Performance Comparison of Bone Segmentation on Dental CT Images

Abstract

Computed tomography (CT) images used in the dental application mainly consist of bone and soft tissue regions. Segmentation of bones from the soft tissues can help clinicians to assess the quality of the patient’s jawbone more easily. Moreover, the identifiable bone region can help creating the better results of 3D surface/volume rendering. In the dental CT images, the bone regions are usually surrounded by other regions which are in a larger area and darker colors. The application of Local Entropy Equalization is usually suitable for the images with the mentioned characteristics. It has been reported that different thresholding techniques, namely Otsu and Kittler, are used to segment the bone regions. In this paper, we present the experimental studies for bone segmentation of CT images in the dental application using three different techniques based on i) Otsu’s thresholding, ii) Kittler’s thresholding, and iii) Local Entropy thresholding. The performances of these three different techniques on the bone segmentation and 3D rendering, which are applied on several CT datasets, are compared and presented. The experimental results show that segmentation of bones using the Local Entropy thresholding technique provides better quality than the Otsu’s and Kittler’s methods.