This study evaluated the combined effect of the modified atmosphere packaging (MAP1: 60% CO
2, 10% O
2/30% N
2 & MAP2: 40% CO
2, 30% O
2/30% N
2), and active packaging of oregano essential oil (1%
v/
w) used as a natural preservative, on the quality and shelf-life extension of fresh sea bream fillets. The samples were stored at four different temperatures (0, 4, 8, and 12 °C), and a microbiological analysis, pH measurements, and sensory evaluations were performed. In parallel, spectral data were obtained using three different spectroscopic sensors (two MultiSpectral Imaging devices and an FT-IR one), and nine different machine-learning regression models were applied to predict the microbiological counts. Oregano essential oil positively affected preservation, reducing microbial growth by 0.5 to 2 log CFU/g, and extending the fillets’ shelf life by up to 48 h based on sensory evaluation. Regarding the sensors’ data, the examined nine models exhibited encouraging results for the rapid microbiological assessment, with the FT-IR data showing the best performance for evaluating the microbiological population. Among the tested algorithms, the least Angle Regression (lars) achieved the best performance for both the flesh and skin datasets, with RMSE values of 0.6075 and 0.5953, MAE of 0.3008 and 0.4567, R
2 of 0.8858 and 0.7532, and accuracy of 87% and 91%, respectively. The Benchtop-MSI showed the best predictive performance for flesh (RMSE = 0.5926, MAE = 0.4876, R
2 = 0.7338, and Accuracy = 92%), while the artificial neural network (nnet) performed best for skin (RMSE = 0.6761, MAE = 0.5247, R
2 = 0.6560, and Accuracy = 84%). Regarding the Portable-MSI, the artificial neural network model gave the highest accuracy for flesh (RMSE = 0.5908, MAE = 0.4663, R
2 = 0.5903, and Accuracy = 87%), whereas principal component regression was the most effective for skin (RMSE = 0.6600, MAE = 0.5413, R
2 = 0.5534, and Accuracy = 83%).
Full article