Predicting and Visualizing the Distribution of Cryoprotectant L-Proline in Litopenaeus Vannamei by Machine Learning and Hyperspectral Imaging

Cryoprotectants are vital for preserving Litopenaeus vannamei at low temperature, with their penetration in shrimp muscle being crucial for exerting cryoprotective effect. However, traditional methods for detecting cryoprotectants are often destructive, time-consuming, and impractical for real-time applications. To address this limitation, this study proposes a non-destructive approach combining hyperspectral imaging (370–1030 nm) and machine learning for rapid quantification and spatial visualization of cryoprotectant L-Proline in peeled shrimp. Three hundred predictive models were constructed and compared through different preprocessing methods, feature extraction techniques, and algorithms. The optimal model (F-SG-1st-IFS-BP-ANN) achieved a coefficient of determination (R 2 ) of 0.98, a root mean square error (RMSE) of 0.42 mg/g, and a residual prediction deviation (RPD) of 8.06 on the prediction set, with training and prediction times of 0.59 s and 0.0001 s, respectively. Based on this model, the proline distribution within shrimp muscle was clearly visualized, allowing for precise discrimination of concentration gradients. This work demonstrates a novel and efficient strategy for monitoring cryoprotectant diffusion, which can significantly contribute to process optimization and online quality control in the production of frozen shrimp products.