Segmenting Tooth Components in Dental X-Ray Images Using Gaussian Kernel- Based Conditional Spatial Fuzzy C-Means Clustering Algorithm

Abstract

Tooth component segmentation is a crucial task in computer-aided design for forensic odontology, especially to estimate chronological age. Tooth segmentation on radiographic data is a very challenging task due to noise, low contrast, and uneven illumination. The Fuzzy C-Means clustering is generally used for image segmentation that allow pixels to be classified into one or more clusters according to their membership value. However, this clustering method still has problems associated with the shifting of cluster centers and sensitivity to the overlapping intensity distributions between classes. This paper proposes a modified strategy of the conditional spatial Fuzzy C-Means (csFCM) that incorporates global and spatial information into a weighted membership function by replacing the Euclidean distance with the Gaussian kernel distance to increase insensitivity to noise and outliers. The aim of this paper is to divide the tooth into 3 components, i.e. enamel, dentine, and pulp. Therefore, this modified algorithm is preceded by dental X-ray image pre-processing and continued by combining each dental component clusters into one composite image. The tooth image is pre-processed using Contrast Limited Adequate Histogram Equalization (CLAHE) and gamma adjustment to enhance the dental X-ray images quality from the non-uniform lighting. The Gausian kernel-based conditional spatial Fuzzy C-Means (GK-csFCM) segments the dental image into four class clusters, namely enamel, dentine, pulp, and background. Through iterations, the resulting cluster centers are more convergent with real cluster centers, thus ensuring the proposed method improves the drawback of inherent FCM-based methods and further promoting image segmentation performance. The experimental results on the real dental X-ray images showed that GK-csFCM has better performance than the typical FCM and csFCM clustering algorithms in terms of both qualitative and quantitative metrics, i.e. accuracy, specificity, sensitivity, and precision.

Descriptive analysis of dental X-ray images using various practical methods: A review

“…In Mahdi & Kobashi (2018) , it sets a multi-threshold by applying quantum particle swarm optimization to improve the accuracy. Fariza et al (2019) employed a method to extract dentin, enamel, pulp, and other surrounding dental structures using conditional spatial fuzzy C-means clustering. Subsequently, the performance improved as compared to inherently used FCM approaches.…”

“…Adaptive histogram equalization and median filtering are combinedly applied. (Fariza et al, 2019) Dental X-ray image is processed using CLAHE, and gamma correction is done to improve the contrast. (Avuçlu & Bacsçiftçi, 2020) Median softening filter applied after contrast stretching.…”

“…The dental X-ray image analysis methods can be categorized in several categories: region growing techniques, edge detection methods, thresholding based, clustering techniques, level set, and active contour, etc., are presented in ‘Image processing methods for dental image analysis’ ( Mahoor & Abdel-Mottaleb, 2004 ; Zhou & Abdel-Mottaleb, 2005 ; Nomir & Abdel-Mottaleb, 2005 , 2007 ; Gao & Chae, 2008 ; Oprea et al, 2008 ; Patanachai, Covavisaruch & Sinthanayothin, 2010 ; Harandi & Pourghassem, 2011 ; Hu et al, 2014 ; Amer & Aqel, 2015 ; Zak et al, 2017 ; Avuçlu & Bacsçiftçi, 2020 ) ( Rad et al, 2015 ; Tuan, Ngan & Son, 2016 ; Poonsri et al, 2016 ; Son & Tuan, 2016 , 2017 ; Ali et al, 2018 ; Alsmadi, 2018 ; Obuchowicz Rafałand Nurzynska et al, 2018 ; Tuan et al, 2018 ; Fariza et al, 2019 ; Kumar, Bhadauria & Singh, 2020 ).…”

“…Human identification (Fariza et al, 2019) For tooth segmentation, the Gaussian kernel-based conditional spatial fuzzy c-means (GK-csFCM) clustering algorithm is used.…”

Separation of Overlapping Object Segmentation Using Level Set With Automatic Initalization on Dental Panoramic Radiograph

“…Overlap object characteristics are having high grayscale intensity (brightest) from other parts. The part of normal teeth (without overlapping) that has the highest brightness of all parts of the tooth is enamel [19]. The overlapping object is formed by overlapping parts of the enamel.…”