Table 8.
Bibliographic information (i.e., source) published in the last three years regarding the use of NIR for predicting mycotoxin contaminations in different matrices.
| Feed Matrix | Target Mycotoxin | Wavelength | Statistical Model * | Results Obtained | Practical Application | Source |
|---|---|---|---|---|---|---|
| Ground corn samples | Fumonisin B1 and B2 | 900–1700 nm | PLS, SVM, and LPLS-S | R2 prediction = 0.71–0.91 RMSEP = 12.08–22.58 mg/kg |
Pocket-sized NIR spectrometers controlled by a smartphone | [65] |
| PCA, PLS-DA, and SVM-DA | Prediction accuracy = 86.3–88.2% Error in prediction = 11.8–13.7% |
|||||
| Rice (Oryza sativa L.) | Aflatoxin B1 | 400–2498 nm | MSA + PLS | Low-aflatoxin-level (≤35 μg/kg): R2 calibration = 0.72–0.99 RMSEC = 0.11–5.02 μg/kg High-aflatoxin-level (>35 μg/kg): R2 calibration = 0.72–0.99 RMSEC = 0.56–13.74 μg/kg |
Monitoring aflatoxin B1 contamination in milled rice during postharvest storage | [66] |
| Almonds | Aflatoxin B1 | 900–1700 nm | PLS | R2 = 0.786–0.958 RMSEP = 0.089–0.197 μg/g |
Commercial application | [67] |
| Distiller’s dried grains | Fumonisin B1 and B2 | 400–2500 nm | PLS | FB1 R2 = 0.80 FB2 R2 = 0.79 |
Potential to support decision making regarding the use of feed ingredients and, consequently, to protect animal health | [68] |
| Barley (Hordeum vulgare) | Deoxynivalenol (cut off limit cut off 1250 µg/kg) | 10,000 cm−1–4000 cm−1 | PLS-DA | Sensitivity in cross-validation = 90.9% Specificity in cross-validation = 89.9% |
Green technique to monitor DON contamination | [69] |
| Corn products | Fusarium verticillioides and F. graminearum | 1000–2500 nm | PLS-DA | Accuracy = 99.7% | Monitoring the safety of feed and food supply | [70] |
| Wheat flour | Deoxynivalenol | PLS-DA and PC-LDA | Contamination level ≤ 450 μg kg−1 Accuracy (PLS-DA) = 85–87.5% Error (PLS-DA) = 10–15% error; Accuracy (PC-LDA) = 85% Error (PC-LDA) = 10–15% error |
Screening method to evaluate DON contamination to support decision making in industries | [71] |
* PLS = Partial least squares; SVM = Support vector machine; LPLS-S = local PLS based on global PLS score; PCA = principal component analysis; PLS-DA = partial least squares discriminant analysis; SVM-DA = support vector machine discriminant analysis; MSA = modified simulated annealing; PC-LDA = Principal Component Analysis-Linear Discriminant Analysis.