科研之友
微信
新浪微博
Facebook
分享链接
摘要
Forty lateral cephalometric radiographs, taken from the data files of subjects were used in this study. A purpose-made software allowed recording of the cephalometric points and then, with the help of algorithms based on cellular neural networks, to transfer the previously processed radiographs into an embossed image. Five observers recorded 22 landmarks on the displayed images from the two image modalities, i.e. monitor-displayed radiograph (mode A) and monitor-displayed embossed radiograph (mode B). The positions of the landmarks were recorded and saved in the format of x and y co-ordinates and as Euclidean distance. The mean errors and standard deviation of landmarks location according to the two modalities were compared with the 'best estimate' for each landmark and the values were calculated for each of the 22 landmarks. One-way analysis of variance was then used to evaluate any statistically significant differences.
Euclidean distance mean errors were higher for the embossed images (except for Po) than for the unfiltered radiographs. These differences were all statistically significant (P < 0.05) except for Or, Po, PM, Co, APOcc, and PPOcc. On the x and y co-ordinates, the accuracy of the cephalometric landmark detection improved on the embossed radiograph but only for a few points (Or on x axis and Po, PM, Co, and APOcc on y axis), as these were not statistically significant. The use of radiographic enhancement techniques, such as embossing, does not improve the level of accuracy for cephalometric points detection. Unless more precise algorithms are designed, this feature should not be used for clinical or research purposes.
- 出版日期2010-6
