Metal Artifact Removal on Dental CT Scanned Images by Using Multi-Layer Entropic Thresholding and Label Filtering Techniques for 3-D Visualization of CT Images

Abstract

Metal artifact is a significant problem in computed tomography (CT). Degradation of the quality of an image is a direct consequence of the metal artifacts in the image data. A number of papers and articles have been published regarding the subjects. However, none of the approach can be employed to incorporate in commercial CT scanners. Besides, all of the methods, to the best of our knowledge, are performed in the reconstruction process which means the CT images are created at the same time as the metal artifact removal process. In this research, we assume that the CT images with metal artifacts are given and we have no control over the reconstruction approach.Hence, we propose a new method to automatically remove metal artifacts in dental CT images in the post-processing steps. The proposed technique consists of two main steps. First, the local entropy thresholding scheme is employed to automatically segment out the dental region in a CT image. Then, the Label filtering technique is used to remove isolated pixels, which are the metal artifacts, by using the concept of connected pixel labeling.The algorithm has been tested on thirty sets of the dental CT scanned images. The experimental results are compared with hand-labeled dental images and are evaluated in terms of accuracy, sensitivity and specificity. The numerical results showing sensitivity, specificity, and accuracy are 87.89%, 99.54%, and 99.21% respectively. The experiments demonstrate the robustness and effectiveness of the proposed algorithm. The algorithm provides promising performance in detecting and removing metal artifacts from dental CT images. Therefore, automatic artifact removal can greatly help with the 3-D Visualization of CT images.KeywordsArtifact RemovalCT scanned imageDentistryEntropic ThresholdingLabel Filtering