Aflatoxin detection and quality assessment in dried red chilli (Capsicum annum L.): A multifaceted approach using physical, microstructural, and machine learning techniques

Abstract

India contributes nearly 43 % of global dried red chilli production; however, inadequate post-harvest handling and poor storage conditions frequently lead to fungal contamination and nutritional degradation. Dried chillies are highly susceptible to infection by Aspergillus flavus and A. parasiticus, which produce aflatoxins, a serious food safety concern during storage. This study investigated the physicochemical and microstructural deterioration of healthy and infected red chillies and explored an objective image-based grading approach to minimize contamination. Whole chilli samples were analysed for physical attributes such as shape, size, mass, bulk density, and surface characteristics, while powdered samples were evaluated for water activity, colour, ash content, microbial load, and compositional changes using Fourier Transform Infrared (FTIR) spectroscopy. Scanning Electron Microscopy (FESEM) and Energy Dispersive X-ray (EDAX) analyses revealed distinct microstructural and elemental differences between healthy and infected samples, with the latter showing higher water activity and microbial counts indicative of fungal growth. FTIR spectra of infected and market samples exhibited characteristic peaks at 1548 cm⁻¹and 1711 cm⁻¹, confirming aflatoxin-associated functional groups (C=O stretching and amide II vibrations). To enhance pre-storage sorting efficiency, a deep learning model (YOLOv8m) was developed to classify and grade red chilli pods based on colour and morphology, achieving 92 % accuracy. This integrated approach provides a rapid, non-destructive method to identify and remove contaminated pods, thereby reducing aflatoxin risk and improving the safety and storability of dried chillies.