基于带孔U-net神经网络的肺癌危及器官并行分割方法

Parallel segmentation method for organs at risk in lung cancer based on dilated U-net neural network

为了提高肺癌放疗计划危及器官勾画的精度和效率,提出了一种基于带孔U-net神经网络的肺癌放疗计划危及器官肺及心脏的并行分割方法.首先,构建了肺窗、心脏窗以及纵膈窗下的三通道伪彩色图像数据集,将图像数据集分成训练集、验证集以及测试集;然后,搭建了带孔U-net神经网络,利用训练集和验证集对其进行训练和参数调优;最后,利用测试集对训练后的带孔U-net神经网络进行图像分割性能评价,并与U-net神经网络及3种传统图像分割算法进行比较.实验结果表明,带孔U-net神经网络分割性能最优,可有效地完成肺及心脏的自动并行分割,提高勾画效率,分割结果与人工勾画结果相当.

To improve the accuracy and efficiency of the delineation of organs at risk in lung cancer radiotherapy planning, a parallel segmentation method for heart and lung based on the dilated U-net neural network for lung cancer radiotherapy planning was proposed. First, the three-channel pseudo color image data set was constructed with lung window, heart window and mediastinal window, and it was split into the training set, the validation set and the test set. Then, the dilated U-net neural network was built, trained and optimized by the training set and the validation set. Finally, the segmentation performance of the dilated U-net neural network was evaluated by the test data set, and compared with those of the U-net neural network and three traditional image segmentation algorithms. The experimental results show that the dilated U-net neural network has the best segmentation performance. It can effectively achieve the automatic parallel segmentation of lung and heart. The delineation efficiency can be improved and the segmentation result is approximate to the manual segmentation result.