Predicting voxel-level dose distributions for esophageal radiotherapy using densely connected network with dilated convolutions.

This work aims to develop a voxel-level dose prediction framework by integrating distance information between PTV and OARs as well as image information into a densely-connected network (DCNN). At first, a four-channel feature map consisting of PTV image, OARs image, CT image, and distance image is constructed. Then a densely connected neural network is built and trained for voxel-level dose prediction. Considering that the shape and size of OARs are highly inconsistent, the dilated convolution is employed to capture features from multiple scales. Finally, the propose network is evaluated with five-fold cross-validation on ninety-eight clinically approved treatment plans. The voxel-level mean absolute error (MAEV) of DCNN was 2.1% for PTV, 4.6% for left lung, 4.0% for right lung, 5.1% for heart, 6.0% for spinal cord, and 3.4 % for body, which outperforms conventional U-Net, Resnet-antiResnet, U-Resnet-D by 0.1-0.8%. This result shows that with the introduction of distance image and DCNN model, the accuracy of predicted dose distribution could be improved significantly. It provides a new dose prediction tool in supporting quality assurance and automation of treatment planning in esophageal radiotherapy.