Improved 3D Radial Phyllotaxis Trajectories for Uniform Density Distribution of Readout Directions and Sequential Binning

Technical Summary: Improved 3D Radial Phyllotaxis Trajectories for Uniform Density Distribution and Sequential Binning

Problem Statement
Three-dimensional radial trajectories, specifically spiral phyllotaxis, are widely used in dynamic and motion-resolved MRI due to their motion robustness and self-navigation capabilities. However, the standard spiral phyllotaxis trajectory suffers from two primary limitations:

1. Non-Uniform Sampling Density: The standard trajectory does not achieve a uniform spherical distribution of sampling density (SDSD). It tends to cluster readout directions near the equatorial region, leading to anisotropic k-space sampling, direction-dependent variations in effective field of view and spatial resolution, and potential aliasing artifacts.
2. Sequential Binning Artifacts: While global k-space coverage is efficient, many applications (e.g., dynamic imaging, functional MRI, motion-resolved reconstruction) require reconstructing images from temporally contiguous subsets of readouts (sequential binning). In standard phyllotaxis, consecutive interleaves exhibit systematic clustering in polar angles. When these consecutive subsets are binned, they create circular gaps in k-space, manifesting as ringing artifacts and temporal inconsistencies in reconstructed images.

Methodology
The authors propose two methodological advancements to address these limitations, evaluated in vivo on 10 healthy volunteers using a 3T MRI scanner with two gradient-echo sequences (T1-weighted GRE and LIBRE).

1. Uniform Phyllotaxis (UPhy):

   • Goal: To achieve a uniform density distribution of readout directions (uniform SDSD).
   • Mechanism: UPhy redefines the polar angle formulation of the standard phyllotaxis. While the standard model uses $\theta_n = \frac{\pi}{2}\sqrt{\frac{n}{N}}$, UPhy utilizes $\theta_n = \arccos(1 - \frac{n}{N})$.
   • Effect: This modification ensures a uniform distribution of points over the sphere, eliminating the equatorial clustering observed in the standard model, while retaining the original azimuthal (golden angle) ordering.

2. Flexible Phyllotaxis (FlexiPhy):

   • Goal: To improve robustness to arbitrary retrospective sequential binning.
   • Mechanism: FlexiPhy builds upon UPhy but introduces a randomized permutation ($\sigma$) of the interleaves. This decouples the ordering of the polar angle from the azimuthal golden-angle progression. Specifically, the polar angle for the $j$-th point in the $k$-th interleaf is determined by a randomly permuted index $\sigma(k)$ rather than the sequential index $k$.
   • Effect: This randomization breaks the systematic clustering of polar angles within consecutive temporal bins. Consequently, sequential subsets of readouts maintain a well-distributed sampling pattern, preventing the formation of concentric k-space gaps and the associated ringing artifacts.

Key Contributions

• Analytical and Empirical Uniformity: The paper demonstrates that UPhy provides an analytically proven uniform SDSD, correcting the density bias of the standard phyllotaxis.
• Decoupling for Binning: The introduction of FlexiPhy offers a novel strategy to decouple polar and azimuthal ordering via randomization, specifically targeting the degradation of sampling properties in temporally contiguous subsets.
• In Vivo Validation: The trajectories were tested using two distinct sequences (standard GRE and fat-suppressed LIBRE) across static brain imaging and dynamic eye imaging scenarios.

Results

• Static Imaging: Visual inspection of sequential temporal-bin reconstructions revealed that both standard phyllotaxis and UPhy exhibited visible ringing artifacts and temporal inconsistencies. FlexiPhy substantially reduced these artifacts, producing more temporally stable images.
• Dynamic Imaging: In dynamic eye imaging, standard and UPhy trajectories produced structured ringing artifacts that propagated into the orbital region, obscuring anatomical details. FlexiPhy significantly reduced these artifacts, improving the visualization of the eye and surrounding structures.
• Quantitative Metrics:
  • SSIM and L2 Error: FlexiPhy demonstrated significantly higher Structural Similarity Index Measure (SSIM) values and lower relative L2 errors compared to both standard phyllotaxis and UPhy for both GRE and LIBRE sequences (Bonferroni-corrected $p < 0.05$).
  • UPhy vs. Standard: While UPhy improved theoretical angular uniformity, it did not significantly reduce ringing artifacts in sequential binning compared to the standard trajectory, except for a minor improvement in L2 error for the LIBRE sequence. This confirms that global SDSD improvement alone is insufficient to solve sequential binning artifacts.

Significance and Claims
The paper concludes that FlexiPhy enables more reliable sequential binning reconstructions by reducing trajectory-induced ringing artifacts and temporal inconsistencies. Its randomized construction is not tied to a specific binning strategy, making it broadly compatible with retrospective binning approaches used in dynamic and motion-resolved MRI.

The authors emphasize that while UPhy corrects the global density distribution, it is the decoupling of polar and azimuthal ordering in FlexiPhy that is critical for maintaining sampling quality within temporal subsets. The work suggests that FlexiPhy preserves the practical advantages of spiral phyllotaxis (efficient k-space coverage, smooth gradient transitions) while offering superior robustness for applications requiring temporal resolution or retrospective sorting. The study acknowledges limitations, noting that future work should investigate the impact on downstream tasks like motion estimation accuracy and explore the combination of these trajectories with pole-to-pole acquisition techniques.