A Flexible Rapidplan Model for Optimization of High-Grade Glioma Radiotherapy Plans with and without Hippocampal Sparing
Abstract
Purpose
Evidence from randomized trials demonstrates improved memory, cognition, and quality of life in patients whose hippocampi are spared during radiotherapy (RT) to the brain. We present a RapidPlan (RP) model for high-grade glioma RT planning that includes hippocampal sparing (HS), and we demonstrate that the inclusion of HS minimally impacts clinical plan quality.
Methods
The model was trained on 53 high-grade glioma RT plans previously planned and delivered at our institution which were replanned with HS by a single expert dosimetrist. For evaluation, the trained model was used to define objectives, and the optimization was allowed to run without user input for twelve cases. Plans were generated both with and without HS and compared against each other and the dosimetrist’s HS plans. Where the target volumes overlapped the ipsilateral hippocampus, only the contralateral hippocampus was spared. Statistical significance was tested using a Wilcoxon signed-rank test with p<0.05 threshold.
Results
In three out twelve HS plans, the dosimetrist was able to meet clinical objectives that RP could not. However, only the PTV/Global Dmax demonstrated a statistically significant difference (p=0.001). The RP HS model achieved statistically lower overall Optic Nerve Dmax compared to the dosimetrist (p=0.01). When RP with HS was used, only two of twelve plans were unable to meet coverage objectives that were met by non-HS RP plans, however, these coverage differences were less than 0.5%. In general, non-HS RP plans had significantly lower Dmax values for chiasm (p=0.03), eyes (p=0.001), lenses (p<0.001) and optic nerves (p<0.001), but HS plans still met the same clinical planning objectives.
Conclusion
Hippocampal sparing in RT for high-grade glioma can be achieved using an RP-informed planning approach without compromising clinical plan quality.