Hi Jill,
this really sounds reassuring. Please forgive me for misspelling your name in my post - it was your recipe I was referring to.
>
> You are right that some authors (e.g. Gitelman) have argued that deconvolution is important, especially for event related designs. However, what you get in both cases is a convolved PPI regressor in which your physiological data and task regressor were in the same state (either convolved or deconvolved) when you combined them. Thus the assumption you are making by not deconvolving in FSL are that the shape of the HRF should be roughly what you used to combine your task regressor.
I agree that it is essential that you have both regressors in the same state. A constant shape of the HRF is what is assumed in the Gitelman paper, too. Still, they state
"The result of multiplying the hemodynamic responses (BOLD signals), in these two regions, to generate the interaction term is shown in Fig. 2A. In contrast, Fig. 2B shows the result of first obtaining neural signals by deconvolution, multiplying the neural signals to obtain the interaction term, and then reconvolving with a HRF. The results of these operations are clearly different because the interaction term in Fig. 2A is completely insensitive to the relative onsets of the neuronal activities that determine the degree of interaction." (Gitelman et al., 2003, Neuroimage 19, p203).
They go on saying "The Pearson correlation coefficient for the BOLD versus neuronal interaction for block design data (r= 0.948) was significantly better than that for event-related data (r=0.709)" (ibid., p204)
With r=0.948 for block designs, you probably do not need to go all the way of de- and reconvolving. But for ER, while I would say that r=0.709 is not too bad, there is still a difference. What is not clear to me, however, if they also convolved the psychological regressor (which has a somewhat unusual shape, by the way) as they did with the artificial neural vectors. If yes, then this should probably be the way to go, shouldn't it? Perhaps one could include the method proposed in Makni, Beckmann, Smith and Woolrich (2008, Neuroimage 42, 1381-96) in a not-too-future version of FSL?
Best regards,
Cornelius
> This seems to me not very unreasonable, since in most analyses people do just assume a certain shape for the HRF.
>
> Hope that helps
>
> Jill
>
>
>
>
>
> -----------
> Dr Jill O'Reilly
> FMRIB Centre, Dept Clinical Neurology, Oxford University
> Phone +44 1865 222466
>
> ________________________________________
> From: Stephen Smith [[log in to unmask]]
> Sent: 23 February 2010 04:56
> To: Jill OReilly
> Subject: Fwd: [FSL] PPI: deconvolve a timecourse?
>
> Yo - two related queries below....?
> Ta ;-)
>
> Begin forwarded message:
>
> From: Cornelius Werner <[log in to unmask]<mailto:[log in to unmask]>>
> Date: 22 February 2010 09:04:42 GMT
> To: [log in to unmask]<mailto:[log in to unmask]>
> Subject: Re: [FSL] PPI: deconvolve a timecourse?
> Reply-To: FSL - FMRIB's Software Library <[log in to unmask]<mailto:[log in to unmask]>>
>
> Hi,
>
> I am also interested in running a PPI on ER-data. Could you point me
> to the paper in question?
> Could it be that in SPM there is/has been no option to leave
> regressors unconvolved as it is with FSL? Perhaps this would not make
> such a difference with a large block, but rather more so with a stick
> function (if it became convolved twice, once physiologically, and the
> second time by SPM) - could that be the reason? From Jill Kelly's
> instructions I gathered that if you enter a timecourse you extracted
> beforehand and choose "do not convolve", you should be fine, as the
> brain "convolved" the timeseries with its genuine HRF anyway. Any
> experts have an opinion on this?
>
> If you get any news on this, I'd be interested to hear about it.
> Thanks
> Cornelius
>
>
>
>
> On Wed, Feb 17, 2010 at 5:35 PM, Ilya Veer <[log in to unmask]<mailto:[log in to unmask]>> wrote:
> Hi all,
>
>
>
> I would like to do a PPI analysis on event-related task data which have been
> analyzed in FSL already. Following Gitelman’s paper, it seems that
> deconvolution of the physiological time-course (derived from the seed ROI)
> is a prerequisite when doing PPI’s on event-related designs. However, FSL
> doesn’t provide a tool for this. I tried to deconvolve the time-course using
> SPM, but I couldn’t manage to do this without running the entire analysis of
> the task data in SPM. Does anyone have experience in analyzing PPI’s on
> event-related data in FSL? Is it at all possible to obtain a deconvolved
> time-course without running the entire analysis in SPM? Lastly, is there an
> alternative for deconvolution when doing a PPI on event-related data in FSL?
> Any help with this issue would be greatly appreciated!
>
>
>
> Cheers,
>
> Ilya Veer
>
>
>
> ______________________________________
>
>
>
> Ilya Veer, M.Sc.
>
> Leiden Institute for Brain and Cognition (LIBC)
> Postzone C2-S
> P.O. Box 9600
> 2300 RC Leiden
>
> The Netherlands
> Tel. +31 71 526 1375
>
>
>
> --
> Dr. med. Cornelius J. Werner
>
> Department of Neurology
> RWTH Aachen University
> Pauwelsstr. 30
> 52074 Aachen
> Germany
>
> Institute of Neuroscience and Medicine
> MR Physics - INM4
> Research Centre Juelich
> 52425 Juelich
> Germany
>
> ::: Please encrypt confidential data :::
>
>
>
> ---------------------------------------------------------------------------
> 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]<mailto:[log in to unmask]> http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
|