Alginate concentration is an important chemical indicator for evaluating the quality and extraction efficiency of marine biological polysaccharides. In this study, for the first time, a novel Terahertz (THz) spectral imaging technique was applied to the non-destructive quantitative analysis of alginate in Alaria esculenta extract samples . To evaluate the optimal modelling strategy, three machine learning algorithms were developed and compared, including partial least squares regression (PLSR), gaussian process regression (GPR), and a three-dimensional convolutional neural network (3D-CNN). Unlike traditional approaches relying solely on mean spectral features, the 3D-CNN was designed to capture combined spatial-spectral information from the THz imaging data . The results showed that the GPR model achieved the highest prediction performance with an R² of 0.908, while the 3D-CNN model also provided reliable predictions with an R² of 0.841 . This study demonstrates the feasibility and effectiveness of terahertz spectral imaging as a non-destructive method for accurately quantifying biopolymer concentrations in food and pharmaceutical materials, opening new possibilities for rapid and non-invasive quality assessment.
李庆霞 et al. (Wed,) studied this question.