人工神经网络在三维X线头影测量系统中的应用

Application of Artificial Neural Network in 3D X-ray Cephalometrics System

给出一种三维X线头影测量中图形图像重建的方法;由正侧位两张X线头影片上72个标志点,可以重建出颅颌骨三维透视图;利用人工神经网络LMBP算法,以病人颅颌骨72个标志点的三维坐标集合在某一平面的二维投影为理想输出,以人体颅颌骨标本72个同名标志点在相同平面的二维坐标集合为训练样本,在训练结束后即可建立两者之间的近似线性关系;由于网络具有泛化能力,因此可用该网络对颅颌骨标本上每一像素进行变换,结果表明变换后72个标志点坐标和病人很好吻合,变换图像为病人近似的颅颌骨图像;即满足口腔正畸诊断要求又获得较好图像视觉效果。

A methord of rebuilding graphics and images iu 3--D X-ray cephalometrics is presented. By using 72 mark points in two X-- ray images, 3--D graphics can he reconstruct. By using artificial neural network, the patient＇s coordinates of 72 mark points in a ptane are used as designed output of the network and the coordinates of 72 mark points of skull sample in the same ptane are used as learning input of the network, and a learning procedure is carried out. After learning, a linear retationship between input and output of the network is set up. Because of the generatization of the network, pies in sample image can be transformed. Results show that the 72 mark points in transformed image are very close to the patients and the transformed inaage is atike the patient＇s. It is satisfaction for diagnosis and better visual effect.