Hi Theresa,
Looking closely, there are actually many fields that need to be set
properly to distinguish between a time series and not.
-----------
Fields that are required so that it does not seem the data is already
aligned in time:
1. dim_info : this should be 48 (or possibly more) to indicate that the
slice are along the 3rd dimension
2. slice_end : index of the last slice in the timing pattern (probably 32)
3. slice_start : index of first slice in pattern (probably 0)
4. slice_code : set to 3 for alt+z
5. slice_duration : time for single slice (so maybe 0.060606 = 2 sec/33 slices)
-----------
Fields that are required for proper interpretation of data:
a. dim : NIfTI has the time axis as the 4th dimension (always), and if
not time, a multi-volume dataset should count along the 5th axis
b. pixdim : pixdim[4] should equal the TR (with dim[4] equal to NT)
c. xyzt_units : should probably be 10, to mean size in mm, time in seconds
-----------
To display one last example, let me compare to identical NIfTI datasets,
but where one is a time series (TR=2, 33 slices, NT=21) and the other is
just a collection of volumes:
nifti_tool -diff_hdr -infiles volumes.nii time_series.nii
name offset nvals values
------------------- ------ ----- ------
dim_info 39 1 0
dim_info 39 1 48
dim 40 8 5 80 80 33 1 21 1 1
dim 40 8 4 80 80 33 21 1 1 1
pixdim 76 8 1.0 2.75 2.75 3.0 1.0 1.0 0.0 0.0
pixdim 76 8 1.0 2.75 2.75 3.0 2.0 0.0 0.0 0.0
slice_end 120 1 0
slice_end 120 1 32
slice_code 122 1 0
slice_code 122 1 3
xyzt_units 123 1 2
xyzt_units 123 1 10
slice_duration 132 1 0.0
slice_duration 132 1 0.060606
So there is a lot to set to have accurately interpretted data (8 fields).
To answer your last question, no, I cannot fathom how one could tell whether
time shifting had already been done (just by looking at the data).
Hope that helps,
- rick