Automatic Segmentation of Parotids in Head and Neck CT Images using Multi-atlas Fusion

Treatment planning for high precision radiotherapy of head and neck (H&N) cancer patients requires accurate delineation of many critical structures. Manual contouring is tedious and often suffers from large interand intra-rater variability. To reduce manual labor, we have previously developed a fully automated, atlas-based method for H&N CT image segmentation [1, 2]. In this work, we adapt the previous method and apply it to tackle the parotid segmentation problem in the MICCAI 2010 H&N Segmentation Challenge. The proposed method applies a hybrid deformable image registration to map parotid labels from an atlas image to the subject, the result of which is then refined using a deformable surface model approach. Segmentation fusion using multipleatlases is also employed to further improve the segmentation accuracy. Validation results on eight clinical datasets distributed by the MICCAI workshop showed that the proposed method gave accurate segmentation results, with a volume overlap above 85% for most subjects.