Beam geometry optimization and convergence mode evaluation for knowledge‐based planning of hippocampal avoidance whole‐brain radiotherapy: A cross‐platform study with quantitative scorecard assessment

BACKGROUND: Hippocampal avoidance whole-brain radiotherapy (HA-WBRT) is used to treat brain metastases while preserving cognitive function by sparing the hippocampi. Volumetric modulated arc therapy (VMAT) is commonly used due to its ability to generate conformal dose distributions. However, treatment planning remains challenging due to the need for sharp dose fall-off near centrally located hippocampi without compromising whole-brain coverage. PURPOSE: This study evaluated the dosimetric impact of a four-arc hybrid geometry (4A-HG) and convergence mode settings for HA-WBRT using a knowledge-based planning (KBP) workflow across two linac platforms. METHODS: Twenty HA-WBRT patients prescribed 30 Gy in 10 fractions were retrospectively studied. Plans were generated using a KBP model on two linac platforms with different collimation systems: Halcyon and Edge. Four arc configurations were evaluated, including three open-field geometries with different numbers of arcs (2A, 3A, and 4A) and 4A-HG. The 4A-HG consists of two open-field arcs and two partial-field arcs with superior-inferior field restriction near the hippocampal region, which allows improved dose modulation near the hippocampus to potentially enhance hippocampal sparing. Jaw tracking was enabled for all Edge plans. Each arc configuration was combined with three convergence modes ("Off," "On," and "Extended") on both linac platforms, yielding 24 plans per patient. Plans were normalized so that 95% of the planning target volume (PTV) received 100% of prescription dose. Plan quality was evaluated using Radiation Therapy Oncology Group (RTOG) 0933 metrics, as well as by a dosimetric scorecard incorporating PTV D100% and the Paddick conformity index (PCI). Statistical comparisons were performed using Friedman tests with Holm correction. Patient-specific quality assurance (QA) was performed for the five highest-monitor-unit (MU) plans on each platform. RESULTS: The proposed 4A-HG consistently improved target coverage and hippocampal sparing compared with conventional open-field arc configurations. Notably, PTV D100% was significantly improved with 4A-HG on the Edge with the "On" and "Extended" convergence modes (all p < 0.05 compared with other arc configurations). For example, with the "On" convergence mode, mean PTV D100% across 20 patients increased from 19.433 Gy with 2A to 20.011 Gy with 4A-HG (p = 0.009). The 4A-HG also achieved the highest mean PCI using the "Extended" mode (0.940 on Halcyon, 0.944 on Edge). Hippocampus Dmax was significantly reduced with 4A-HG on Halcyon across all modes (p < 0.01) and on Edge for "On" and "Extended" modes (p < 0.05). On Halcyon, the mean hippocampus Dmax decreased from 14.446 Gy to 13.662 Gy when changing the arc configuration from 2A to 4A-HG using the "Extended" mode, with a similar reduction observed on Edge (12.945 Gy to 12.566 Gy). More extensive convergence mode settings further improved plan quality, with the highest mean score (157/184) achieved on Edge using 4A-HG under the "Extended" mode. These improvements were associated with increased MUs; however, all QA results met institutional criteria. CONCLUSIONS: The 4A-HG arc configuration and the more extensive convergence mode settings improved plan quality for HA-WBRT within a KBP workflow across two linac platforms. These findings support the clinical implementation of the proposed planning strategies for VMAT-based HA-WBRT.