Room: AAPM ePoster Library
Purpose: Incorporate model-derived importance data of spatial regions within parotid glands into the optimization of head-and-neck treatment plans.
Methods: Contralateral parotid gland contours for 5 head-and-neck patients were partitioned into 18 sub-regions of equivalent volume, and previously derived importance data  for each sub-region was used to quantify the relative importance of these sub-regions. Artificial base dose plans were constructed to penalize spatial dose delivered within the parotid gland in proportion to the relative importance. Base plan dose within each sub-region was uniform with magnitude determined by relative importance. The most important sub-region received 20Gy.
Standard institutional protocols were followed to retrospectively create treatment plans in the Varian Eclipse?? test system. Whole parotid mean doses were minimized. Plans were then re-optimized using base plans and an additional upper bound constraint imposed on the contralateral parotid. This modified technique imposes a spatially varying dose constraint, determined by relative importance of sub-regions.
Plans created with/without base doses were compared to determine if incorporating importance information into treatment planning could significantly lower dose to important regions of the parotid while continuing to meet other planning constraints. Improvements were assessed using a predictive model for stimulated salivary output loss at one-year post-RT .
Results: Mean dose to important sub-regions, as well as whole-parotid mean dose was found to decrease in all 5 cases when artificial base plans were introduced. Mean dose to the most important sub-region was found to significantly decrease using this new technique (paired t(8)=5.49, p<0.006). Improvements in salivary output retention of up to 20% of baseline were predicted.
Conclusion: This study has demonstrated that incorporating sub-regional importance data into the treatment planning process can significantly lower dose to important sub-regions. This technique has the potential to improve post-therapy patient outcomes.
 Clark et al., doi:10.1088/2057-1739/aac8ea
 Clark et al., doi: 10.1118/1.4915077
Not Applicable / None Entered.
TH- External Beam- Photons: Treatment planning using machine learning/Knowledge Based Planning/automation