DeepDoseNet: A Deep Learning model for 3D Dose Prediction in Radiation\n Therapy

Abstract

The DeepDoseNet 3D dose prediction model based on ResNet and Dilated DenseNet\nis proposed. The 340 head-and-neck datasets from the 2020 AAPM OpenKBP\nchallenge were utilized, with 200 for training, 40 for validation, and 100 for\ntesting. Structures include 56Gy, 63Gy, 70Gy PTVs, and brainstem, spinal cord,\nright parotid, left parotid, larynx, esophagus, and mandible OARs. Mean squared\nerror (MSE) loss, mean absolute error (MAE) loss, and MAE plus dose-volume\nhistogram (DVH) based loss functions were investigated. Each model's\nperformance was compared using a 3D dose score, $\\bar{S_{D}}$, (mean absolute\ndifference between ground truth and predicted 3D dose distributions) and a DVH\nscore, $\\bar{S_{DVH}}$ (mean absolute difference between ground truth and\npredicted dose-volume metrics).Furthermore, DVH metrics Mean[Gy] and D0.1cc\n[Gy] for OARs and D99%, D95%, D1% for PTVs were computed. DeepDoseNet with the\nMAE plus DVH-based loss function had the best dose score performance of the\nOpenKBP entries. MAE+DVH model had the lowest prediction error (P<0.0001,\nWilcoxon test) on validation and test datasets (validation:\n$\\bar{S_{D}}$=2.3Gy, $\\bar{S_{DVH}}$=1.9Gy; test: $\\bar{S_{D}}$=2.0Gy,\n$\\bar{S_{DVH}}$=1.6Gy) followed by the MAE model (validation:\n$\\bar{S_{D}}$=3.6Gy, $\\bar{S_{DVH}}$=2.4Gy; test: $\\bar{S_{D}}$=3.5Gy,\n$\\bar{S_{DVH}}$=2.3Gy). The MSE model had the highest prediction error\n(validation: $\\bar{S_{D}}$=3.7Gy, $\\bar{S_{DVH}}$=3.2Gy; test:\n$\\bar{S_{D}}$=3.6Gy, $\\bar{S_{DVH}}$=3.0Gy). No significant difference was\nfound among models in terms of Mean [Gy], but the MAE+DVH model significantly\noutperformed the MAE and MSE models in terms of D0.1cc[Gy], particularly for\nmandible and parotids on both validation (P<0.01) and test (P<0.0001) datasets.\nMAE+DVH outperformed (P<0.0001) in terms of D99%, D95%, D1% for targets.\nMAE+DVH reduced $\\bar{S_{D}}$ by ~60% and $\\bar{S_{DVH}}$ by ~70%.\n