Thanks Helmut!
Cheers,
Yann
On 15 Aug 2015, at 8:52 pm, H. Nebl <[log in to unmask]> wrote:
> Dear Yann,
>
> The microtime resolution is actually unrelated to the no. of slices acquired. Slices are acquired at discrete time points, and these differ between slices. However, even if we just have data from discrete time points we can predict the BOLD response e.g. based on the temporal resolution of the logfile, e.g. stimulus onset functions with a microtime resolution of 1 ms. A lower microtime resolution means that the predicted value for a certain discrete time point will be less accurate (because durations are rounded to the nearest time bin), and a higher microtime resolution will increase accuracy. Even if we acquired data e.g. every 10 s we could go with a microtime resolution of 1 ms and try to test for effects of different durations (100 ms vs. 101 ms).
>
> In practice, this is unnecessary most of the time though, e.g. because you have constant durations all the time (but it's nonetheless not a bad idea to adjust the microtime resolution). We can come up with super-precise predictors, but what do they reflect? Does it make sense at all to assume constant "on" periods? Does the neural process really start at stimulus onset, does it end when the response is recorded? This is
> 1) why event-related designs often just go with 0 s (resulting in a duration = repetition time / microtime resolution), because for short durations, a different duration mainly affects the scaling of the predictor but not the temporal pattern. It becomes critical if you want to go with different durations for different conditions. Do you want the predictors to take into account different durations / reaction times? This is going to depend on your hypotheses then.
> 2) some people go with a microtime resolution reflecting the no. of slices because this way, it's very easy to specify the microtime onset. If you have acquired 17 slices and want the temporally 2nd to correspond to the microtime onset just go with 2 and 17.
>
> If you have slice-time corrected onto the spatially 1st = temporally middle slice then the microtime onset should also be set to *the middle time bin*, with microtime resolution 16 this would be time bin 8, with microtime resolution 32 this would be time bin 16, with microtime resolution 2000 this would be time bin 1000.
>
> Best
>
> Helmut
>
|