Recent advances in spectroscopy combined with artificial intelligence for detection of food contaminants

Hazardous substances in food, such as heavy metals, pesticide residues, antibiotics, and mycotoxins, pose a significant threat to human health, making their rapid and accurate detection crucial for ensuring food safety. Traditional analytical methods for these hazards are often time-consuming, costly, and require complex sample preparation, limiting their applicability in on-site inspections. To overcome these limitations, novel approaches combining spectroscopic techniques with artificial intelligence (AI) are being actively researched. Spectroscopy allows for the non-destructive acquisition of unique spectral information from samples, while AI can learn meaningful patterns from complex spectral data to identify the presence or predict the concentration of hazardous substances with precision. This review examines recent research trends in the application of various spectroscopic techniques—including near-infrared (NIR) spectroscopy, hyperspectral imaging (HSI), surfaceenhanced Raman spectroscopy (SERS), and laser-induced breakdown spectroscopy (LIBS)—combined with AI models such as principal component analysis (PCA), support vector machines (SVM), random forests (RF), convolutional neural networks (CNN), and transfer learning. Case studies on the AI-based spectroscopic detection of major food contaminants, including heavy metals (e.g., lead, cadmium), pesticide residues, antibiotic residues (e.g., tetracycline, norfloxacin, ciprofloxacin), and mycotoxins (e.g., aflatoxin B1, zearalenone), demonstrate improved detection sensitivity, accuracy, and speed compared to conventional methods. In particular, data fusion strategies and advanced AI algorithms have significantly enhanced the performance of detecting trace hazardous substances within complex food matrices. However, challenges such as the standardization of spectral data, construction of large-scale training datasets, ensuring model interpretability, and obtaining regulatory approval remain for the practical field application of AI-based spectroscopic technologies. Future advancements, including multimodal spectral data integration, reinforcement learning-based feature exploration, and the development of real-time monitoring systems linked with portable sensors, are expected to further advance AI-spectroscopy fusion technology, playing a key role in upgrading safety management systems throughout the food supply chain and protecting public health.