Combining CNN and Hybrid Active Contours for Head and Neck Tumor Segmentation in CT and PET images
This method offers an incremental improvement in tumor segmentation accuracy for clinicians and researchers in head and neck cancer radiomics.
This paper proposes an automatic method for segmenting head and neck tumors from PET and CT images, combining a multi-channel 3D U-Net with hybrid active contours. The method achieved second place in the MICCAI 2020 HECKTOR challenge, with a Dice Similarity Coefficient of 0.752, precision of 0.838, and recall of 0.717.
Automatic segmentation of head and neck tumors plays an important role in radiomics analysis. In this short paper, we propose an automatic segmentation method for head and neck tumors from PET and CT images based on the combination of convolutional neural networks (CNNs) and hybrid active contours. Specifically, we first introduce a multi-channel 3D U-Net to segment the tumor with the concatenated PET and CT images. Then, we estimate the segmentation uncertainty by model ensembles and define a segmentation quality score to select the cases with high uncertainties. Finally, we develop a hybrid active contour model to refine the high uncertainty cases. Our method ranked second place in the MICCAI 2020 HECKTOR challenge with average Dice Similarity Coefficient, precision, and recall of 0.752, 0.838, and 0.717, respectively.