Room: Exhibit Hall | Forum 3
Purpose: To identify the collimator angle and field size settings that, for a large cohort of patients, consistently produce the best head-and-neck (H&N) VMAT plans for knowledge-based automated optimization.
Methods: A generalized automated strategy for H&N VMAT optimization was developed to give the most clinically acceptable plans. We employed RapidPlan knowledge-based dose objectives for OARs, and generalized optimization weights for planning structures. This template was then used to create 4-arc VMAT plans for 27 H&N patients. Each plan consisted of two arcs with collimator angles of Â±5Â°, Â±10Â°, Â±20Â°, or Â±30Â°, and an x-direction field size of 15cm, 18cm, 20cm, or 22cm. Within the same plan, two additional arcs with 90Â° collimator angle and 15cm or 18cm z-direction field size were created where jaw positions were offset between the arcs to achieve complete cranial-caudal target coverage. As a result, 32 beam settings were generated per patient, and a total of 864 plans were analyzed. All plans were normalized such that 98% of the prescribed dose covered 95% of the PTV (D95 = 98%). Plan quality was compared via DVH metrics.
Results: All plans met clinical dose constraints. The average range of spinal cord maximum dose across all patients was 1.79 Gy. Small variation was also seen for other investigated planning structures (spinal cord +7mm, brainstem, and brainstem +7mm) as well as for mean contralateral parotid dose. For patients with many hot spots (average V105 > 5%), V105 varied significantly between plans. This is because the steepness of the DVH for PTVs made V105 very sensitive to slight changes in normalization. No trend for optimal beam settings could be identified.
Conclusion: The quality of knowledge-based automated H&N VMAT plans is not sensitive to the selection of collimator angle and field size.
Funding Support, Disclosures, and Conflict of Interest: This work was partially funded by the NCI (UH2 CA202665).