Integration of Raman spectroscopy and machine learning for species-specific assessment of fish freshness with insights into biogenic amines

Ensuring fish quality and safety is critical for protecting consumer health, particularly with respect to freshness loss and biogenic amines (BAs) accumulation. In this study, Raman spectroscopy was integrated with machine learning to enable rapid, non-destructive assessment of fish freshness and prediction of BAs. Raman spectra were collected from five fish species at different freshness stages and analyzed using a combination of unsupervised and supervised learning approaches to discriminate freshness levels and species. Spectral analysis revealed prominent freshness markers, characterized by intensity changes in specific Raman bands, particularly the Amide I band (1654 cm −1 ) and aliphatic C-H stretching vibrations (1451 cm −1 ), which were closely associated with protein degradation and lipid oxidation during storage and contributed significantly to freshness discrimination, and were further supported by two-dimensional correlation spectroscopy, confirming their sequential evolution during spoilage. Compared with conventional chemometric methods, machine learning models demonstrated superior classification performance, with a convolutional neural network model achieving the highest accuracy of 94.80%. For safety assessment, partial least squares regression and support vector regression models were developed to predict BAs, both exhibiting strong predictive performance (Rp > 0.9). Based on histamine concentrations, samples could be reliably classified into three categories: high histamine (>50 mg/kg), medium histamine (10-50 mg/kg) and low histamine (<10 mg/kg), with classification accuracies of 80.00%, 88.23% and 100.00%, respectively. Overall, the integration of Raman spectroscopy with machine learning provides an effective framework for rapid evaluation of fish freshness and early identification of safety risks, supporting quality monitoring and control in aquatic products.