Alternative approaches for accelerated shelf life testing (ASLT) of low-moisture food products: A case study of dried shrimp.

Alternative approaches for shelf life evaluation of food based on accelerated shelf life testing (ASLT) assisted with principal component analysis-based machine learning (ML-ASLT), and hyperspectral imaging integrated with ML-ASLT (HSI-ML-ASLT) were developed and compared with the conventional ASLT (C-ASLT). Dried shrimp was stored at 15, 30, and 40 °C and monitored for changes in sensory characteristics including odor, color and overall acceptability during the storage. Physicochemical properties of the dried shrimp were monitored to describe the changes in sensory attributes. Volatile compounds of initial and unacceptable dried shrimp were analyzed using gas chromatograph-mass spectrometer (GC-MS). The samples were also analyzed using HSI in the wavelength range of 400-1000 nm. PCA-based machine learning was applied with ASLT and HSI-ASLT to determine quality degradation rate constant and shelf life. The ML-ASLT gave the predicted shelf life of 9.87, 6.13, and 3.34 wk., while HSI-ML-ASLT gave the shelf life of 9.71, 6.51 and 3.31 wk. for dried shrimp stored at 15, 30, and 40 °C, respectively, which were more accurate than C-ASLT (10.29, 6.80 and 3.49 wk), when compared to the actual shelf life (10, 6 and 3 wk). By integrating multivariate quality attributes, the PCA-based machine learning model enhances the ASLT approach to yield precise, practical shelf life predictions for dried shrimp. Furthermore, HSI-ML-ASLT provides a rapid, simple, and nondestructive method for evaluating food shelf life without the need for chemical reagents.