Background: Contamination by mycotoxins is a major concern to the maize industry in north-east Italy where maize grain is often spoiled by Fusarium spp. In this work, fumonisins, deoxynivalenol and zearalenone were determined and an artificial neural network (ANN) model suitable for predicting mycotoxin contamination of maize at harvest time was developed.
Results: The occurrence of deoxynivalenol and zearalenone was very limited, while fumonisins concentration ranged from 163 and to 3663 µg kg(-1) in 2007, and from 333 to 11473 µg kg(-1) in 2008. Statistical data analysis of factors affecting fumonisins concentration revealed that irrigation, chemical treatment against the European corn borer and harvest date significantly affected the level of contamination (P < 0.05), although the relevance of the factors was different in 2007 and 2008. The neural network approach showed a significant correlation between ascertained values and predictions based on agronomic data.
Conclusion: This is the first time that an artificial neural network has been used to predict fumonisin accumulation in maize: the prediction has been shown to have the potential for the development of a new approach for the rapid cataloging of grain lots.
Copyright © 2012 Society of Chemical Industry.