I haven't analyzed ASL data, but don't you have to subtract the control images from the label images, and concatenate those images? So then your TR would equal 2.5s. I am not sure if you did this already, or even if it is appropriate, but something to think about. See also Wang Z et al (2008, MRI).
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] on behalf of Joao Pereira [[log in to unmask]]
Sent: Wednesday, February 27, 2013 9:15 AM
To: [log in to unmask]
Subject: Re: [SPM] Too much lag in the design matrix?
Dear Andreas, and SPM list,
Thank you for having pointed me in the direction of the microtime resolution and onset. Fiddling with these values, I was able to extract better results.
However, I still have some doubts concerning the best choice for these. As I am working with ASL data, my equivalent TR is 2x2.5s = 5s. Due to the greater scan duration, I have increased the microtime resolution to - I've tested results with 40 and 80, and the latter gave me better results. The onset puzzles me though - this is ASL data, so no timing correction was applied. As the acquisition was interleaved [2:2:9, 1:2:9], my guess was to choose t0 = t, i.e. 40 or 80, and indeed the results vastly improved. I must ask if this makes sense though, and if there is a some other way of addressing the choice of these values that is less heuristic.
Best,
Joćo
On 20 February 2013 22:55, Joao Pereira <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Thanks for quick reply, Andreas. I haven't tried specifying seconds yet, but will do that next. My microtime onset and resolution were left at their default values (16).
Good point re the modelling of the "fixation" condition. I'll correct that in my model (In any case, the underlying issue re the odd lag will persist)
Best,
Joćo
On 20 February 2013 22:23, Andreas Finkelmeyer <[log in to unmask]<mailto:[log in to unmask]>> wrote:
It's hard to say what went wrong here, but you're right that an onset at 0 should produce an earlier deflection. Have you tried specifiying the units in "seconds" instead of "scans" i.e. for the movement condition onsets of [15 45 75 ...] and durations of 15? Also, what are your settings for microtime onset and resolution?
Andreas
*the reason I chose the movement blocks as example is that you probably shouldn't actually specify onsets for 'fixation' anyways. Right now your two "conditions" are probably highly (negatively) correlated, which is bad. Also, you wouldn't expect a hemodynamic 'response' to your baseline, but simply a return to some equilibrium state (i.e. the periods of 0's in between your movement blocks).
________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]<mailto:[log in to unmask]>] on behalf of Joao Pereira [[log in to unmask]<mailto:[log in to unmask]>]
Sent: 20 February 2013 18:10
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: [SPM] Too much lag in the design matrix?
Dear SPMers,
I have been trying to perform a simple analysis with a fixation vs movement visual task with ASL. My TR is 2.5s, and due to the control-tag nature of the ASL scans I say in my model that the interscan interval is 5s. I am using the standard double-gamma canonical HRF. My design matrix (image below), however, seems odd to me as the fixation task (starting at scan "0") only starts to show non-zero values in X(3,1), i.e. at 3*interscan interval = 15 seconds! Given the canonical HRF peak at 6-8s or so, this seems a bit off. The results confirm this: if I change the onsets so as to subtract 2 scans from each onset, the results are as expect - as it is, I see an empty glass brain. Using the derivatives hasn't helped either.
The details of the blocks are: Fixation = [ 0 6 12 18 24 30 36 42 48 54 60 66 72 78 84]; Activation = [3 9 15 21 27 33 39 45 51 57 63 69 75 81] (all block durations are 3 perfusion scans, i.e. 6 "normal" scans). Other parameters were left as default.
Any help will be much appreciated.
Thank you!
Joao
[Inline images 1]
|