ABSTRACT

Determination of bone age from radiography is important in pediatric development follow-up and forensic medicine. Bone age determination is usually made by radiological examination of the left hand using the Greulich and Pyle method or the TannerWhitehouse method. These methods produce results based on observational matches, which can cause differences in the determinations of radiologists. The aim of our study is to provide a supportive method that physicians can use in age determination, allowing them to make a more successful estimation. In this study, a method is proposed in which the calculated areas of carpal bones and the distal epiphyseal region of the radius are used together to assess bone age automatically. A native data set containing left hand graphics of boys and girls aged 1-7 was used. Carpal bones were decomposed using DICOM (Digital Imaging and Communications in Medicine) image window variables, edge and contour detectors. Areas were calculated by manually selecting the dissociated carpal bones. The areas and the distal epiphyseal region of the radius were given to the modelled artificial neural network and the network was trained with an accuracy of 87%. The success rate of the model on the test data was measured as 85%. As a result of the study, it was concluded that the proposed method is effective in determining bone age.