Application of laser-induced breakdown spectroscopy and neural networks on archaeological human bones for the discrimination of distinct individuals.

The use of elemental analysis based on Laser-Induced Breakdown Spectroscopy (LIBS) combined with Neural Networks (NN) is being evaluated as a method for assigning archaeological bone remains to individuals. The bone samples examined originate from excavations of burials at the Cross Street Unitarian Chapel, Manchester (United Kingdom) that date from the 17th to the 19th century. In this study, we critically assess the influence of soil contaminants, by separating the bone elemental fingerprint into two groups of different components prior to the NN analysis. The first group includes elements related to the bone matrix (Ca and P) as well as elements that are regularly incorporated in the living bone tissues (Mg, Na, Sr, and Ba). The second group includes metals with a low probability of accumulation in living bone tissues whose presence is more likely to be related to diagenesis and the chemical composition of the burial soil (Al, Fe, Mn). The NN analysis of the spectral data, based on the use of an open access software, provided accurate results, indicating that it can be a promising tool for enhancing LIBS applications in osteoarchaeology. The influence of bone diagenesis and soil contaminants is significant. False classifications occurred exclusively in the NN analyses that relied partially on elemental peaks from the second group of elements. Overall, the present study indicates that discrimination between individuals through LIBS and NN analysis of bone material in an archaeological setting is possible, but a targeted approach based on selected elements is required and the influence of bone diagenesis will have to be assessed on a case-by-case basis. The proposed LIBS-NN method has potential as a tool capable for distinguishing distinct individuals in disarticulated or commingled human skeletal assemblages particularly if combined with standard osteometric methods.