(Tuesday, 7/31/2018) 3:45 PM - 4:15 PM
Room: Exhibit Hall | Forum 4
Purpose: To evaluate the feasibility of generating high quality plan based on three dimensional (3D) predicted dose distribution.
Methods: A novel deep-learning approach is developed that can predict the achievable 3D dose distribution for the new patient based on the geometrical features. The predicted dose distribution is used as the optimization objective function to generate high quality plans automatically. Ten head-and-neck patients with different prescription doses, target numbers and positions are enrolled in this study. For each patient, the dose distribution from automated plan (Auto-plan) is compared to the manually optimized plan (MO-plan) and the predicted plan from deep learning method (Pred-plan). Additionally, we find that the automated plan can also be refined or re-optimized by adding more objective functions for further plan quality improvement.
Results: Compared to the MO-plan, the Auto-plan does not show significant statistical differences. After re-optimization, the re-optimized plans (Re-O plan) show no difference compared to the MO-plans but with better OAR dose sparing.
Conclusion: A deep-learning based dose prediction model for IMRT treatment is established. A novel method of implementing the predicted 3D dose distribution in the treatment planning is realized. The planning process and efficiency is improved by reducing the contouring of dose shaping structures. This new method will help to further realize the goal of automatic treatment planning.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
