Predicting the Freshness of Starch-Coated Snakehead Fish Fillets During Storage Using Hyperspectral Imaging Combined...

Area of Science:

• Food Science
• Analytical Chemistry
• Machine Learning

Background:

• Assessing fish fillet freshness is crucial for food safety and quality.
• Starch coating can extend fish shelf life, but predicting its effectiveness requires advanced methods.
• Traditional methods for freshness evaluation are often destructive and time-consuming.

Purpose of the Study:

• To develop a robust method for predicting the freshness of starch-coated snakehead fish fillets.
• To evaluate the effectiveness of hyperspectral imaging (HSI) combined with transfer learning (TL) for this purpose.
• To compare prediction models across different storage durations (refrigerated and frozen).

Main Methods:

• Hyperspectral imaging (HSI) was used to capture spectral data from fish fillets.
• Transfer learning (TL) techniques, specifically a domain transfer convolutional neural network (DT-CNN), were applied.
• Models were trained and validated using data from both short-term refrigerated and long-term frozen storage conditions.

Main Results:

• Starch coating significantly delayed quality degradation, including texture loss, pH increase, TVB-N accumulation, and lipid oxidation.
• The DT-CNN model demonstrated improved robustness in freshness prediction across different storage stages compared to models using only short-term or long-term data.
• The best performing DT-CNN model, utilizing VIS spectra, achieved high accuracy for TBA and TVB-N prediction in starch-coated fillets (R_P² = 0.76 for TBA, R_P² = 0.85 for TVB-N).

Conclusions:

• HSI combined with TL offers a promising non-destructive approach for evaluating the freshness of starch-coated snakehead fish fillets.
• The DT-CNN model effectively addresses spectral distribution shifts and complex quality changes during storage.
• This technology can aid in ensuring the quality and safety of seafood products during storage and distribution.