BACKGROUND: Rice grains can be contaminated easily by certain fungi during storage and in the market chain, thus generating a risk for humans. Most classical methods for identifying and rectifying this problem are complex and time-consuming for manufacturers and consumers. However, E-nose technology provides analytical information in a non-destructive and environmentally friendly manner. Two-feature fusion data combined with chemometrics were employed for the determination of Aspergillus spp. contamination in milled rice. RESULTS: Linear discriminant analysis (LDA) indicated that the efficiency of fusion signals ('80th s values' and 'area values') outperformed that of independent E-nose signals. Linear discriminant analysis showed clear discrimination of fungal species in stored milled rice for four groups on day 2, and the discrimination accuracy reached 92.86% by using an extreme learning machine (ELM). Gas chromatography-mass spectrometry (GC-MS) analysis showed that the volatile compounds had close relationships with fungal species in rice. The quantification results of colony counts in milled rice showed that the monitoring models based on ELM and the genetic algorithm optimized support vector machine (GA-SVM) (R2 = 0.924-0.983) achieved better performances than those based on partial least squares regression (PLSR) (R2 = 0.877-0.913). The ability of the E-nose to monitor fungal infection at an early stage would help to prevent contaminated rice grains from entering the food chains. CONCLUSIONS: The results indicated that an E-nose coupled with ELM or GA-SVM algorithm could be a useful tool for the rapid detection of fungal infection in milled rice, to prevent contaminated rice from entering the food chain.
BACKGROUND:Rice grains can be contaminated easily by certain fungi during storage and in the market chain, thus generating a risk for humans. Most classical methods for identifying and rectifying this problem are complex and time-consuming for manufacturers and consumers. However, E-nose technology provides analytical information in a non-destructive and environmentally friendly manner. Two-feature fusion data combined with chemometrics were employed for the determination of Aspergillus spp. contamination in milled rice. RESULTS: Linear discriminant analysis (LDA) indicated that the efficiency of fusion signals ('80th s values' and 'area values') outperformed that of independent E-nose signals. Linear discriminant analysis showed clear discrimination of fungal species in stored milled rice for four groups on day 2, and the discrimination accuracy reached 92.86% by using an extreme learning machine (ELM). Gas chromatography-mass spectrometry (GC-MS) analysis showed that the volatile compounds had close relationships with fungal species in rice. The quantification results of colony counts in milled rice showed that the monitoring models based on ELM and the genetic algorithm optimized support vector machine (GA-SVM) (R2 = 0.924-0.983) achieved better performances than those based on partial least squares regression (PLSR) (R2 = 0.877-0.913). The ability of the E-nose to monitor fungal infection at an early stage would help to prevent contaminated rice grains from entering the food chains. CONCLUSIONS: The results indicated that an E-nose coupled with ELM or GA-SVM algorithm could be a useful tool for the rapid detection of fungal infection in milled rice, to prevent contaminated rice from entering the food chain.
Authors: Ali Khorramifar; Mansour Rasekh; Hamed Karami; James A Covington; Sayed M Derakhshani; Jose Ramos; Marek Gancarz Journal: Molecules Date: 2022-05-30 Impact factor: 4.927