Title:Optical detection of Aflatoxins B in grained almonds using fluorescence spectroscopy and machine learning algorithms

Abstract:Aflatoxins are fungal metabolites extensively produced by many different fungal species that may contaminate a wide range of agricultural food products. They have been studied extensively because of being associated with various chronic and acute diseases especially immunosuppression and cancer and their presence in food is strictly monitored and regulated worldwide. Aflatoxin detection and measurement relies mainly on chemical methods usually based on chromatography approaches, and recently developed immunochemical based assays that have advantages but also limitations, since these are expensive and destructive techniques. Nondestructive, optical approaches are recently being developed to assess presence of contamination in a cost and time effective way, maintaining acceptable accuracy and reproducibility. In this paper are presented the results obtained with a simple portable device for nondestructive detection of aflatoxins in almonds. The presented approach is based on the analysis of fluorescence spectra of slurried almonds under 375 nm wavelength excitation. Experiments were conducted with almonds contaminated in the range of 2.7-320.2 ng/g total aflatoxins B (AFB1 + AFB2) as determined by HPLC/FLD. After applying pre-processing steps, spectral analysis was carried out by a binary classification model based on SVM algorithm. A majority vote procedure was then performed on the classification results. In this way we could achieve, as best result, a classification accuracy of 94% (and false negative rate 5%) with a threshold set at 6.4 ng/g. These results illustrate the feasibility of such an approach in the great challenge of aflatoxin detection for food and feed safety.