Layered data fusion using spectral imaging to assess myoglobin content of vacuum-packaged veal meat during storage

Myoglobin content and status of myoglobin changes act as an indicator of the colour and oxidative status of meat, and the non-invasive prediction of myoglobin in meat products is vital for real-time assessment. This study aims to develop a methodology to predict the myoglobin content and forms in vacuum-packaged veal products (animal of controlled sex, age and source) at different storage durations (day 0 to 15) using spectral imaging sensors (one type of the non-destructive spectroscopic sensors). Three levels of data fusion (data-level, feature-level, and decision-level) were tested on multi-parameters extracted from the raw 3D spectral data: average spectra, standard deviation (STD) spectra, spatial information (Gray Level Co-occurrence Matrix (GLCM)), and the hyperspectral imaging-root mean square (HSI-RMS). Data analysis pipeline was embedded into different data fusion strategy, including pre-processing, wavelength selection, and modelling methods. Among the various combinations, decision-level fusion achieves the highest performance for predicting oxymyoglobin content (%) (R 2 CV=0.72, RMSECV=1.08%) by reassigning the modelling results from single-parameter. The key novelty is the data processing pipeline from spectral imaging sensors to evaluate packaged meat products directly. Rather than proposing a new chemometric algorithm, the study tests how established spectral and imaging modelling methods perform in a through-package veal application. These findings can support future development of rapid screening tools for monitoring the colour and oxidative status of packaged meat products after broader validation.