Examination and Analysis on Oxygen and Hydrogen Stable Isotope Ratios of Resident Hair br

Objective The stable isotopic composition of human tissue is related to the individual's diet,environment and metabolic status.Once grown,human hair is no longer exchanged with the body and ischemically stable and easy to collect,making it a good subject for studying the stable isotope composition ofhuman tissue.The oxygen and hydrogen elements that make up the human body are mainly derived from thewater and food,of which the oxygen and hydrogen stable isotope composition is recorded in hair keratin.Differences in the stable isotope composition of oxygen and hydrogen in the human hair of different regions canbe used to infer information about people's diets,living areas and activity trajectories,which is of greatimportance in forensic science and other research fields.MethodsIn this study,oxygen and hydrogen stableisotope ratios of hair samples from permanent residents in different regions of China were detected and analyzedusing an elemental analyzer-isotope ratio mass spectrometer(EA-IRMS).Results The results showed that therewere significant differences in delta 18O and delta 2H between some cities,and a significant positive correlation betweenoverall delta 18O and delta 2H.The overall discrimination accuracy of the stable isotope data was63.9% for the cross-validation,and increased significantly when combined with the carbon and nitrogen stable isotope data,reachingan overall discrimination accuracy of76.0% for the cross-validation.As the type of stable isotopes used in thediscriminant analysis increased,the model's discriminant ability was significantly enhanced.Conclusion The overall discriminative accuracy of the multi-layer perceptron neural network model built by using the stableisotope data of the4elements was82.8%,and the overall discriminative accuracy of the radial basis functionneural network model was78.8%.The multilayer perceptron neural network model had the highest discriminativeaccuracy among the three traceability mathematical models