Abstract:
Radiological age estimation of living subjects from MR images has recently become very popular. Besides skeletal ossification this can be done using the mineralization status of wisdom teeth. To support potential automatic age estimation, an important preliminary step is a reliable and automatic localization of the wisdom teeth. Therefore, we propose a random regression forest framework to localize third molars, which is capable to predict landmarks up to an error of 3.55 ± 2.62 mm in mean and standard deviation in a challenging 3D MRI dataset.