The Integrated Cephalometric Analysis System

• Main Authors: Yen-ting Chen, 陳彥廷
• Other Authors: Kuo-sheng Cheng
• Format: Others
• Language: en_US
• Published: 2000
• Online Access: http://ndltd.ncl.edu.tw/handle/87343515076587616340

博士 === 國立成功大學 === 醫學工程研究所 === 88 === This purpose of this study is to develop the integrated computerized system for cephalometry. This proposed system contains two major parts as follows. One subsystem is developed for computer-assisted analysis, and the other for automatic computerization. The former one includes an intelligent view box and the identification software for manual landmarking so as to be directly applied in actual clinical analysis. The latter one includes two parts: One is the feature subimage extracting for automatic landmarking, and the other is computerized cephalometic superimposition for tracings. The intelligent view box combining a touch panel and view box extends the application of touch panel for subtle pointing. The design and performances of this system are investigated. A coordinate look-up table can actually provide a reference for the adjustment of the position drift. The results of performance tests reveal that the proposed intelligent view box has high resolutions, good stability, and good reproducibility. The landmark identification software for manual landmarking is integrated into the intelligent view box to facilitate the work for manual landmarking. A genetic algorithm is implemented to “guess” the manual located landmarks by evaluating the difference of the angular correlation matrices between the located points and the template landmarks. This angular correlation matrix represents the angular relationships for the spatial locations of these landmarks. The results of statistical analysis show that the algorithm well performs well in all landmarks identification. The method of multilayer perceptron with genetic algorithm searching is proposed for feature subimage extraction. This method is used to consolidate the performance for automatic landmarking with cephalogram. From the experimental results of the comparison with the cross correlation, this multilayer perceptron with genetic algorithm shows a better performance for feature subimage extraction. In the computerized cephalometric superimposition for the tracings, fuzzy theory is introduced to represent the spatial weighting functions for features. This algorithm finds the best match of the two tracings by finding highest similarity between the fuzzy-weighted feature curves. The appraisals of two senior orthodontists indicate that the performance of this system is good for clinical application. In words, the completely automatic cephalometric analysis system is the long-term goal of this study. Up to now, we have designed the intelligent view box system, developed the clinical assistant system for cephalometric analysis with the automatic identification of manual located landmarks, and investigated the theories for the feature subimage extraction and the computerized superimposition with tracings. The cephalomatric analysis system by using tracings has been used in the Department of Dentistry of National Chang Kung University with good performance in clinical application.