PurposeAs stereotactic ablative radiotherapy (SABR) is being used to treat greater numbers of lung metastases, selecting the optimal dose and fractionation to balance local failure and treatment toxicity becomes increasingly challenging. Multi-lesion lung SABR plans include spatially diverse lesions with heterogenous prescriptions and interacting dose distributions. In this study, we developed and evaluated a generative adversarial network (GAN) to provide real-time dosimetry predictions for these complex cases. Methods and MaterialsA GAN was trained to predict dosimetry on a dataset of patients who received SABR for lung lesions at a tertiary center. Model input included the planning CT scan, the organs at risk and (OARs) target structures, and an initial estimate of exponential dose fall-off. Multi-lesion plans were split 80/20 for training and evaluation. Models were evaluated on voxel-voxel, clinical dose-volume-histogram, and conformality metrics. An out-of-sample validation and analysis of model variance were performed. ResultsThere were 125 multi-lesion plans from 102 patients with 357 lesions. Patients were treated to 2-7 lesions, with 19 unique dose-fractionation schemes over 1-3 courses of treatment. The out-of-sample validation set contained an additional 90 plans from 80 patients. The mean absolute difference (MAD) and gamma pass fraction (GPF) between the predicted and true dosimetry was <3 Gy and > 90% for all OARs. The absolute differences in lung V20 and CV14 were 1.40±0.99% and 75.8±42.0 cc respectively. The ratios of predicted to true R50%, R100% and D2cm were 1.00±0.16, 0.96±0.32, and 1.01±0.36 respectively. The out-of-sample validation set maintained MAD and GPF of <3 Gy and >90% for all OARs. The median standard deviation of variance in V20 and CV14 prediction was 0.49% and 22.2 cc respectively. ConclusionsA GAN for predicting the 3-D dosimetry of complex multi-lesion lung SABR is presented. Rapid dosimetry prediction can be used to assess treatment feasibility and explore dosimetric differences between varying prescriptions.