 |
 |
See below for comments. >>> Jeni Chen <...> 09.12.14 7.04 Uhr >>> Dear Helmut, Thank you very much for the detailed explanation regarding this subject! It's always been unclear to me and certainly confusing at times when we come across older posts suggesting different things. So just to make sure I 'finally' understood it :) : In summary, if we use slice timing then the microtime onset should reflect the acquisition time of the reference slice. In that case the onsets don't have to be modified. If the reference slice is the temporally first slice then the microtime onset is 1, if it's the temporally middle the microtime onset is (microtime resolution)/2, for example 16/2 = 8. If you use any other slice as a reference you have to determine the most appropriate microtime onset yourself = (acquisition time of the slice)/TR * microtime resolution. For example with an ascending interleaved scheme 2 4 6 ... 30 1 3 5 ... 29 the spatially middle slice is 15 due to the 30 slices overall, and it is acquired at the beginning of the third quarter of the TR, resulting in a microtime onset of 3/4 * 16 = 12. In principle you could also use slice timing (working on the acquired data, e.g. onto the temporally middle slice), set the microtime onset to 1 (the predictor would then be "too early") and adjust the onsets (e.g. minus TR/2). But it's much more likely to introduce errors this way as you might accidentally correct the onsets into the wrong direction. Thus regarding setting the reference slice- whether using the middle slice in time or in space, if I understood it correctly, it is really not a problem in a sequential acquisition but for an interleaved acquisition, one must be careful when setting up the microtime onset, depending on what was chosen as their reference slice in the STC. Yes, in sequential acquistion schemes the temporally middle slice should correspond to the spatially middle one. But as any slice can serve as a reference (although the temporally middle one seems to be the most plausible one as long as you don't have any specific reasons as pointed out before) you should always make sure that the microtime onset fits well. Otherwise you have a different prediction (expected time course is shifted in time), and this is independent of the acquisition scheme. The default microtime settings (8 and 16) correspond to STC onto the temporally middle one. If you haven't done so during preprocessing then you have to adjust the microtime onset yourself. For STC onto the temporally first slice you would e.g. go with a microtime onset of 1 during model specification, when correcting onto the temporally last one you would go with a microtime onset of 16 (assuming the default resolution of 16), and if you plan to use a specific spatial one you have to find out yourself in most instances (except if you're lucky and the spatially middle one corresponds to the temporally first/middle/last). So in your example (ascending interleaved), the temporally first slice would be slice #2, correct? Yes, if it's 2 4 6 ... 1 3 5 .... and if you want to use the temporally first one for STC you have to enter "2" as reference slice since the numbers during slice timing reflect the spatial property of the slices (also called slice index in the manual). In that example slice "2" as the spatially second (usually the second most ventral) is the temporally first slice. You would NOT enter "1" because this would reflect the spatially first one, which would have been acquired in the middle of the TR. However, if it's 2 4 6 ... 1 3 5 .... and if you want to correct onto the temporally middle slice you would enter "1" or the highest even number (because these two slices, the spatially lowest = 1 and the spatially highest = e.g. 30 in case of 30 slices overall, would have been acquired in the middle of the TR). If we want to choose the temporally first slice as reference slice, then the corresponding microtime onset would then be 2 as well, right? No. The numbers entered for microtime onset / microtime resolution are based on temporal aspects = the time bins of a TR. A microtime onset "2" means that the acquired data (its acquistion time) corresponds to the second time bin out of the default 16. But when correcting the data onto the temporally first slice you would of course also want to go with the first time bin, thus microtime onset "1". If you don't use any slice timing at all then a microtime onset of 8 (or microtime resolution/2) should be the best, as all the data has been acquired "somewhen during the TR". I think this was not the case in SPM5 at least, as far as I remember the default value was still 1 there. If we don't use STC at all, then wouldn't it make more sense to set the microtime onset to the slice that represents t0, in your example (ascending interleaved), slice #2, which is the first slice being acquired? (so that it aligns with the event times). Well, if you don't go with STC at all, then most of the slices are slightly shifted in time. Assuming all the data of a volume to be acquired in the middle of the TR should lead to the least misalignment (- TR/2 for the first slice, ... + TR/2 for the last one, if you sum up the absolute values this is smaller than 0 for the first slice, ... + TR for the last one). Thank you again for your help! kind regards, Jeni