2588 - Evaluation of Biological Dose DVH Estimates from a Commercially Available DVH Estimation Algorithm for Head and Neck Cancer Proton Radiation Therapy

Purpose/Objective(s): Proton radiation therapy may have enhanced relative biologic effectiveness based on increased LET at the distal edge of the spread-out Bragg peak. Our goal was to evaluate the accuracy of DVH estimates from a commercially available DVH estimation algorithm compared to ground truth Monte Carlo calculated biological dose that accounts for LET in head and neck cancer patients treated with spot-scanning proton radiation therapy.

Materials/Methods: Eighty-seven bilateral oropharynx cancer patients treated at a large American academic center using spot-scanning proton therapy were identified to construct the DVH estimation model. All subjects had treatment plan biological dose calculated using an in-house, validated Monte Carlo calculation algorithm. The biological dose was used in the DVH estimation model training. Eleven additional patients not included in model training were used to evaluate the accuracy of DVH band estimates. The accuracy of the model was measured by determining if ground truth clinical metrics fell within the model’s DVH estimation bands. DVH metrics included brainstem D_max and V_30%, esophagus D_mean and V_55%, larynx D_mean and V_50%, mandible D_max, spinal cord D_max, and brachial plexus D_max.

Results: The model's DVH estimate bands accurately captured most delivered clinical DVH metrics. Larynx D_mean and V_50% had the most ground truth clinical metrics within DVH estimates with nine out of eleven metrics each within estimate bands. Mandible D_max had the smallest number of metrics within DVH estimate bands with two out of eleven plans within estimates. However, mandible DVH estimate bands were the smallest of all metrics studied.

Conclusion: The constructed DVH estimation model showed agreement with ground truth DVH metrics for most anatomical structures and DVH metrics studied. DVH estimation models have the potential to estimate biological dose to anatomical structures before time-consuming treatment planning and Monte Carlo calculations have been completed and allow for swift treatment planning objective decision making.