Rapid Detection of Aflatoxins in Ground Maize Using Near Infrared Spectroscopy

Abstract

Aflatoxins are carcinogenic mycotoxins that may contaminate many crops and more especially maize. To protect consumers from these contaminants, many countries set up low regulatory thresholds of few µg/kg. The control of food requires time-consuming analysis for which sampling is a key step. It would therefore of key sanitary and economic relevance to develop rapid, sensitive and accurate methods that could even be applied on line at harvest, to identify batches to be excluded as soon as possible. In this study, we analyzed more than 500 maize samples taken at harvest during 3 years for their aflatoxin contamination using HPLC-MS. Among them, only 7% were contaminated but sometimes at levels largely exceeding European regulations. We demonstrate that Near InfraRed Spectroscopy (NIRS) could be of great help to classify cereal samples according to their level of aflatoxin contamination (below or higher than E.U. regulation). To build the model, all AF contaminated samples as well as an equivalent number of AF free samples were used. NIRS performance was not sufficient to quantify the toxins with adequate precision. However, its ability to discriminate naturally contaminated maize samples according to their level of contamination with aflatoxins in relation to European regulations using a quadratic PCA-DA model was excellent. Accuracy of the model was 97.4% for aflatoxin B1 and 100% for total aflatoxins.

Keywords: aflatoxins; maize; near infrared spectroscopy; rapid detection.

Figure 1

Raw spectra of maize samples across Wavelengths 908–1650 nm.

Figure 2

Performance of the ANN model for predicting AFB1 content in ground maize samples. The scatter plot shows the predicted versus reference values for the training set (grey circles) and the validation set (green circles).

Figure 3

Performance of the ANN model for predicting Total AF content in maize samples. The scatter plot shows the predicted versus reference values for the training set (grey circles) and the validation set (green circles).