Key points are not available for this paper at this time.
Background Food systems contain a wide array of bioactive compounds that exert beneficial effects on human health, among which antioxidant activity (AA) plays a pivotal role in mitigating oxidative stress-related disorders. Conventional analytical methods for evaluating AA are often labor-intensive, destructive and unsuitable for high-throughput or large-scale assessment, whereas AI-enhanced vibrational spectroscopy enables rapid, non-destructive, and sustainable evaluation of AA across diverse food matrices. Scope and approach This review systematically summarizes the principles, applications, and limitations of promising vibrational spectroscopic techniques, including near- and mid-infrared spectroscopy, Raman spectroscopy, and hyperspectral imaging, in recent applications for predicting AA in foods. Notably, advances in AI-enabled strategies integrating vibrational spectroscopy, hyperparameter optimization, explainable artificial intelligence, multimodal data fusion, and automated machine learning are leveraged to enhance the accuracy and reliability of AA assessment. Finally, this review outlines future directions toward large-scale and real-world AA sensing, emphasizing end-edge-cloud collaborative modeling, blockchain-enabled e-labeling, and portable smartphone spectrometers for consumer-level implementation. Key findings and conclusion The integration of vibrational spectroscopy with advanced AI techniques enables rapid, non-destructive, and scalable evaluation of AA in complex food systems, bridging the gap between laboratory analysis and real-world applications. Deep learning-based spectral optimization, data fusion, explainable artificial intelligence, and automated machine learning frameworks significantly enhance model robustness, interpretability, and automation. Looking ahead, end-edge-cloud collaboration, blockchain-enabled traceability, and portable smartphone-based spectrometers will underpin large-scale AA prediction in food systems, fostering AI-driven evaluation of food nutritional quality and functional value.
Zha et al. (Tue,) studied this question.