การตรวจจับฟัน และการระบุมุมของฟันล้มจากภาพเอกซเรย์พาโนรามิก

Abstract

This research project aims to develop a method for analyzing and identifying problems with tilted and impacted teeth to be useful for decision-making in dentists' treatment options. Therefore, this research preposes the process of teeth detection, identifying the inclination, and identifying the angle of tilted and impacted teeth. It consists of 5 steps follows: 1. Preparation of template 10 single and double roots teeth were cropped from the panoramic images, then the templates were averaged, and the templates were rotated by 5 degree each to obtain 72 templates. 2. Image acquisition, two methods of this step, method 1 cropping the image size 800x2000, 900x1900 and 1000x2100 and method 2 using image contrast by otsu's thresholding method together with finding the Mahalanobis distance and cut the area to leave only single root teeth or double root teeth. 3. Template matching, Comparison of similarity between templates and panoramic images using the correlation-based template matching method. 4. Teeth detection is divided into 3 sub-steps, segmentation of the detection coordinate data group, framing the area of the detected tooth and image segmentation 5. Identify the major axis and degree of inclination, according to Evan Ruth's theory. The results of 3 processes, each of which was tested with 50 panoramic images follows: The teeth detection process can be detected 160 single roots, 208 double roots teeth out of a total of 366 single roots and 364 double roots teeth with an accuracy of 57.14%. The measuring of teeth axis process was able to find 368 tooth axes with the angular difference (error) of 24.24% and the mean distance (d) of 25.92 pixels. And the measuring of the impacted teeth angle process can be found in 17 cases from 14 panoramic images with an error of 38.32%