Development of a Real-Time Aflatoxin Detection System Using Near-Infrared Spectroscopy for Nigerian Grain Markets

This study develops and validates a real-time aflatoxin detection system using near-infrared spectroscopy combined with chemometric modelling for point-of-market grain quality assessment in Nigerian grain markets. Aflatoxin contamination of maize, groundnut, and sorghum is a major public health and trade problem in Nigeria, where mycotoxin levels frequently exceed domestic and export permissible limits. Current testing methods including ELISA and HPLC are laboratory-based, slow, and expensive, precluding their use at the grain sale point. NIR spectroscopy offers rapid, non-destructive aflatoxin estimation from spectral signatures associated with mould growth and chemical changes in grain. This study collects a reference dataset of 1,800 grain samples from maize, groundnut, and sorghum marketed in Kano, Abuja, and Ibadan, including laboratory-determined aflatoxin concentrations across the full contamination range. NIR spectra are collected using a portable instrument. Partial least squares regression, support vector regression, and deep learning spectral models are developed and compared. Model validation uses independent sample sets from different market seasons. A field-deployable prototype device is designed incorporating the validated model. Findings reveal that the deep learning NIR model achieves aflatoxin prediction accuracy with RMSEP of 3.2 micrograms per kilogram and correlation coefficient of 0.91 in independent validation, adequate for regulatory screening at the 10 micrograms per kilogram European Union limit. The portable device enables testing in less than 30 seconds per sample. The study recommends regulatory adoption of the NIR screening device for national market surveillance programmes targeting high-risk grain collection points.

Keywords: aflatoxin detection, near-infrared spectroscopy, grain quality, chemometrics, Nigeria.