Abstract
Bone age (BA) estimation is one of the important applications in the area of pediatrics, especially in the diagnosis of endocrinological problems and growth disorders. Because manual BA estimation (BAE) is tedious and time consuming and heavily dependent upon doctor's experiences, we attempt to construct a computerized BAE system to help reduce the doctor's burden. It is known that sex, race, and nutrition will affect the results of BAE. So far to now, most of BAE standards are American standards. There is no BAE standard for Taiwanese people. With the aid of the proposed computerized BAE system, a standard for Taiwanese people can be easily built which will greatly help doctors to estimate BA in Taiwan. Our BAE system is based on the carpal-bone information. In 1993, Ewa Pietka, Lotfi Kaabi, M.L. Kuo and H.K. Huang had developed a feature extraction method for carpal-bone[3]. They use two-step local thresholding method to extract the area of a carpal bone. However, the method did not work well in our database. Carpal bones can not be extracted simply by the local thresholding method. Therefore, we develop a new method for carpal-bone feature extraction, which we call the two-stage edge detection algorithm, and a new method for carpal-bone ROI division. After the image is manually equalized, our BAE system can extract the features of the carpal-bone and estimate the bone age automatically. Three different classifiers: the minimum distance classifier, Bayes classifier, and a neural network classifier are tested in our experiments. The results of our feature extraction are quite satisfactory. More than 90% of the classification are acceptable in our experiments.