Surimi manufacturing involves complex, multi-step operations in which small changes in raw material condition, formulation, and heating history can markedly alter texture, water retention, and visual quality. This review critically examines peer-reviewed studies that apply artificial intelligence to surimi and surimi-based products, focusing on work validated directly in surimi systems. Current evidence mainly supports non-destructive quality evaluation and integrity screening using imaging and vibrational spectroscopy. These applications include deep learning for classifying gel surface images, as well as chemometric and machine learning analysis of infrared, near-infrared, and hyperspectral data for quality prediction and adulteration detection. Process-linked monitoring during thermal treatment is also beginning to emerge, with one time-resolved hyperspectral imaging study demonstrating quality tracking during heating. Major barriers to industrial adoption include limited and narrowly sampled datasets, batch effects and validation designs that may overestimate predictive performance, and practical deployment challenges such as stable sensing in wet environments, instrument drift, and calibration transfer across devices and sites. The review also outlines forward-looking directions, including digital twins, adaptive control strategies, and automation, and identifies data standardization, external validation, and maintenance strategies as priorities for translating laboratory demonstrations into reliable industrial applications.
Oyinloye et al. (Thu,) studied this question.