Portable Visible/Near-Infrared Spectroscopic Protocol for the Rapid Classification of Aflatoxin B1 Contamination Levels in Corn

Aflatoxin B1 (AFB1) was identified as the analyte for rapid, nondestructive screening in corn using a newly developed portable visible/near-infrared (Vis-NIR) spectroscopic protocol. To address the limitations of conventional laboratory-based analyses, a portable Vis-NIR device was fabricated utilizing 17 characteristic wavelengths selected via the competitive adaptive reweighted sampling (CARS) algorithm. A deep learning-based framework was employed for spectral data augmentation and classification to categorize AFB1 contamination severity into semi-quantitative levels. Analytically, the protocol demonstrated robust performance with a classification accuracy of 93.61% on the independent test set and an outstanding identification accuracy of 98.44% for severe contamination levels (>20 ppb). The instrumental reliability was further validated by a relative standard deviation (RSD) of 1.24% across replicate measurements. These results indicate that the developed spectroscopic protocol serves as an efficient analytical tool for the primary on-site screening of AFB1 in agricultural commodities, successfully bridging the gap between sophisticated laboratory instrumentation and rapid field detection needs.