Exploring the Use of Raman Spectroscopy and Covariate-Adjusted Multivariate Analysis for the Detection of Irradiated Blood.

Biological dosimetry is a key technique for retrospective radiation dosimetry that provides individual estimates of absorbed dose of ionizing radiation, applicable for use in a large scale radiological/nuclear event. Current techniques for biodosimetry are labor intensive and time consuming and not high through-put. In this proof-of-concept study, we developed a new approach for detecting irradiated blood based on Raman spectroscopy of blood combined with multivariate analysis. Peripheral blood samples from 8 healthy male and female, anonymous donors, were exposed to either 5 Gy X ray radiation or unirradiated (0 Gy). At 3 h postirradiation, the blood was immediately frozen at -80°C. Raman spectra were measured from thawed blood using a portable spectrometer system. Data were preprocessed and analyzed using principal component analysis (PCA) to observe trends in the data, and by using partial least squares-discriminant analysis (PLS-DA) to build a model to discriminate between Raman spectra of control (0 Gy) and irradiated (5 Gy) blood. We found strong evidence of inter-donor variability in the form of donor-wise clustering of PCA scores corresponding to the control Raman spectra, in addition to the poor separation of PLS-DA scores associated with Raman intensities of 0 Gy vs. 5 Gy spectra. However, after adjustment for donor covariates using a linear mixed-effects model, we obtained a better separation between control and irradiated blood using PLS-DA. Evaluation of the coefficients of the PLS-DA loading vectors indicated radiation-induced changes in proteins, lipids and hemoglobin to be major contributors for this discrimination. This pilot study demonstrates the potential of application of Raman spectroscopy to support biodosimetry of blood and blood components.