Low-cost grain sorting technologies to reduce mycotoxin contamination in maize and groundnut

Authors: AOUN, MERIEM - Cornell University; STAFSTROM, WILLIAM - Cornell University; PRIEST, PAIGE - Cornell University; FUCHS, JOHN - The Widget Factory; WINDHAM, GARY - Retired ARS Employee; Williams, William; NELSON, REBECCA - Cornell University

Submitted to: Food Control
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/14/2020
Publication Date: 5/22/2020
Citation: Aoun, M., Stafstrom, W., Priest, P., Fuchs, J., Windham, G.L., Williams, W.P., Nelson, R.J. 2020. Low-cost grain sorting technologies to reduce mycotoxin contamination in maize and groundnut. Food Control. 118:1-15. https://doi.org/10.1016/j.foodcont.2020.107363.
DOI: https://doi.org/10.1016/j.foodcont.2020.107363

Interpretive Summary: Widespread contamination of foods by mycotoxins continues to be a public health hazard in sub-Saharan Africa, with maize and groundnut being the major sources of contamination. This study was undertaken to determine whether grain sorting can reduce mycotoxin contamination in grain lots by removing toxic kernels. We tested a set of sorting methods for effectiveness in reducing mycotoxin concentration in maize and groundnut from a variety of genotypes and environments. Kernel bulk density (KBD) and 100-kernel weight (HKW) were associated with aflatoxins (AF) and fumonisins (FUM) concentrations. A low-cost sorter prototype (the ‘DropSort’ device) that stratified maize grain based on KBD and HKW was more effective in reducing FUM than AF, although neither toxin was consistently reduced below regulatory limits. DropSorting following size sorting was the fastest and most effective method in reducing FUM to under 2 ppm. None of the sorting methods reduced AF levels in maize grain to acceptable levels in heavily AF-contaminated grain. Analysis of individual kernels showed that kernels with low concentrations of AF had higher weight, volume, density, length, and width and lower sphericity than those with high AF. Single kernel weight was the most significant predictor of AF concentration. The DropSort excluded kernels with lower single kernel weight, volume, width, depth, and sphericity. For groundnut, the DropSort separated grain based on HKW and did not significantly reduce AF concentrations. Size sorting and visual sorting effectively separated groundnut grain based on AF levels. Although sorting based on physical kernel attributes was not consistently successful in reducing mycotoxins to acceptable levels, it provides a valuable tool for reducing sampling error associated with mycotoxin quantification.

Technical Abstract: The widespread contamination of foods by mycotoxins is a public health hazard in sub-Saharan Africa, with maize and groundnut being the major sources of contamination. This study was undertaken to assess the hypothesis that grain sorting can be used to reduce mycotoxin contamination in grain lots by removing toxic kernels. We tested a set of sorting methods for reducing mycotoxin levels in maize and groundnut from a variety of genotypes and environments. We found that kernel bulk density (KBD) and 100-kernel weight (HKW) were associated with aflatoxins (AF) and fumonisins (FUM) levels in maize grain. A low-cost sorter prototype (the ‘DropSort’ device) that stratified maize grain based on KBD and HKW was more effective in reducing FUM than AF, though neither toxin was consistently reduced below the regulatory limits. We then evaluated the effectiveness of the DropSort when combined with either size or visual sorting. DropSorting following size sorting was the fastest and most effective method in reducing FUM to under 2ng/g. None of the sorting methods was effective in reducing AF levels in maize grain to acceptable levels, especially in the heavily AF-contaminated grain. Analysis of individual kernels showed that low-AF maize kernels had higher weight, volume, density, length, and width and lower sphericity than those with high AF. Single kernel weight was the most significant predictor of AF concentration. The DropSort excluded kernels with lower single kernel weight, volume, width, depth, and sphericity. For groundnut, the DropSort separated grain based on HKW and did not significantly reduce AF concentrations. Size sorting and visual sorting effectively separated groundnut grain based on AF levels. Although sorting based on physical kernel attributes was not consistently successful in reducing mycotoxins to acceptable levels, we propose that it may provide a valuable tool for reducing sampling error associated with mycotoxin quantification.