Analysis of polynomial behavior of the C3 cervical concavity to bone age estimation using artificial neural networks

The bone age estimation is obtained by radiographic analysis of any bone in human body, which is considered as important information for medical and orthodontic diagnosis. The use of lateral cephalograms for bone age reports reduces the patient's exposure to ionizing radiation because this radiography is usually requested for orthodontic treatments. This paper presents a novel methodology for mapping the behavior of C3 vertebra concavity (presented in lateral cephalograms) related to bone age. A 4
^th
degree polynomial is used to describe the concavity. Its coefficients are the input of an artificial neural network. Thus, the best topology is chosen based on the correlation coefficient between the output of the network and bone age. After obtaining and analyzing several metrics, the results related to neural network application are expressive, achieving near 0.9 of correlation and the results related to bone age estimation are promising, with mean absolute error around 0.8 years. The results allow the conclusion that just one feature (C3 concavity) has high correlation with bone age.