Room: Karl Dean Ballroom C
Purpose: To develop a knowledge-guided planning strategy for efficient adaptive planning to reduce organ-at-risk (OAR) dose during intensity modulated radiotherapy (IMRT) of oropharynx cancer.
Methods: A dose volume histogram(DVH) prediction model was formulized for oropharynx IMRT with sequential boost scheme (44/50Gy to low-risk PTV, 70Gy to high-risk PTV, 2Gy/fraction). OAR anatomical factors and DVH features contributing to OAR dose sparing were identified and analyzed. Acquired knowledge from 142 previous plans were studied in the model. Sixteen patients prescribed with sequential boost underwent Â¹â?¸FDG-PET and contrasted-CT after 24Gy for adaptive planning. The model yielded DVH predictions and confidence intervals of 4 OARs (pharynx, larynx, oral cavity, contralateral parotid), which were subsequently provided to human planners as guidance for adaptive planning. Adapted plans were generated using the same beam settings as original plans with prioritized optimization constraints predicted by the model. OAR mean dose reductions predicted by the model and that achieved by the adapted plan were analyzed respectively against the original plan. Dosimetric differences between model predictions and adapted plans were compared. Significance was set to p<0.05 in Wilcoxon signed rank test.
Results: The formulized model predicted mean dose reductions to pharynx (1.42Â±0.61Gy, p=0.02), larynx (0.82Â±0.48Gy, p=0.04), oral cavity (1.80Â±0.94Gy, p=0.04) and contralateral parotid (0.46Â±0.42Gy, p=0.06). Model prediction-guided adapted plans successfully reduced mean dose of pharynx (2.64Â±0.44Gy, p<0.01), larynx (1.06Â±0.44Gy, p=0.01), oral cavity (2.59Â±0.52Gy, p<0.01) and contralateral parotid (3.18Â±0.75Gy, p<0.01). No significant difference was observed between model predictions and adapted plans except contralateral parotid (p<0.01). Planning time per patient was about 15-20min, which is shorter than current standard workflow.
Conclusion: The developed knowledge-guided DVH prediction model accurately predicted OAR dose reduction in oropharynx adaptive planning. Using model predictions as guidance, adapted plans could be generated with effective OAR dose reduction and improved planning efficiency.