15-10-2016, 11:51 AM
An Edge Detection Scheme for Endodontic Working Length Measurement in Root Canal Treatment for Succedaneous Teeth
1459179240-wd1.docx (Size: 22.75 KB / Downloads: 3)
Abstract— The process of edge detection is useful in processing the dental x-ray images, specifically in determining the root morphology and the length to which the root has extended, for the purpose of providing root canal treatment. Generally, a manual analysis of an expert is made to determine these two features. However, the human analysis has two disadvantages: intra subject variability and inter subject variability in analyzing the features. Therefore, we propose an algorithm: Multi-scale and Multi Directional Analysis based Edge Detection with Statistical Thresholding (MMST), to determine these two features and were quite successful in their extraction. The results of the proposed algorithm are compared in terms of both qualitative and quantitative analysis with Canny method and are found to be doing appreciably well. The dental images used in this paper are taken from the database of the patients of Vishnu Dental College, Bhimavaram. Key-Words: - Endodontic working length; edge detection; Thresholding; Succedenous; Radiovisiography 1. INTRODUCTION A root canal is nothing but the space within the root of a tooth and consists of pulp chamber, the main canals and more complex anatomical branches that connect root canals to each other or to the surface of the root. Accessory canals are small branches found near the apex of the root. There may be one or more canals within each root. The root canal is filled with a highly vascular, loose connective tissue called the dental pulp. However, root canal is also common name for “Endodontic Therapy”, which is a treatment given for the pulp to remove infection and to give protection for the disinfected tooth from further microbial attacks. In the situations in which a decayed or cracked tooth is found or if there is a scope for further infection to that damaged tooth where, the doctors’ advice for a “Pulpectamy” which is the removal of pulp tissue. As a first step to cure the infection and save the tooth, the dentist drills into the pulp chamber and removes the infected pulp and then drills the nerve out of the root canal with long needle shaped drills. Then, the dentist fills each of the root canals and the chamber with an inert material and seals up the opening. This entire procedure is called the root canal or endodontic therapy. The standard filling material is “gutta- percha”, natural non elastic latex taken out of the percha or the Palaquium Gutta tree. The standard endodontic technique involves inserting a gutta – percha cone into the cleaned out root canal along with cement and sealer. The proposed method is useful at this stage of root canal treatment where the filling of the root canal with gutta –percha requires the shape of the passage and also the length to which the filling should be made. Generally, in endo-dontics, conventional film-based radiography is an important resource for diagnosis, trans-operative procedures, and treatment control. The digital radiography obtained through intra-buccal sensors rather than radiographic films represents technological progress that allows qualitative and quantitative analyses of all stages of endodontic therapy [11, 12]. Regarding radiographic estimation of endodontic working length, direct digital imaging provides measurement tools that facilitate the definition of the apical limit of root canal instrumentation. Moreover, there is substantial reduction in image processing time with the acquisition of digital radiographs. Hence, the clinical procedures are performed more quickly, with reduced radiation. However, when using conventional or digital techniques for working length determination, problems can occur when the endodontic treatment is performed in atresic and/or curved root canals. In these situations, frequently, the use of small instruments is required and they are barely perceptible in the radiographic image. Latest Trends in Circuits, Systems, Signal Processing and Automatic Control ISBN: 978-960-474-374-2 306 Furthermore, a manual analysis of the dental X-rays is made to know how to fill the root canal which may lead to errors because of the intra subject variability in the interpretation of results. So, the proposed Multi-scale Multi Directional Edge Detection Scheme with a Statistical Thresholding is used in this paper to determine the root morphology and also the length of the root especially for atresic and curved root canals. In the proposed method Laplacian Pyramidal decomposition is used to decompose the X-ray image of the tooth into number of levels, a Directional Filter Bank is used for the directional decomposition of the image. Moreover, the variability in the absolute magnitudes of the neighborhood pixels is considered to determine a threshold which is used to find the boundaries of the root canal. Thus, the skeleton of the root canal is determined using the proposed method and finally the distance between the CVG and the apex of the root is found out to determine the endodontic working length. The rest of the paper is organized as follows: In section 2, the details of the proposed methodology are presented, in section 3 the results obtained and the comparisons against the Canny algorithm are presented. Finally, the conclusions drawn from the experiments and results obtained in section 3 will be provided in section4. 2. PROPOSED METHOD The block diagram shown in Fig.1 provides the outline of the proposed method. The major sources of noise in a digital radiography system are quantum noise, TV Camera electronic noise, quantization noise from analog to digital converter, time jitter and video recording electronic noise. The X-ray quantum noise is determined by the X-ray dose delivered to the patient and by the beam energy and absorption properties of the image intensifier. TV camera electronic noise is made up of noise from the pre amplifier and beam shot noise in the camera tube. This coherent or electronic noise from the TV camera causes a speckle noise which gives a grainy appearance to the image [4]. The speckle noise is assumed to have multiplicative error model and must be reduced before the data can be utilized. Ideally, speckle noise in medical images must completely be removed. The removal of speckle noise in the medical images guarantees for an efficient and easier automatic image segmentation. Speckle noise can be reduced by spatial filtering. The spatial filters are categorized in to adaptive and nonadaptive filters. Non-adaptive filters do global processing of the image and leave out the local properties of the sensor. The non-adaptive filters are not suitable for stationery scene signals. Conversely, adaptive filters adjust themselves for changes in the local properties of the image [3]. These filters reduce speckles while preserving the edge information of the images. These filters modify the images based on the statistics extracted from the local neighborhood of each pixel. They vary the contrast for each pixel depending on the Digital Number (DN) values in the surrounding. As a result, adaptive filtering causes a better enhancement in the quality of an image. A number of speckle reduction filters are available in the literature. The mean filter is supposed to be the most classical one amongst them. The mean filter does not remove the speckle but it averages in to the data. The application of mean filter results in loss of details and resolution. The Median filter is also a simple filter that can remove pulse and speckle noise. Pulse functions which have less than fifty percent width than that of the kernel width are efficiently removed by using this filter. However, step and ramp functions will be retained. Therefore, it is essential to select a filter that is simple to realize but is efficient in avoiding the speckle noise completely. So, experiments are conducted to understand which filter will be most appropriate for the dental radio graphic images. The data set of images chosen for comparison is taken from the data base of Vishnu Dental Hospital at Bhimavaram in India. The data base consists of Radio Vision Graphs of the tooth taken for a number of patients for whom the root canal treatment is to be made. A dataset of more than 100 images is considered and the plot depicts the evaluation for a sample of test images. A plot depicting the performance evaluation of the filters is illustrated in Fig.2. As the performance of the Frost filter is the best among the three filters, this filter is chosen in the proposed algorithm to remove the speckle noise. X-Ray Image Frost Filter Cluster Adaptive Histogram Equalization DFB LP Statistical Thresholding Edge Map Root Length Fig.1: Block Diagram of Proposed Method Latest Trends in Circuits, Systems, Signal Processing and Automatic Control ISBN: 978-960-474-374-2 307 2.1. Frost Filter The Frost filter uses a kernel that is moved along the pixels in a local neighborhood of the image. Then, the pixel of interest is replaced with a weighted sum of the values within the n x n kernel. The weighting factors decrease with distance from the pixel of interest. The weighting factors increase for the central pixels as variance within the kernel increases. The filter assumes multiplicative and statistics. The filter is realized by using the expression H= ∑ - t nxn k e α α (1) Fig.2: Performance Evaluation of Filters 2.2. Contrast Enhancement Radiographic images usually have poor contrast. Therefore, it is necessary to enhance the contrast of the image before further processing. Histogram equalization is one of the most commonly used techniques for contrast enhancement. But, as a digital radiographic image have only a finite number of gray scales, an ideal equalization is not possible. The global histogram equalization amplifies the image noise and increases visual graininess or patchiness. Furthermore, this technique improves the contrast for brightness values closer to maxima in the image histogram and decrease contrast near minima. That means it improves the contrast in the areas of poor contrast of the image at the expense of areas of good contrast. A generalization of adaptive histogram equalization called Contrast Limited Adaptive Histogram Equalization, also known as CLAHE was developed to address the problem of noise amplification. The CLAHE algorithm partitions the images in to contextual regions and applies the histogram equalization to each one. The homogeneous areas can be characterized by a high peak in the histogram associated with contextual regions, as many pixels fall inside the same gray level range. High peaks in the histogram lead to large values in the final image because of integration. This problem is overcome by clipping the original histogram to a limit. The clip limit is defined as a multiple of the average histogram contents and is actually a contrast factor. If this limit is set to a higher value, the process becomes a standard Adaptive Histogram Equalization method while a low contrast factor avoids contrast enhancement. Therefore, generally a contrast factor normalized in a range of ‘0’ and ‘1’ is chosen for a good contrast enhancement. The dental x-rays acquired are generally low in contrast with spatially varying average brightness and noisy back ground. The example of one such image is shown in Fig.. 3 (a). Primarily, we are concerned here with enhancing the dental image to make it suitable for further processing. The dental image enhanced by using the histogram equalization method (HE), the adaptive histogram equalization method (AHE) and the CLAHE method is shown in Fig.3(b), Fig.3 © and Fig. 3(d) respectively. As the performance of the CLAHE is visibly better when compared to the other two algorithms, this technique is employed in our algorithm as a preprocessing step to enhance the contrast of the low contrast dental x-ray images. 2.3. Laplacian Pyramid Laplacian Pyramid (LP) is a non-orthogonal pass- band pyramid structure which was proposed by Burt and Adelson. The Laplacian pyramid decomposition is 0 5 10 15 20 25 30 35 40 45 50 Dent1Dent2Dent3Dent4Dent5Dent6Dent7Dent8 psnr(dB) Image Psnr(dB) with Mean filter Psnr(dB) with Median filter Psnr(dB) with Frost filter (a) (b) © (d) Fig.3: (a) Low Contrast Image,(b) Image enhanced by “HE”,© Image enhanced by “AHE”,(d) Image enhanced by “CLAHE” Latest Trends in Circuits, Systems,