Automatic Segmentation of the Mandible for Three-Dimensional Virtual Surgical Planning

Abstract

Three-dimensional (3D) medical imaging techniques have a fundamental role in the field of oral and maxillofacial surgery (OMFS). 3D images are used to guide diagnosis, assess the severity of disease, for pre-operative planning, per-operative guidance and virtual surgical planning (VSP). In the field of oral cancer, where surgical resection requiring the partial removal of the mandible is a common treatment, resection surgery is often based on 3D VSP to accurately design a resection plan around tumor margins. In orthognathic surgery and dental implant surgery, 3D VSP is also extensively used to precisely guide mandibular surgery. Image segmentation from the radiography images of the head and neck, which is a process to create a 3D volume of the target tissue, is a useful tool to visualize the mandible and quantify geometric parameters. Studies have shown that 3D VSP requires accurate segmentation of the mandible, which is currently performed by medical technicians. Mandible segmentation was usually done manually, which is a time-consuming and poorly reproducible process. This thesis presents four algorithms for mandible segmentation from CT and CBCT and contributes to some novel ideas for the development of automatic mandible segmentation for 3D VSP. We implement the segmentation approaches on head and neck CT/CBCT datasets and then evaluate the performance. Experimental results show that our proposed approaches for mandible segmentation in CT/CBCT datasets exhibit high accuracy.