Room: Stars at Night Ballroom 1
Purpose: We present a real-time Pareto surface dose generation deep learning neural network that can be used immediately after segmentation by the physician. This adds a tangible and quantifiable endpoint for the physician to portray to the planner, potentially improving plan quality and saving planning time.
Methods: From 70 prostate patients, we first generated 84,000 pseudo random 7-beam IMRT plans sampling the Pareto surface, representing various tradeoffs between the planning target volume (PTV) and the various organs at risk (OAR), including bladder, rectum, left femur, right femur, and body. Ring and skin structures were added as tuning structures. We divided the data to 10 test, 54 training, and 6 validation patients. We then trained a hierarchically densely connected convolutional U-net (HD U-net), to take the PTV and an avoidance map representing the OARs, and predict the optimized plan. We trained the model for 100,000 iterations using the Adam optimizer, with a learning rate of 1x10â?»â?´.
Results: The HD U-net is capable of accurately predicting the Pareto optimal 3D dose distributions, with mean dose errors of 3.4% (PTV), 1.6% (bladder), 3.7% (rectum), 3.2% (left femur), 2.9% (right femur), and 0.04% (body) of the prescription dose, as compared to the optimized plans. The PTV dose coverage prediction was also very similar, with errors of 1.3% (D98) and 2.0% (D99) of the prescription dose. Given any structure weights set, the neural network is capable of predicting the dose distribution in 1.7 seconds. Clinically, the optimization and dose calculation for IMRT takes approximately 5-10 minutes to complete.
Conclusion: We expect that the implementation of such a framework would drastically reduce the number of feedback loops between the planner and physician. The valuable time saved would allow for the physician and planner to focus on more challenging cases and produce the best achievable plan.
Funding Support, Disclosures, and Conflict of Interest: This study is supported by Cancer Prevention and Research Institute of Texas (CPRIT) (IIRA RP150485, MIRA RP160661)