Automatic nasopharyngeal carcinoma segmentation in MR images with convolutional neural networks

Abstract

Automatic and reliable segmentation of Nasopharyngeal Carcinoma (NPC) is an important but difficult task for various clinical applications such as Nasopharyngeal Carcinoma radiotherapy. In this paper, a novel automatic nasopharyngeal carcinoma segmentation method using deep Convolutional Neural Network (CNN) is proposed. Three deep single-view CNNs are trained separately on patches extracted from axial, sagittal and coronal view respectively, then their predicted classification information are integrated for NPC segmentation. The proposed method has been evaluated on the T1-weighted MRI images of 18 NPC patients using leave-one-out method. Both qualitative and quantitative results show significant improvement compared to currently used method such as graph cut.