1. Certain sequences (e.g., PROPELLER, SNAILS) rely on multiple acquisitions. To account for phase differences in the sequential acquisitions that arise from motion, correction is done based off of the repeatedly-acquired center of k-space (a major recurring theme in reconstruction).

2. Aliasing and blurring throw off the accuracy of this phase correction.

3. However, the correction still works IF you include it in the reconstruction. Mathematically, this works the same way as a parallel imaging reconstruction, with each of the low-resolution phase maps serving the same purpose as the images from the separate coils.

The rest of the paper is straightforward testing of the idea.

It’s a very elegant approach. It seamlessly integrates PI and phase correction–it should complement almost anything else you want to do with PROPELLER, really (or any other technique that repeatedly acquires the center of k-space).

Simultaneous phase correction and SENSE reconstruction for navigated multi-shot DWI with non-cartesian k-space sampling

Liu et al. Magn Reson Med. 2005 Dec;54(6):1412-22. PMID 16276497

I emailed you about it, hopefully you can comment on it?

Once they have all the corrected blades, they need to make sure that the blade orientations are also exact. They basically do a correlation-based correction for k-space rotation using the magnitude of the k-space data. Rebecca improved on that for her MS thesis project if I recall and can comment further.

Next up is bulk translation, which they do by creating a reference image formed by the average of all the k-space centers from each blade, though in my dissertation I showed better results using the first blade as the reference.

Finally, they do a correlation between all the corrected blades to try and reject blades with too much through-plane motion. After that, they finally get to the re-gridding process and combine the blades all together.

So, as to the question of how homodyne and PROPELLER would play together, you tell me It does seem fiendishly complex. A casual PubMed search shows no results for “PROPELLER homodyne” but there mihgt be something out there, maybe buried in ISMRM abstracts.

The authors here avoid throwing a monkey wrench into these nontrivial works by separating the recon into stages. They are doing parallel imaging to acquire each blade, so the entire PROPELLER recon is a black box downstream. So, if you wanted to include homodyne, I imagine that would be the best way to do that also - use it to acquire a single blade.

Ah, I see I made a conceptual mistake in my post though. By using homodyne for a single blade, they wont really see any significant decrease in scan time, any more than they do by using parallel imaging. So homodyne isn’t the answer to reduced scan time unless you can do the Hard problem of integrating it into the PROPELLER recon.

BTW, without cheating by looking at the reference, I bet you can tell me why they use a triangular function for the filter. You’re the recon God around here, Dustin