Direct Analysis of Blood for Diagnostic Metals for Malaria by Peak-Free Laser-Induced Breakdown Spectroscopy (LIBS) with Artificial Neural Networks (ANN) and Partial Least Squares (PLS)

Abstract

Although laser induced breakdown spectroscopy (LIBS) spectra are rich in latent chemical information, biometal spectral signatures from body fluids are subtle (peak-free) and hardly discernible against the background. Trace analysis requires considerable development in the measurement strategy. The preliminary performance of peak-free LIBS was explored with chemometrics for the direct rapid determination of malarial diagnostic trace biometals copper (Cu), zinc (Zn), iron (Fe), and magnesium (Mg) in blood. Artificial neural networks (ANN) and partial least squares (PLS) calibration strategies exploiting spectral feature selection were developed and their analytical performance evaluated. The model regression coefficients were 0.990, 1.000, 1.000, and 0.973 for Zn, Cu, Fe, and Mg respectively for ANN and 0.960, 0.985, 0.986, and 0.993 for PLS. The root mean square errors were respectively 0.012, 0.026, 0.036, and 0.870 for Zn, Cu, Fe, and Mg for ANN and 0.056, 0.032, 0.030, and 0.243 for Zn, Cu, Fe, and Mg for PLS. The accuracy by which the biometals may be determined in blood was established by analyzing animal serum and was 73%, 68%, and 99% for Zn, Cu, and Fe, respectively, for ANN; and 70%, 68% and 95% for PLS. These results show proof-of-concept for single-shot chemometric peak-free LIBS for direct determination of biometals in blood with potential for malaria diagnostics based on the concentrations together with analyte alterations and multivariate correlations.