Developing a MLC modifier program to improve fiducial detection for MV/kV imaging during hypofractionated prostate volumetric modulated arc therapy

Laura Happersett, Ping Wang, Pengpeng Zhang, James Mechalakos, Guang Li, Eleanor Eley, Michael J. Zeléfsky, G Mageras, Antonio L. Damato, Margie HuntJournal of Applied Clinical Medical Physics 2019 · DOI

TL;DR

This paper introduces MLC_MODIFIER, an automated script that modifies radiation therapy plans to improve the visibility of implanted fiducial markers during MV/kV imaging. This automation significantly reduces registration failures, a critical step for accurate image-guided radiotherapy, while minimally impacting treatment plan quality.

What Problem Does This Paper Address?

Image-guided radiation therapy (IGRT) for prostate cancer relies on accurately tracking the prostate's position during treatment. A common method involves implanting fiducial markers within or near the prostate, which are then visualized using X-ray (kV) and megavoltage (MV) imaging. Volumetric Modulated Arc Therapy (VMAT), a sophisticated radiation delivery technique, uses a moving multileaf collimator (MLC) to shape the radiation beam. However, the MLC's complex movements can inadvertently block the view of these fiducial markers on MV images, leading to "registration failures." These failures can necessitate treatment delays or compromises in accuracy. Manually adjusting VMAT plans to ensure fiducial visibility is time-consuming and prone to variability. This paper addresses the need for an automated, efficient, and consistent method to improve fiducial visibility during MV/kV imaging in VMAT for hypofractionated prostate treatments.

Methods

The authors developed an Eclipse scripting plug-in called MLC_MODIFIER. This program automates the process of modifying VMAT control points to expose fiducial markers. The core modifications involve:

  1. Angular Compression: Reducing the angular separation between treatment control points (TxCPs) to create smaller, more manageable segments.
  2. Insertion of "Imaging" Control Points (ImCPs): Introducing low-dose (approximately 3 MU) control points specifically for imaging. The dose delivered at these ImCPs is subtracted from the main treatment dose to maintain the overall treatment plan.
  3. MLC Aperture Modification: At these ImCPs, MLC_MODIFIER automatically adjusts the MLC positions to expose at least one fiducial marker with a 1 cm margin.

These modifications are applied to specific angles where imaging is triggered. The program evaluates MLC apertures and ensures fiducial visibility. The modified plans are then imported back into Eclipse for dose calculation and normalization. The authors compared this automated method to a manual modification approach where a physicist manually adjusted MLC positions for a few control points.

Datasets

The study evaluated data from 27 hypofractionated prostate cancer patients treated between October 2014 and November 2017. A total of 135 treatment sessions were analyzed, with 1960 registrations evaluated for the manually modified cohort and 2376 for the MLC_MODIFIER cohort.

Key Insights and Ideas

  • Automation is Key: The development of an automated script (MLC_MODIFIER) is crucial for streamlining a complex and time-consuming manual process, reducing human error and variability.
  • Imaging Control Points are Effective: Introducing dedicated low-dose "imaging" control points within the VMAT arc provides a window for fiducial visualization without significantly impacting the overall treatment dose delivery.
  • Minimal Plan Degradation: The automated modification process, while altering MLC positions, results in only minor changes to critical dosimetric parameters like PTV coverage and dose homogeneity, and negligible changes to organ-at-risk doses.
  • Fiducial Blockage is the Primary Failure Mode: The study highlights that MLC blockage of fiducials is a significant cause of registration failures, and the automated method effectively addresses this.

Key Results

The MLC_MODIFIER program demonstrated a significant improvement in fiducial detection and registration success rates compared to manual modifications:

  • Overall Registration Failures: Reduced from 33% (manual) to 8.9% (MLC_MODIFIER) (P < 0.0001).
  • Failures due to MLC Blockage: Drastically reduced from 13.6% (manual) to 0.1% (MLC_MODIFIER) (P < 0.0001).
  • Failures due to Unknown Reasons: Reduced from 7.6% (manual) to 2.4% (MLC_MODIFIER) (P < 0.0001).
  • PTV Maximum Dose: Increased by an average of 2.1% (manual) vs. 0.3% (MLC_MODIFIER) (P < 0.004).
  • PTV Dose Homogeneity (HI): Increased by an average of 22.0% (manual) vs. 3.3% (MLC_MODIFIER) (P < 0.004).
  • Organ-at-Risk Doses (Rectum, Bladder, Urethra D1CC): Changes were minimal (<0.7%) and similar between methods.
  • Time Savings: The MLC_MODIFIER script takes approximately 10 minutes to run, compared to 45-75 minutes for a manual modification by a physicist.

Limitations and Future Work

  • Manual Margin Variability: The manual modification method had variable margins (0-5 mm), whereas MLC_MODIFIER used a fixed 1 cm margin. The authors hypothesize that increasing manual margins could yield similar results, but this was found to increase dose to critical structures and degrade PTV coverage.
  • Image Quality: While MLC_MODIFIER reduced failures due to blockage, failures due to poor DTS image quality remained similar between methods.
  • Future Work:
    • Investigate an optimal balance between margin size and detection rate to further improve plan quality.
  • Develop a next version of MLC_MODIFIER that dynamically adjusts MU at imaging control points based on patient separation to potentially improve image quality.
  • Further refine the modification process to ensure modified plans are as close as possible to the original plan quality.

The authors have already treated over 230 patients using the MLC_MODIFIER method, indicating its successful integration into clinical practice.

Summary updated

Abstract

Increasing fiducial visibility via an automated process comprised of angular compression of control points and insertion of additional "imaging" control points is feasible. Degradation of plan quality is minimal. Fiducial detection and registration success rates are significantly improved compared to manually edited apertures.
BibTeX
@article{happersett2019developing,
  author = {Laura Happersett and Ping Wang and Pengpeng Zhang and James Mechalakos and Guang Li and Eleanor Eley and Michael J. Zeléfsky and G Mageras and Antonio L. Damato and Margie Hunt},
  title = {{Developing a MLC modifier program to improve fiducial detection for MV/kV imaging during hypofractionated prostate volumetric modulated arc therapy}},
  journal = {Journal of Applied Clinical Medical Physics},
  year = {2019},
  doi = {10.1002/acm2.12614},
}
Figure