Microplastic quantification and chemical characterization in salt samples, using stereomicroscopy, smartphone camera and supervised machine learning tools.

The microplastic (MPs) pollution, especially in the oceans, has a direct impact on seafood products, particularly sea salt. This is a growing concern due to the high daily consumption of salt and the potential adverse health effects associated with MPs exposure. This study presents a rapid, cost-effective, and green approach for MPs-detection in food samples employing stereomicroscopy, smartphone-cameras, image analysis, and Machine-Learning tools. The method integrates accessible image acquisition with open-source software (Ilastik/FIJI) to enable automated detection and measurement of two common morphologies: filaments and fragments, providing a straightforward alternative for laboratories with limited resources. The developed models demonstrated high sensitivity (100% filaments, 76.5% fragments) and acceptable specificity (76.4% filaments, 72.2% fragments), with an average recovery of 108.6% for spiked samples. Upon application to sea salt samples from México, the model identified 1950 ± 354 MPs pieces/kg of salt. Among the detected MPs, filaments were the predominant morphology (1235 ± 263 MPs pieces/kg), while fragment count was 715 ± 167 MPs pieces/kg. Polypropylene, polyethylene, Polyvinyl chloride, and cellophane were confirmed via FTIR analysis. The method was successfully applied to 6 additional food matrices (e.g., sugar, meat tenderizer). The proposed methodology is delimited to the detection of microplastics ≥300 μm (filaments) and ≥200 μm (fragments), and therefore does not account for smaller particles. The proposed method offers a rapid, replicable and accessible alternative that enables high-throughput preliminary microplastic detection while reserving chemical identification for complementary spectroscopic analysis, thus suggesting a potential for routine application in microplastic monitoring.