Patient setup variation on Elekta Unity and its impact on adaptive planning

M. Yan, Erika Kollitz, Sheng‐Hsuan Sun, Kathryn Hitchcock, Alexandra De Leo, Amanda Schwarz, Luke Maloney, Jonathan Li, Chihray Liu, Guanghua YanJournal of Applied Clinical Medical Physics 2025 · DOI

TL;DR

This study investigates patient setup variations (PSVs) on the Elekta Unity MR-linac, a system with unique immobilization challenges. The findings demonstrate that while adaptive planning effectively compensates for these variations, minimizing PSVs, especially rotations, remains crucial for optimizing treatment efficiency and accuracy.

What Problem Does This Paper Address?

Modern radiotherapy relies heavily on precise patient positioning to ensure radiation targets are accurately hit while sparing healthy tissues. The Elekta Unity, an advanced MR-linac, integrates an MRI scanner directly into the treatment unit. This integration, while offering real-time imaging for adaptive planning, introduces significant constraints on patient immobilization. The limited gantry size and integrated MR coils restrict the use of traditional immobilization devices, potentially leading to larger patient setup variations (PSVs) compared to conventional linear accelerators. These PSVs can complicate treatment delivery and necessitate adaptive planning, which itself has time and resource implications. This paper aims to quantify these PSVs on the Elekta Unity across different treatment sites and assess how effectively the system's adaptive planning capabilities can accommodate them, exploring the impact of PSV magnitude on planning time and dosimetric outcomes.

Methods

The researchers analyzed PSVs from actual patient treatments on the Elekta Unity, including data from brain, pancreas, and pelvis (prostate and rectum) patients. To simulate PSVs and evaluate their impact on adaptive planning, they artificially shifted reference MR scans by varying amounts (1-3 cm) in the left-right (LR), superior-inferior (SI), and anterior-posterior (AP) directions. Two adaptive planning workflows were investigated: "adapt to position" (ATP) and "adapt to shape" (ATS). The ATP workflow reoptimizes MLC leaf positions and segment weights for minor setup variations, while the ATS workflow allows for contour adjustments and re-optimization for more significant changes. The study compared adaptive planning time, monitor units (MUs), and dosimetric metrics (target coverage and organ-at-risk sparing) between the two workflows and across different simulated PSV magnitudes.

Datasets

The study utilized retrospective data from 10 brain patients, 10 pancreas patients, and 10 pelvis patients (5 prostate, 5 rectum) treated on the Elekta Unity system. For the adaptive planning simulations, five prostate and five pancreas plans were selected. The PSV data was collected from daily treatment fractions, and simulated PSVs were generated by artificially shifting reference MR scans.

Key Insights and Ideas

  • MR-linac Immobilization Challenges are Real: The Elekta Unity's design inherently leads to larger PSVs compared to conventional linacs, primarily due to limitations in immobilization devices and the absence of lateral lasers.
  • Adaptive Planning is Effective, But Not a Panacea: Both ATP and ATS workflows successfully compensated for the simulated PSVs, maintaining target coverage and OAR sparing. This highlights the power of adaptive planning in overcoming setup uncertainties.
  • ATS is More Time-Consuming: The "adapt to shape" workflow consistently took significantly longer than the "adapt to position" workflow, a critical factor in clinical workflow efficiency.
  • PSV Magnitude Has Limited Impact on Dosimetry: Surprisingly, the magnitude of the simulated PSVs (up to 3 cm) had a minimal impact on target coverage and OAR sparing in the adaptive plans. This suggests that the adaptive algorithms are robust.
  • Rotations are the Unaddressed Elephant: While translational shifts were well-managed, the study emphasizes that the rigid registration used in the adaptive workflows does not account for rotational errors, which remain a significant challenge and can necessitate more time-consuming ATS workflows or even patient repositioning.

Key Results

  • Significant PSVs Observed: Brain treatments showed average PSVs of -0.2 ± 0.3 cm (LR), 0.3 ± 0.7 cm (SI), and 0.8 ± 0.7 cm (AP), with maximums up to 1.8 cm. Pancreas and pelvis treatments exhibited larger variations, with mean magnitudes around 1.2-1.5 cm and maximums up to 3.9 cm.
  • ATS Workflow is 2-3 Times Longer: The adaptive-to-shape (ATS) workflow took approximately two to three times longer than the adaptive-to-position (ATP) workflow.
  • MUs Slightly Increased in ATP for Prostate: For prostate patients, the ATP workflow showed a slight increase in MUs (up to 10%) with larger PSVs, while no significant trend was observed for pancreas patients or in the ATS workflow.
  • Dosimetric Outcomes Maintained: Both adaptive workflows generally maintained target coverage (D0.5cc within 3% and D99% within 2% of reference values) and OAR sparing, regardless of the simulated PSV magnitude.

Limitations and Future Work

  • Small Sample Size: The study used a limited number of IMRT plans (five per site), which may reduce the statistical power to detect subtle differences. Future work should involve larger cohorts.
  • Focus on Translational Shifts: The study primarily simulated translational shifts. The impact of rotational errors, which are not corrected by rigid registration, was highlighted as a significant limitation and an area for future investigation.
  • No Intra-fraction Motion: The study focused on inter-fraction setup variations and did not account for intra-fraction motion, which can further complicate treatment delivery.
  • Clinical Workflow Nuances: While the study used complex adaptive methods, real-world clinical workflows might involve adjustments to IMRT objectives that could influence outcomes, particularly for complex cases like pancreas treatments.

Future work should focus on developing methods to mitigate rotational errors, exploring more advanced adaptive strategies, and investigating the impact of PSVs in more complex treatment scenarios and with larger patient cohorts.

Summary updated

Abstract

Significant PSVs were observed on Elekta Unity due to suboptimal patient immobilization. Using prostate and pancreas treatments as examples, we demonstrated that adaptive planning can effectively accommodate such PSVs. Nevertheless, efforts should be made to minimize PSVs-particularly rotations-to mitigate intra-fraction motion and reduce treatment time.
BibTeX
@article{yan2025patient,
  author = {M. Yan and Erika Kollitz and Sheng‐Hsuan Sun and Kathryn Hitchcock and Alexandra De Leo and Amanda Schwarz and Luke Maloney and Jonathan Li and Chihray Liu and Guanghua Yan},
  title = {{Patient setup variation on Elekta Unity and its impact on adaptive planning}},
  journal = {Journal of Applied Clinical Medical Physics},
  year = {2025},
  doi = {10.1002/acm2.70016},
}
Figure