Fluorometric determination of mixtures of quinolones by means of partial least squares and neural networks.

Multivariate calibration methods (partial least squares calibration, back propagation multilayer perceptrons networks, radial basis functions and generalized regression neural networks) were applied to the simultaneous fluorometric quantification of levofloxacin, garenoxacin and grepafloxacin, without previous separation steps. A data matrix was obtained by registering the emission spectra of mixtures of the three quinolones in urine (with concentrations ranging over 0.00-0.40 microg mL-1 for each quinolone) with a 283 nm excitation at pH 4.0. The generalized regression neural network model proved to be the most adequate model for simultaneous quantification of the three quinolones in urine samples.