Measurement of Carboxyhemoglobin in Forensic Blood Samples Using UV-Visible Spectrometry and Improved Principal Component Regression

Abstract

The forensic determination of carboxyhemoglobin (COHb) in blood was performed by using an improved principal component regression (PCR) technique applied to UV-visible spectra. Calibration data were decomposed into principal components, and the principal components useful for prediction were selected by their correlation with calibration spectra. Cross-validation of prediction results was done by leverage-corrected residuals. Confidence and prediction intervals derived from classical regression theory were found to be reasonable in size. The results compared favorably to a comparison study conducted by using a CO Oximeter method. In analysis of forensic case study samples, the improved PCR method allowed detection of abnormal samples and successfully predicted percentages of COHb and methemoglobin (MetHb), and provided error estimates for those predictions.