Automatic Segmentation of Organs-at-Risk from Head-and-Neck CT using Separable Convolutional Neural Network with Hard-Region-Weighted Loss

Abstract Accurate segmentation of Organs-at-Risk (OAR) from Head and Neck (HAN) Computed Tomography (CT) images with uncertainty information is critical for effective planning of radiation therapy for Nasopharyngeal Carcinoma (NPC) treatment. Despite the state-of-the-art performance achieved by Convolutional Neural Networks (CNNs) for the segmentation task, existing methods do not provide uncertainty estimation of the segmentation results for treatment planning, and their accuracy is still limited by the low contrast of soft tissues in CT, highly imbalanced sizes of OARs and large inter-slice spacing. To address these problems, we propose a novel framework for accurate OAR segmentation with reliable uncertainty estimation. First, we propose a Segmental Linear Function (SLF) to transform the intensity of CT images to make multiple organs more distinguishable than existing simple window width/level-based methods. Second, we introduce a novel 2.5D network (named as 3D-SepNet) specially designed for dealing with clinic CT scans with anisotropic spacing. Thirdly, we propose a novel hardness-aware loss function that pays attention to hard voxels for accurate segmentation. We also use an ensemble of models trained with different loss functions and intensity transforms to obtain robust results, which also leads to segmentation uncertainty without extra efforts. Our method won the third place of the HAN OAR segmentation task in StructSeg 2019 challenge and it achieved weighted average Dice of 80.52% and 95 % Hausdorff Distance of 3.043 mm. Experimental results show that 1) our SLF for intensity transform helps to improve the accuracy of OAR segmentation from CT images; 2) With only 1/3 parameters of 3D UNet, our 3D-SepNet obtains better segmentation results for most OARs; 3) The proposed hard voxel weighting strategy used for training effectively improves the segmentation accuracy; 4) The segmentation uncertainty obtained by our method has a high correlation to mis-segmentations, which has a potential to assist more informed decisions in clinical practice. Our code is available at https://github.com/HiLab-git/SepNet.