AI-Assisted Quantification of Hydroxyapatite in Treated Archaeological Samples

This study highlights the value of infrared spectroscopy for analysing archaeological sediments, focusing on 11 samples from Neolithic ash-rich layers at the Riparo Gaban site near Trento. After chemical treatment to remove calcite and enhance hydroxyapatite signals, FTIR spectra were acquired using KBr pellets. A spectral database of standard minerals and their mixtures was created to support quantitative analysis. One key challenge is the overlap of absorption peaks among target minerals. To address this issue, the study took advantage of a neural-network based post-processing. Due to the large number of spectra typically needed to train neural networks, an algorithm was developed to generate synthetic spectra, expanding the available dataset. This innovation enabled accurate quantitative analysis of hydroxyapatite, demonstrating the potential of combining spectroscopy with AI to overcome limitations in archaeological material analysis.