FT-NIR spectroscopy for rapid analysis of deoxynivalenol in wheat and wheat bran

Deoxynivalenol (DON) is a mycotoxin mainly produced by several Fusarium species occurring in cereals and derived products. Rapid, robust and inexpensive methods using Fourier-Transform-Near Infrared (FT-NIR) spectroscopy have been recently developed at ISPA-CNR to predict DON levels in durum wheat. Linear Discriminant Analysis (LDA) models were developed based on different cut-off limits (i.e. 1000, 1200 and 1400 ?g/kg DON) that were set at levels lower than the EC maximum limit for DON in unprocessed durum wheat (i.e. 1750 ?g/kg). The overall classification rates of models were 89-91% with false compliant values of 3-7%. Model using a cut-off of 1400 ?g/kg fulfilled the requirement of the European official guidelines for screening methods. Partial Least-Squares (PLS) regression analysis was also used to determine DON content in wheat samples in the range of <50-6000 ?g/kg (as determined by a reference HPLC method). The model displayed good regression quality with a root mean square error (RMSE) of prediction of 868 ?g/kg. The feasibility of using FT-NIR spectroscopy was also investigated to rapidly predict DON in durum wheat bran at levels up to 1600 ?g/kg by both LDA and PLS analysis. The LDA model used a cut-off value of 400 ?g/kg that was lower than the EC maximum limit for DON in bran (i.e. 750 ?g/kg) and displayed a classification rate of 80% with 5% of false compliant samples. Good performance results were also obtained by applying the PLS statistical model, confirming a good fit between HPLC and FT-NIR data in the tested range with an RMSE of cross-validation of 191 ?g/kg. These findings confirmed the suitability of FT-NIR to rapidly screen a large number of wheat samples for DON contamination and to verify the compliance with EU regulation.