Exploiting Hyperspectral Imaging (HSI) and chemometrics for the rapid on-site measurement of aflatoxin in chillies

Naturally occurring toxins (e.g., mycotoxins) remain one of the greatest food safety concerns globally. Chilli (Capsicum spp.) is one of the most important and largest produced spice in India, however, they are continuously susceptible to aflatoxin (AF) contamination at every stage of production due to the climatic conditions. High temperatures, high rainfall and humidity in these areas are the perfect conditions for fungal proliferation and toxin production. The main technological challenge is that AF testing commonly relies on analytical techniques such as, liquid chromatography with tandem mass spectrometry (LC-MS/MS) or high-performance liquid chromatography combined with fluorescence (HPLC-fluor) which are extremely accurate and sensitive. However, they are not applicable outside of the laboratory and require highly skilled and trained personnel to conduct. Immunosensor based tests such as, ELISAs and lateral flow tests (LFT) are also available. However, these tests require lengthy procedures and incubations, multiple steps, and the incorporation of recognition elements, which can be costly and ultimately lead to cross reactivity and/or matrix effects. Therefore, there is a demand to improve on-site analysis using rapid, portable, and cost-effective techniques. Hyperspectral Imaging (HSI) is rapid, portable and can provide enhanced information about the chemical and physical properties of samples through spatial imaging combined with spectroscopy. During this preliminary study, chilli samples from India (n=300) were analysed using two analytical techniques. HSI was conducted alongside confirmatory analysis, LC-MS/MS to quantify levels of AF and kojic acid. The correlation between the two analytes was addressed and the LC-MS/MS results were fused with the spectral data obtained using HSI for statistical modelling. Several statistical models were developed using chemometrics to determine low, medium and high-quality chillies and AF contamination with detection limits down to 20 ppb. A leave-30%-out cross validation was conducted to assess model predictability and accuracy. Overall, the results from HSI (combined with LC-MS/MS) and chemometric modelling revealed much promise for the future on-site determination of quality parameters (e.g., discoloration and degradation) and AF contamination in chillies. Future work will focus on improving datasets and model robustness.