Objective To apply deep neural networks to predict high-precision dose distribution in volume modulated arc therapy(VMAT)plans for non-small cell lung cancer(NSCLC)patients.Methods This study developed a U-shaped network called MHA-resunet,which incorporated a large kernel dilated convolution module and a multi-head attention module.The model was trained from 151 VMAT plans of NSCLC patients.CT images,planning target volume(PTV)and organs at risk(OARs)were fed into the independent input channel.The dose distribution was taken as the output to train the model.The performance of this network was compared with that of several commonly used networks,and the networks'performance was evaluated based on the voxel-level mean absolute error(MAE)within the PTV and OARs,as well as the error in clinical dose-volume metrics.Results The MAE between the dose distribution predicted by MHA-resunet network and the manually planned dose distribution within the PTV area was 1.51 Gy,and the D98 and D95 errors in the PTV area were both＜1 Gy.Compared with the other three commonly used networks,the dose error of the MHA-resunet was the smallest in the PTV area and in OARs except for the heart.Conclusions The proposed MHA-resunet network improves the receptive field to learn the relative spatial relationship between the PTV area and the OARs,enabling accurate prediction of dose distribution in NSCLC patients undergoing VMAT radiotherapy.
万方数据