Monitoring of protein secondary structure during surimi thermal processing using hyperspectral imaging: comparison of spectral, image-texture, and data-fusion strategies.

Surimi, a processed seafood product known for its distinctive texture and high protein content, poses significant challenges for maintaining quality during its complex processing. This study investigates changes in protein secondary structure content during two-stage water bath (TSWB) heating of surimi, utilizing hyperspectral imaging (HSI) for rapid, non-destructive monitoring. Surimi samples were heated at 40 °C for 30 min followed by 90 °C for 20 min, with protein secondary structure data recorded at 5-min intervals. Three predictive models were developed using hyperspectral data: spectral-based, image-based, and data fusion approaches, all based on partial least squares (PLS) regression. The spectral-based model achieved optimal performance (Rp2 > 0.98, RMSEP <0.13), while the image-based and data fusion models demonstrated comparatively lower predictive accuracy (0.55 < Rp2 < 0.78, 0.12 < RMSEP <0.96). Among the three modelling strategies, spectral models showed the best predictive performance. Furthermore, the protein secondary structure changes observed during heating enable subsequent visualization analysis. This study demonstrates the feasibility of HSI combined with chemometrics as a non-destructive approach for monitoring protein structural changes during surimi thermal processing under controlled laboratory conditions. The proposed framework provides a useful basis for improved quality assessment and may support future development of online monitoring applications in surimi processing.