Slice Timing HOWTO

How-Tos - Slice Timing Correction

How do I...

Do slice timing correction in SPM?

This should be done before any other preprocessing steps (coregistration, normalization, realignment, etc.) Slice timing will produce aV*.img files.

Number of subjects/sessions You'd typically enter 1 here. However:

  1. If you have multiple sessions, enter the number of sessions.
  2. If you have multiple subjects, enter the number of subjects multiplied by the number of sessions to get the total number of sessions you need slice-time corrected. (For example, you'd enter 12 if you had 3 subjects, 4 sessions per subject).

Select all SCAN1/V* images for your first session, then all SCAN2/V* images, etc., then repeat for each additional subject.

Select ascending if the way you prescribed and collected slices was starting from the bottom of the brain and going to the top (inferior to superior). Select descending if you prescribed and collected top-down. For any spiral imaging, select ascending (see Why ascending in spiral?).

Selecting user specified or interleaved would only be necessary if the slices were not collected sequentially (in space, that is). See the how-to for specifying a sequence in this case: Do slice timing correction with an interleaved slice sequence?

Reference slice (1=bottom): Enter the slice you want to consider as a reference point. All other slices will be corrected to what they would have been if they were acquired when the reference slice was acquired. The default is the middle slice (although, please make sure the default value given is indeed the middle slice for the number of slices you have). The default is generally fine, but make sure you write down what the value was! You should adjust your defaults for it down the road. Also read the SliceTimingFaq for thoughts on choosing your reference slice. If you have a structure of a priori interest (e.g., the hippocampus or something), it may make sense to choose a different slice than the default.

NB: The number you enter here is selected regardless of the acquisition sequences specified in 'Select sequence type' above. To decide what number you should enter here, disregard how your sequence was collected, and imagine a brain with slices numbered sequentially from bottom to top : 1 2 3 4 5 6 ... up to your #_of_slices. Now select the number corresponding to the slice you would like to use as reference. The number you have now is correct if you are using the standard spm99 release.

Interscan interval (TR) {sec} Enter the TR in seconds.

Acquisition time (TA) {sec} Enter the time between beginning of acquisition of the first slice and the beginning of acquisition of the last slice of one scan. Typically, i.e. for continuous acquisition protocols, TA = (TR/#_slices)*(#_slices - 1). The default value is calculated according to this formula.

For some specific acquisition protocols however (for instance clustered acquisition), the gap between acquiring the last slice of one scan and the first slice of the next scan may be considerably longer than the gap between acquiring sequential slices within the same scan. In such cases, you should calculate and enter a TA different than the default value.

Why ascending in spiral?

Note that this may be specific to spiral sequences acquired at the Lucas center.

From a question posed to G. Glover:
Q: In all of our protocols we specify our slice thickness (ex: 5mm) and our spacing (ex: interleave) in our scanning range. With the spiral-in-out functional scans, our question is whether our slices are acquired in a sequential (1,2,3...) or in an interleaved (1,3,5...2,4,6...) fashion from the seed slice. While under "Spacing" in the Scanning Range we specify "interleaved", we are not sure how this relates to the acquisition order if at all.

A: A lovely question. In my sequences the acquisition order is always sequential, whether you have skip 0, intlvd, or a number. So for time slice correction you need simply use a sequential schedule, where slice 1 is acquired TE + 10 ms after the start of TR, and successive slices are acquired in spatially contiguous order TR/nslices apart (unless its clustered).

Possibly Too Much Information (TMI):
The reason for interleaving in normal sequences is to try to mitigate as much as possible the influence of imperfect slice selection profiles, i.e. "slice bleed". If you have contiguous slice spacing (skip 0) but the slice oozes into the adjacent ones, then spin saturation effects reduce the net SNR. This effect can be reduced by acquiring the slices as far apart in time as possible, e.g. 1-3-5... 2-4-6.... The trouble is the interleave factor depends on the number of slices, so sometimes it could be 1-4-7... 2-5-8.... I simply avoid this by the use of a very high performance RF pulse so that slice bleed isn't a problem, and then I can use the simpler sequential order for your preprocessing pleasure.

Do slice timing correction with an interleaved slice sequence?

Just a stub - no info yet...