Estimation of bloodstains time since deposition based on ATR-FTIR spectroscopy in forensic laboratorie.

Abstract

The age of bloodstains at a crime scene provides key information for criminal investigation and interpretation, with important implications in forensic medicine. In this study, silica gel was used as a carrier for bloodstains with different ages to simulate a porous wall surface at an indoor crime scene. A method was developed for bloodstain dating based on attenuated total reflection infrared spectroscopy (ATR-FTIR) and neural networks. Venous blood samples were collected from nine healthy volunteers, and ATR spectra were recorded at five points for each sample during a period of 7 days. The neural networks TRAINSCG, TRAINLM, and TRANGDM were constructed. The training dataset was the ATR spectra (4,000-600 cm^- 1) of samples collected from seven participants (YP1-YP7) and recorded at five points over 7 days (a total of 245 spectra). The prediction dataset was 70 spectra from two participants (YP8 and YP9). The prediction accuracy of the neural networks was compared with different numbers of hidden layers and neurons. The key absorption peaks at 1800-1300 cm^- 1 were used for neural network training and bloodstain dating. The neural network trained using the Levenberg-Marquardt algorithm based on ATR spectra (1800-1300 cm^- 1) was used for predicting the age of bloodstains on silica gel. The coefficient of determination (R²) for predicted and actual bloodstain ages was up to 0.9215 after removing outliers. ATR used in combination with neural networks provides a non-destructive and rapid method for bloodstain dating. Neural networks constructed using different algorithms showed varying performance in bloodstain dating with ATR. Prediction accuracy was improved with the Levenberg-Marquardt algorithm and key peaks.