Assessment of cold-induced sweetening in stored potatoes using hyperspectral imaging and elastic-net logistic regression with coefficient thresholding.

Cold storage improves the storability of potatoes; however, it can cause cold-induced sweetening (CIS), which promotes the formation of harmful compounds during processing. CIS-affected potatoes are difficult to distinguish based solely on the external color of unpeeled potatoes. In this study, hyperspectral imaging was applied to non-destructively assess CIS levels in potatoes. Potatoes were stored at different temperatures (1, 5, and 10 °C) with or without reconditioning; changes in color and metabolites were analyzed. Low storage temperatures (≤5 °C) increased chilling-related metabolite levels, accompanied by an increased browning index in potato flesh, whereas reconditioning decreased them. Based on reducing sugar profiles, CIS levels were classified into four stages (CS0-CS3). An elastic-net logistic regression model with coefficient thresholding (ENLR-CT) classified CIS levels with 92.6% accuracy using 55.2% of the total wavelengths, demonstrating the potential of hyperspectral imaging for rapid and non-destructive potato quality monitoring.