> depending on your question and study design. >>>>> >>>>> Best Regards, Donald McLaren >>>>> ================= >>>>> D.G. McLaren, Ph.D. >>>>> Research Fellow, Department of Neurology, Massachusetts General >>>>> Hospital and >>>>> Harvard Medical School >>>>> Postdoctoral Research Fellow, GRECC, Bedford VA >>>>> Website: http://www.martinos.org/~mclaren >>>>> Office: (773) 406-2464 >>>>> ===================== >>>>> This e-mail contains CONFIDENTIAL INFORMATION which may contain >>>>> PROTECTED >>>>> HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is >>>>> intended only for the use of the individual or entity named above. If >>>>> the >>>>> reader of the e-mail is not the intended recipient or the employee or >>>>> agent >>>>> responsible for delivering it to the intended recipient, you are hereby >>>>> notified that you are in possession of confidential and privileged >>>>> information. Any unauthorized use, disclosure, copying or the taking >>>>> of any >>>>> action in reliance on the contents of this information is strictly >>>>> prohibited and may be unlawful. If you have received this e-mail >>>>> unintentionally, please immediately notify the sender via telephone at >>>>> (773) >>>>> 406-2464 or email. >>>>> >>>>> >>>>> On Thu, Jul 3, 2014 at 9:17 AM, David Hofmann <[log in to unmask] >>>>> > wrote: >>>>> >>>>>> Hi @all, >>>>>> >>>>>> I have some fMRI data consisting of 30 testsubject where each subject >>>>>> has been tested three times. So there are multiple sessions. >>>>>> >>>>>> Now I don't know how to combine/merge those three session per subject >>>>>> in order to do my 2nd-level analysis. Can someone give me some tips how to >>>>>> proceed in such a case? >>>>>> >>>>>> greetings >>>>>> >>>>>> David >>>>>> >>>>>> >>>>>> >>>>>> >>>>>> >>>> >>> >