Detection of fungal infection in almond kernels using near-infrared reflectance spectroscopy

Aspergillus flavus and Aspergillus parasiticus are ubiquitous in natural environments and have the potential to produce aflatoxins that can cause both chronic and acute diseases in humans and animals. The near-infrared (800-2500 nm) reflectance spectra of almonds from the 2012 and 2013 harvest seasons inoculated with one selected isolate of either A. flavus or A. parasiticus were analysed to determine if the spectra could be used to distinguish infected almonds from uninfected control almonds. A canonical classifier was developed that could discriminate infected almonds from the uninfected control kernels with a total cross-validation error rate of 0.26% and zero false negative errors. Additionally, a follow-up canonical classifier was developed to evaluate the potential to discriminate between infections by the two Aspergillus species, once infection was detected. Spectral analysis indicated that the NIR wavebands associated with lipid content were the most indicative factors in both identifying the infected almonds as well as discriminating between almonds infected by the A. flavus or the A. parasiticus isolate. Further comparison of the canonical discriminant analysis results from the 2nd derivative spectra and the original spectra also suggested that the chemical differences caused by the fungal metabolism had a much greater contribution to the discriminant function than the structural changes caused by the fungal invasion. The wavelengths identified in this study could be helpful in developing non-destructive multispectral sensing systems to effectively scan and detect almonds with fungal infection.