Data-Driven Prediction of Microbiological Quality in Oregano-Active Packaged Gilthead Sea Bream Fillets using...

Area of Science:

• Food Science
• Analytical Chemistry
• Microbiology

Background:

• Seafood quality monitoring is crucial for consumer safety and shelf-life prediction.
• Traditional microbiological methods are time-consuming and destructive.
• Non-destructive techniques are needed for rapid assessment of fish spoilage.

Purpose of the Study:

• To evaluate Fourier-transform infrared (FTIR) spectroscopy and multispectral imaging (MSI) combined with machine learning (ML) for monitoring microbiological quality changes in gilthead sea bream fillets.
• To assess the impact of edible films (Na-alginate with oregano essential oil) and storage conditions (aerobic vs. vacuum, various temperatures) on fish spoilage.
• To develop predictive models for total viable counts (TVC) using spectral data.

Main Methods:

• Gilthead sea bream fillets were packaged with edible films (with/without oregano essential oil) and stored under aerobic and vacuum conditions at different temperatures (0, 4, 8, 12°C).
• FTIR spectroscopy and MSI were used to collect spectral data from the fillets during storage.
• Machine learning regression models (linear and non-linear) were developed to predict TVC from spectral data, with emphasis on data preprocessing and model optimization.

Main Results:

• Pseudomonas spp. and H2S-producing bacteria were identified as key spoilage organisms, with spoilage accelerated by higher temperatures.
• Vacuum packaging significantly delayed spoilage compared to aerobic storage.
• Both FTIR and MSI, when combined with optimized ML models, demonstrated potential for predicting TVC, with linear models showing consistent performance.
• Essential oil treatment had minor antimicrobial effects but influenced sensory and spectral properties.

Conclusions:

• FTIR spectroscopy and MSI coupled with machine learning are practical, non-destructive tools for real-time prediction of microbial spoilage in gilthead sea bream fillets.
• Optimized preprocessing and modeling significantly enhance the predictive accuracy of spectral data for seafood quality assessment.
• These technologies offer a promising alternative to traditional methods for monitoring seafood quality and ensuring safety.