Detection of residues of chlorine-containing pesticides on tomato fruits by machine learning evaluation of laser-induced breakdown spectroscopy data.

In this pilot analytical study, we propose the machine learning evaluation of laser-induced breakdown spectroscopy (LIBS) data as a new, fast screening methodology for the detection of pesticides on tomato fruits. Five pesticides (acetamiprid, chlorpyrifos, λ-cyhalothrin, tebuconazole, tefluthrin) at three concentration levels down to 0.01 m/v% corresponding to their real application (spraying) concentrations were tested on fully ripe fruits of tomato (Solanum lycopersicum cv. Money maker). Direct (on the fruit peel) and indirect (sampling-based) analytical approaches were both tested. Linear discriminant analysis and classification tree machine learners, coupled with data normalization were applied to the LIBS data collected. It was found that the proposed method can provide a fairly robust classification with up to about 90% balanced accuracy. The results indicate that LIBS-based qualitative discrimination analysis is a promising candidate for further development towards the detection of hazardous pesticide residues.