3D segmentation of nasopharyngeal carcinoma from CT images using cascade deep learning.

摘要: Abstract In the paper, we propose a new deep learning-based method for segmenting nasopharyngeal carcinoma (NPC) in nasopharynx from three orthogonal CT images. The proposed introduces cascade strategy composed of two-phase manners. images, there are organs, called non-target which NPC never invades. Therefore, first phase is to detect and eliminate organ regions second phase, extracted remained Convolutional neural networks (CNNs) applied organs NPCs. system determines final segmentation by integrating results obtained coronal, axial sagittal Moreover, construct two CNN-based detection systems using one kind overlapping patches with fixed size various different sizes. From experiments images 70 patients, our systems, especially patches, achieves best performance detecting compared conventional methods.