NCJ Number
252634
Date Published
August 2016
Length
8 pages
Annotation
In order to develop a universal, confirmatory, and nondestructive approach that can be used to differentiate and identify body fluids, this study combined the specificity of Raman spectroscopy with the analytical power of statistical modeling.
Abstract
The ability to identify body fluid traces at crime scenes and preserve any DNA present is critically important in forensic science. Identification can be difficult because many of the current techniques are specific to one body fluid, and typical biochemical methods are destructive, preventing any further analysis. In the current study, Raman spectra were collected from 75 body-fluid samples, including peripheral blood, saliva, semen, sweat, and vaginal fluid. After preprocessing, samples were split into calibration and validation datasets. Several chemometric analysis techniques were trained and tested to find the best model. Combining classification modeling with variable selection resulted in a single, robust, technique. This enhanced model accurately predicted the identity of 99.9 percent of the spectra from the calibration dataset after cross-validation. More importantly, it correctly predicted the identity of 100 percent of the spectra in the external validation dataset. All five body fluids were successfully discriminated by coupling Raman spectroscopy and chemometrics. This technique is both reliable and nondestructive, offering substantial advantages over the current techniques used to identify body fluids. (Publisher abstract modified)