Hi,
If you had absolute trust in your HRF then I would agree with you.
However, as the delay in the canonical HRF is know to vary between
individual, brain region, session, caffeine level, etc, etc then we advise
people to include the temporal derivative as it not only accounts for
changes in slice timing, but it also accounts for any difference in the
real HRF delay and the one specified when generating the model.
This isn't explicitly stated on the slides, but I do make that point when I
give the talk. I'll try to incorporate that information on the slides too.
All the best,
Mark
Alle Meije Wink wrote:
> Dear Mark,
>
> Thanks for your response!
>
> Would slice timing correction by means of shifting the design (which
> can be computed exactly) not be superior to using the time derivative?
>
> What I understand from the derivative story is that its use is
> actually mainly to clean the residual -it's far worse in modelling the
> response than an analytically computed correction to the design would be.
>
> Does that make sense?
> Alle Meije
>
>> Yes - you're right. My mistake. It should read that slice 9 is
>> acquired 1.5 seconds
>> *after* slice 10. I'll fix that.
>>
>> Rest assured, however, that the implementation of slicetimer in FSL
>> is fully
>> tested and correct, even though we do not recommend using slice
>> timing correction.
>> The use of temporal derivative instead, is of course driven by the
>> data and hence
>> without problems of this kind.
>>
>> All the best,
>> Mark
>
>
|