S. Reid1, X. Li1, Q. J. J. Wu2, and Y. Sheng1; 1Duke University Medical Center, Durham, NC, 2Duke University, Durham, NC
Purpose/Objective(s): To improve the output of the Head-and-Neck (HN) fluence map generator using Diffusion-GAN. Intro: The current head-and-neck (HN) fluence map generator tends to produce highly modulated fluence maps and therefore high monitor units (MUs) for each beam, which leads to more delivery uncertainty and leakage dose. This project implements diffusion into the training process to mitigate this effect.Materials/
Methods: The dataset consists of 200 head-and-neck (HN) patients receiving intensity modulated radiation therapy (IMRT) for training, 16 for validation, and 15 for testing. Two models were trained, one with diffusion and one without. The original model was a conditional generative adversarial network (GAN) written in TensorFlow, the model without diffusion was written to be the PyTorch equivalent of the original model. After confirming the model was properly converted to PyTorch by comparing outputs, both new models were modified to use binary cross entropy for the GAN loss and mean absolute error as a third loss function for the generator. Hyperparameters were carefully selected based on the training script for the original model, and further tuned with trial and error. The diffusion was implemented based on Diffusion-GAN and the associated GitHub repository. The two new models were compared by plotting training loss vs epoch over 500 epochs. The two models were compared to the original model by comparing the output fluence maps to the ground truth using similarity index and comparing DVH statistics among the three models. Results: The diffusion model and no diffusion model achieved similar training loss and similar organ-at-risk (OAR) DVH statistics. The diffusion model and no diffusion model had consistently delivered better parotid sparing than the original model. The diffusion model had the least MUs: 23% less than the original model and 3% less than the no diffusion model. The diffusion model had lower D2cc: 4% less than the original model and 1% less than the no diffusion model on average. The diffusion model had lowerPTV V110% on average: 29% less than the original model and 12% less than the no diffusion model. All three plans deliver 100% of the prescription dose to nearly the same percentage of PTV volume. Conclusion: Implementing diffusion does not provide significant impact on training time, training loss, or OAR dose. However, it does enable comparable dose performance to both the no diffusion and original models, while significantly reducing the total MUs, 3D max 2cc, and PTV44 V110% relative to these two models, indicating smoother fluence modulation. In addition, both new models reduced right and left parotid dose relative to the original model. This indicates that the diffusion model can reduce overall dose to the OARs, while preserving dose conformity around the target.
