Computerized Automatic Recognition System Of Cephalometry

Cephalometry has been a basic method of clinical diagnosis, treatment design and scientific research in orthodontics and orthopedics since it came out in 1931. There are mainly three approaches to perform cephalometry: dentifying landmarks and measuring, analyzing manually; identifying landmarks manually but performing computer assisted measurement and analysis; identifying landmarks and measuring, analyzing completely automatically. The first two methods are time-consuming and take much effort. Additionally, it will bring about subjective errors during the process of locating landmarks manually. Therefore, it has been a research focus on developing computerized automatic measurement and analysis system of cephalometry in recent years. It's expected to release clinicians from heavy work and eliminate the subjective errors during landmarks locating process. The development of modern computer imaging process makes it possible to solve this problem, and the development of digital X ray radiography technique puts forward an urgent demand for computerized automatic measurementand analysis system of cephalometry. At the present time, computerized automatic measurement and analysis system of cephalometry is still at the stage of research in laboratory, with no successful software system coming out in the world.The objective of this study is to establish computerized automatic recognition system of cephalometry, to identify cephalometric landmarks automatically, to evaluate the accuracy of automaticly identifying landmarks by comparing it with identifying landmarks manually, so as to lay a foundation for the establishment of computerized automatic measurement and analysis system of cephalomtry.1. Extraction of soft-tissue contour and hard-tissue contour in cephalometry:Materials and methods: 10 cephalometric radiographs of patients in the Department of Orthodontics, Stomatology College, the Fourth Military Medical University were selected. The digital images of these cephalometric radiographs were produced by flat scanner and stored in the computer. After image pretreatment, a dynamic threshold that is based on different districts was used to extract the external contour of craniofacial soft tissues and hard tissues and the internal contour of craniofacial hard tissues.Results: Craniofacial external soft tissue contour, hard tissue contour and craniofcial internal contour were extracted.2. Automatic recognition of cephalometric landmarks: Materials and methods: Craniofacial external soft tissue contour,hard tissue contour and craniofcial internal contour generated by computer were used. Edge detection, curve thinning, curve tracking, sampling and curve fitting are a series of methods acted on thecraniofacial contours to get several curves respectively. Then, cephalometric landmarks are redefined according to their spatial positions and geometric characters. The new definitions are in accordance with the original definitions in anatomy. The exclusive coordinates values of cephalometric landmarks are generated by computer automatically.Results: 19 cephalometric landmarks were identified and their coordinates values were generated automatically. These landmarks are as follows: Gs, Ns, UL, UL, LL, LL, Si, Pos, Gns, Mes, G, A, UI, B, Pm, Po, Gn, Me, D.3. Establishment of computerized automatic recognition system of cephalometry:Materials and methods: A microcomputer was used designing programme. The CPU is celeron 1.7GHz, RAM is 256M, operation system is windows 2000. Borland C++ Builder 6.0 was used as programming tool.Results: Computerized automatic recognition system of cephalometry was set up. Its function includes opening image, saving image, showing the histogram of image and identifying 19 cephalometric landmarks automatically.4. Comparison between identifying landmarks automatically and manually:Materials and methods: 10 cephalometric radiographs of patients and their digital images were used. Cephalometric landmarks were identified manually on the original 10 cephalometric radiographs...