Phase encoding direction from Philips Achieva

Hi
I’d be grateful for some guidance in figuring out the correct phase-encoding directoin from a Philips Achieva 3T diffusion dataset.
The dcm2niix .json file says
“PhaseEncodingAxis”: “j”,
“ImageOrientationPatientDICOM”: [
1,
0,
0,
0,
1,
0 ],
“InPlanePhaseEncodingDirectionDICOM”: “COL”,

The original protocol file says
fold-over direction =|“AP”
fat shift direction =|“P”
and this post
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=fsl;7311a4e2.1308)
indicates that the pe is therefore pa, while this post
https://www.nitrc.org/pipermail/mrtrix-discussion/2014-January/000866.html)
suggests the opposite, ap.

sorry, sent without finishing
Any guidance will be appreciated!
Cheers
James

I’m not sure we’ve ever verified what the actual sign of the PE direction was (@rsmith will not doubt correct me if I’m wrong here) – we only need to ensure it’s consistent. For the purposes of distortion correction, it doesn’t make any difference, as long as the PE direction we do extract is consistent across series (if we get it the wrong way around, we estimate the opposite B0 field, but the correction is appropriate nonetheless). So if you have a different application where you need to know the sign of the PE direction, then I’m afraid I’m not sure we can say for sure… You may have more luck on more BIDS-focused fora, e.g. this thread.

I would also guess that @lconcha’s post from the old mrtrix archives only deals with inverting the PE direction, I don’t think it makes any claims as to which direction the PE gradient actually points towards – only that you can reverse it using the setting mentioned. Again, if I’m wrong, I’m sure someone will (kindly :pray:) correct me…

Hi James. In which direction do the eyes appear to be distorted? If they look to be pulled/stretched anteriorly then PE is PA. If they appear to be flattened towards the brain then PE direction is AP.
Do you have blipped data as well?

Jerome

1 Like

Thanks Jerome and J-Donald, super helpful.

The eyes distort anteriorly so it’s definitely PA phase encoding. Unfortunately we don’t have the blipped data so we have to just work with what we have for this dataset.

Cheers

James

Note:

“PhaseEncodingAxis”: “j”,

is not equivalent to:

“PhaseEncodingDirection”: “j”,

The former is provided by dcm2niix presumably because it cannot ascertain the sign of phase encoding from the DICOM data. In the case of MRtrix3 currently, loading directly from the DICOM data would result in an absence of phase encoding information because of the inavailability of that sign information.

Hi Robert
Thanks for the additional clarification.
Does PhaseEncodingDirection “j-” equal PA (and PhaseEncodingDirection “j” equal AP)?
Cheers
James

Does PhaseEncodingDirection “j-” equal PA (and PhaseEncodingDirection “j” equal AP)?

Other way: A>>P is “j-”, P>>A is “j”. NIfTI coordinate system is RAS, so “increasing toward A” is positive.

This is however only correct if the NIfTI image transform is approximately axial. “j” indicates the axis with the second smallest stride, and for an axial image that will be anterior-posterior, but that’s not guaranteed to be the case. If understanding these mechanisms is important for you I’ve tried to explain it as best I could here.

Great, thanks Robert.