Forensic age estimation with Bayesian convolutional neural networks based on panoramic dental X-ray imaging


Forensic age estimation is the medical assessment of age in individuals for legal purposes. In current practice, medical experts use scoring-based methods to assess the most probable age and provide a minimum age estimate, which requires expensive training and is known to have poor inter-rater reliability. As an initial step towards a more objective, quantitative age estimation system, we use Bayesian convolutional neural networks to perform age and uncertainty estimation using a large data set of 12000 panoramic radiographs of the upper and lower jaws, orthopantomograms (OPTs), one of the most accurate indicators of age in subadults. Our system achieves a concordance correlation coefficient ccc=0.91 on the validation set. Importantly, our method provides quantitative estimation of prediction uncertainty, which is imperative within a legal context.

Medical Imaging with Deep Learning (MIDL)