Location of craniofacial landmarks on X-ray images by employing fuzzy neural network

Analysis of cephalometric X-rays and measurements of parameters based on detected features play an important role in monitoring the treatment. The analysis is based on location of landmarks that form the basis of linear and angular measurements. Landmarks are difficult to distinguish in the image because their location depends on external forces such as growth, rotation and shifting in the skull X-ray. In this paper we present a new method for estimating locations of the landmarks by employing a fuzzy neural network. This method has a very high non-linear mapping ability, a fast learning time and guarantee of global minima. In the proposed method fuzzy sets are formed from the training data. Using fuzzy linguistics if-then rules, membership degrees are assigned to the extracted features and fed to the network for training utilizing gradient descent scheme. After training the network is tested on estimating the locations of the landmarks on target images (not used for training). The results moreover, are promising.