Estudo de manchas de sangue: uma abordagem forense empregando espectroscopia Raman, imagens digitais e ferramentas quimiométricas

Abstract

The study of evidence and traces found at the scene of a crime is one of the essential steps in a criminal investigation. Thus, the purpose of this work was to perform a bloodstain study using Raman spectroscopy and digital images, along with chemometric tools. Bloodstains were analyzed over time and the different environmental exposure conditions. In the study, samples were collected from 10 voluntary donors, men andwomen aged between 20 and 50 years. The chemometric tools were used for exploratory analysis, in which case the principal component analysis (PCA) was employed. Multivariate regression models were constructed using partial least squares regression (PLS) to estimate the exposure time of the stains. In the methodology using Ramanspectroscopy, the time interval studied was 269 hours and the regression model was constructed employing 7 latent variables with a variance explained in the X matrix of 84.5% and in the vector y of 98.49%, which provides exposure time information in hours. The mean error of prediction was 19h and the correlation coefficient of 0.91. The mostimportant Raman spectrum band for the regression model were assigned to deoxyhemoglobin (met-Hb) at 377 cm-1, since this band are related to chemical changes occurring in the blood over the exposure time. The digital images were obtained with a smartphone Galaxy S8 and 9 color systems were used for the construction of the regression models, as well as different substrates and environmental conditions. For thebest regression model, we selected 6 latent variables with 98.26% of the variance explained in matrix X and 89.19% in vector y, which describes the time of logarithmic minute exposure. The mean prediction error of the model was 0.43 and the time interval monitored was 3967 hours. The results obtained in this work show that environmental conditions influence the speed of the chemical reactions that occur in the bloodstain and cause high prediction errors, therefore this information should be considered when the regression models are constructed. Finally, the proposed methodologies present potential to accompany the chemical and physical changes that occur in bloodstains as a functionof the exposure time and can be applied as screening tests for forensic purposes.