Inferring GE Slice Timing From Protocol Data Block

[neurolabusc]

There are three methods to determine slice times for GE EPI data:

1. TriggerTime (0018,1060)
2. RTIA Timer (0021,105E)
3. Private Protocol Data Block (0025,101B)

The methods are described here.

In the past, dcm2niix only attempted the first two methods. The latest commit will attempt the third method if the first two fail (just like dicm2nii). The private data block only reports whether the data was acquired in sequential or interleaved order. We can infer the slice times by knowing the order, repetition time and number of slices. Unfortunately, there does not seem to be a reliable way to determine the acquisition time with GE, so this formula will only be precise for continuous fMRI (with no temporal gap between volumes) and will fail for sparse acquisitions (though we do not apply slice time correction when there are large gaps). When this method is applied, dcm2niix will generate the warning Guessing slice times using ProtocolBlock SliceOrder to warn the user that this is an estimate.

I have added examples to dcm_qa_stc that show this method for sequential and interleaved series, each acquired in inferior-to-superior and superior-to-inferior order. dcm2niix stores GE axial slices to disk in inferior-to-superior order regardless of acquisiton order, and when it flips the slices it transforms the sform, qform and slice timing so they remain in register.

This should resolve issues for many users.