Visible-near-infrared hyperspectral imaging for prediction-guided control of degreasing and degossypolization solvent dosage during concentrated degossypolized cottonseed protein preparation

Cottonseed is a useful protein resource, but the presence of gossypol limits its utilization in industry. Concentrated degossypolized cottonseed protein (CDCP) reduces gossypol content while increasing protein concentration. However, traditional processing uses fixed parameters, which may overlook cottonseed variability and could lead to solvent waste, potentially raising environmental concerns. This study aimed to establish a rapid prediction model for cottonseed fat and free gossypol content using Visible-near-infrared hyperspectral imaging (Vis-NIR-HSI), enabling precise solvent regulation during CDCP processing. Spectral data from 23 cottonseed varieties were analyzed using partial least squares regression (PLSR). The fat prediction model achieved a test set coefficient of determination (R²) of 0.9147, root mean square error (RMSE) of 1.1492, and residual prediction deviation (RPD) of 3.50. The free gossypol model achieved a R² of 0.8502, RMSE of 526.5713, and RPD of 2.64. Leveraging the linear relationship between solvent consumption and cottonseed composition, the CDCP contained < 250 mg/kg free gossypol. During the degreasing process, the consumption of n-hexane ranged from 7.75 to 13.90 L/kg, while that of 90% methanol during the degossypolization process ranged from 4.60 to 11.67 L/kg. Hyperspectral imaging combined with chemometrics enabled rapid detection of key components and dynamic solvent regulation. This approach may reduce solvent consumption and its associated environmental burden. However, given the limited sample size (only 23 varieties and four mixed validation batches), the predictive models developed in this study should be considered laboratory-validated (preliminary) rather than industrially validated. Further external validation with larger and more diverse datasets is required before industrial deployment.