Research
Filter
Filter All Projects

Project Details

Bravo Morante Guillermo | Writing-Up Fellow
2020-01-15 - 2020-07-14 | Research area: EvoDevo
Predicting Age at Death from the Shape of the Human Pubic Symphysis by Bandpass Filtering of Bending Energy

In forensics, establishing the biological profile is the first and most crucial step toward successful identification. One of the most difficult parameters to estimate with accuracy is age at death. For this purpose, we exploit the state of development of the pubic symphysis, due to its robustness, to be extended by parallel analyses of other remains when available. Our proposed method is based on contrasts between the process of osteogenesis in the first 30 years of life and the later degenerative processes of aging, during which the shape of the symphyseal surface changes notably. A purely visual classification of this progression is unreliable. The use of a surface scanner to model the pubic symphysis makes it possible to carry out this task more objectively.
For this, we used an Artec Spider (Artec Group) to scan a sample of 400 male individuals aged between 14 and 82 years from the collection of the Laboratory of Anthropology at the University of Granada, all of them Mediterranean and from the last century; each pubis was landmarked following a 102-point template of two fixed landmarks on the top and the bottom of the symphysis along with 100 surface semilandmarks. From the sample we selected the 381 specimens within Procrustes distance 0.05 of the side-specific average, rotated to the standard geometric morphometrics basis of partial warp scores, and then, separately by side for ages under 50, correlated age with summed squared partial warps (PW) amplitudes over a wide range of plausible bandpass filters omitting the uniform term. Peak correlations were -0.524 both for the PW1-PW6 band on the right side and for the PW1-PW7 band on the left side, and the geometry of singular warps PW by PW was virtually identical between the two analyses.
We calculated a standard error of prediction of 8.3 although, for the complexity of the regression, we consider it an inappropriate way of reviewing the accuracy of the method. Instead of than that we predicted the logarithm of age instead of age per se, obtaining a higher correlation, -0.568, with a standard error of -23% to 29% around the estimated age. Thus, there is a strong age signal in symphyseal surface shape that is not a reduced roughness per se but rather a flattening at large scales.
The method is implemented in a version of the easy-to-use R statistical software package, making it easy to combine with other modern geometric morphometrics methods and ready to apply in other samples.