Optimized Spectral Processing of Infrared Diffuse Reflection Spectra for the Detection of Blood on Fabric

The detection of bodily fluids such as blood on interfering backgrounds is important to the forensic community. Luminol, which participates in a chemiluminescent reaction with the heme groups of blood, is one of the most commonly used presumptive tests. Luminol has a few drawbacks, including the requirement of use in a dark environment and potential to degrade the amount of recoverable DNA in blood stains. A potential complementary method is infrared (IR) spectroscopy. These two methods are compared in this work. Infrared diffuse reflection (DR) spectroscopy works well to measure the IR spectrum of samples that are highly absorbing or scattering, such as fabric. In the IR DR spectrum of blood on fabric, the contribution of analyte signal to the total signal is weak, and many of the characteristic amide absorbance bands of blood proteins overlap with the spectral features of fabrics. Derivative transformations are commonly applied to resolve overlapping spectral peaks. These transformations are typically implemented as Savitzky-Golay (SG) derivatives. The performance of optimized higher-order gap derivatives (GDs) and SG derivatives are compared here as preprocessing methods for partial least-squares regression (PLSR), a multivariate calibration technique. Optimized GD processing is found to behave similarly to a matched filter to highlight spectral features of the analyte relative to an interfering background. Derivatives can result in complicated spectra and regression vectors (RV) from the PLSR calibration. To enable better interpretation of the RVs, it is useful to examine the RVs in the original spectral space, which is more familiar to spectroscopists. To that end, we offer a method of calculating higher-order GDs that allows the resulting GD and RVs to be exactly integrated to spectral space. Infrared detection limits (DLs) for blood on four fabric types (acrylic, cotton, nylon, and polyester) were estimated using optimized GD processing and PLSR. The best IR DLs for blood on fabric were found in the mid-IR spectral region. The DLs for acrylic, cotton, and polyester fabrics were blood diluted by factors of 2300, 610, and 900, respectively. Due to the similarity between the IR spectra of blood solids and nylon, no satisfactory IR DLs were determined for the calibration of blood on nylon. These DLs are on the order of the most commonly reported DLs (1000x dilute) for blood on fabric using the standard luminol method. An approach to further improve the DL by accounting for known sources of extraneous variance in the spectra is briefly presented