Author: Eric C. Ford, Yulun He, Minsun Kim, Dustin Melancon, Juergen Meyer, Dong Joo Rhee, Yinghua Tao 👨🔬
Affiliation: Department of Radiation Oncology, Fred Hutchinson Cancer Center, University of Washington, MD Anderson Cancer Center, University of Washington 🌍
Purpose: To develop and evaluate an automated-planning technique capable of generating high-quality treatment plans for hippocampal-sparing-whole-brain radiation therapy.
Methods: An auto-planning script for hippocampal-sparing-whole-brain VMAT (Rx: 30Gy/10fx) was developed in RayStation (RaySearch Laboratories, 2023B). The script prompts users to select the PTV, OARs, and beam model before autonomously generating treatment plans. It initializes beams, creates helper structures, and adds clinical goals and optimization objectives within approximately one minute. Notably, the script generates dedicated helper structures and employs well-balanced optimization objectives to guide plan generation.
The script then pauses to allow manual adjustment of isocenter if desired (e.g., aligning with BBs during CT simulation) to avoid setup shifts. This is the only intervention required. Once resumed, the script performs optimization iterations autonomously until a final plan is generated with no further human input.
The script was validated on seven anonymized datasets, including two simultaneous-integrated-boost (SIB, PTV35Gy) cases. Plan quality was assessed based on PTV coverage, uniformity, and hippocampal-sparing. Additional metrics, such as total monitor units (MU), PTV volume, and hippocampal volumes, were also evaluated.
Results: The script consistently generated high-quality plans, achieving PTV coverage (V30Gy: 95.7 ± 0.2%, V25Gy: 99.1 ± 0.1%), hippocampal-sparing (left/right D0.03cc: 1402.6 ± 38.9/1402.2 ± 40.7cGy; D100%: 815.9 ± 11.8/826.9 ± 12.9cGy), and global maximum dose (3298.9 ± 7.9cGy). Total MU was 1364.2 ± 98.7. PTV volume was 1559.6 ± 381.8cc, and hippocampal-avoidance volume was 34.3 ± 4.2cc.
For SIB cases, boost PTV(PTV35Gy) volumes were 55.46 cc and 10.78 cc. Both PTV30 and PTV35 coverage exceeded 95%, with hippocampal-sparing maintained.
Conclusion: The auto-planning script consistently produced high-quality plans across varying PTV and hippocampal sizes, significantly reducing human planning time (from days to under 1 hour). Its innovative framework streamlines workflows and ensures robust plan quality, offering substantial benefits to busy radiation oncology clinics.