Multispectral image analysis for robust prediction of astaxanthin coating

Area of Science:

• Aquaculture Nutrition
• Food Science
• Spectroscopy

Background:

• Astaxanthin is a vital pigment in aquaculture feeds, influencing fish health and coloration.
• Accurate quantification and identification of astaxanthin are crucial for feed quality control and production efficiency.
• Current methods for astaxanthin analysis can be time-consuming and may not be suitable for rapid screening.

Purpose of the Study:

• To explore the potential of multispectral image analysis for predicting astaxanthin type and concentration in aquaculture feed pellets.
• To evaluate the efficacy of different chemometric methods for astaxanthin quantification and classification.
• To identify optimal spectral regions for astaxanthin analysis in feed pellets.

Main Methods:

• Utilized a VideometerLab system capturing spectral data across 385-1050 nm.
• Employed linear discriminant analysis (LDA) and sparse LDA for classification and variable selection.
• Applied partial least squares regression (PLSR) for predicting astaxanthin concentration levels.
• Tested natural and synthetic astaxanthin across various concentrations and four feed pellet recipes.

Main Results:

• PLSR successfully predicted synthetic astaxanthin concentration, even when models were calibrated across all feed recipes.
• The developed prediction models demonstrated adequate performance for screening astaxanthin concentration across all tested recipes.
• Classification models accurately predicted the type of astaxanthin (natural vs. synthetic) using only ten spectral bands.
• Variable selection indicated that spectral bands within the visible range were most influential for astaxanthin prediction.

Conclusions:

• Multispectral imaging offers a viable, non-destructive method for predicting astaxanthin concentration and type in aquaculture feed pellets.
• The findings support the use of PLSR and LDA for rapid quality control and screening of astaxanthin in feed production.
• Visible spectral bands are key for developing accurate models for astaxanthin analysis in aquaculture feeds.