Discrimination of Substandard and Falsified Formulations from Genuine Pharmaceuticals Using NIR Spectra and Machine Learning
Near-infrared (NIR) spectroscopy is a low-cost analytical tool for rapid characterization of pharmaceuticals. Usually, it's necessary to build up a "library" of NIR spectra from authentic products in order to identify real and fake formulations. In this paper, we developed a simple training method that uses binary mixtures of a pharmaceutical (acetaminophen) with two different cutting agents, and used several different data analysis models to extrapolate the training samples to real and fake samples of popular brands of acetaminophen tablets. While none of the models was able to correctly classify all the samples, a simple voting algorithm was able to compensate for the weaknesses of the models and classify 93% of the samples correctly.
Graduate student Olatunde Awotunde was the lead author, working with researchers at Precise Software Solutions and with undergraduates Nicholas Roseboom and Jin Cai at Notre Dame.