Age estimation from human skeleton by machine learning

We are conducting research on techniques for estimating age, gender, reconstruction of 3D face, etc., based on 3D-CT images of unidentified skeletal remains. In this paper, we report the technology for estimating age by machine learning using 2D rendered images generated from 3D-CT images. By training the variables selected by LASSO with SVR, we achieved prediction accuracy of less than 9 years for femur, less than 7 years for lumbar, and less than 8 years for ischium.