Mapping Lipid Oxidation in Chicken Meat by Multispectral Imaging of Autofluorescence

Abstract

Multispectral imaging of autofluorescence was carried out to investigate the feasibility of mapping the degree of lipid oxidation in ground chicken meat. Meat samples from both breast and thighs were collected from 32 chickens, ground and freeze-stored for different time intervals. Sixteen samples were imaged at the time, making up two sets of multispectral images, A and B. Lipid oxidation was measured by a method using 2-thiobarbituric acid reactive substances (TBARS), and samples were in the range 0.15–3.23. Principal component analysis was performed on image set A, and variation in score images of the two first components corresponded well with lipid oxidation reference values. The multivariate image regression model based on image set A was tested on set B. Pixel-wise prediction gave large individual errors, but averaging predicted values within samples improved accuracy and resulted in a correlation of 0.98. Increasing the amount of spatial variation (number of pixel vectors) in the regression models led to more robust models with lower prediction errors. The technique has potential for nondestructive investigation of distribution and kinetics of lipid oxidation in food.

Fluorescence in aldehyde model systems related to lipid oxidation

“…Many of these methods have been developed for use in pure oil systems, and then adapted to foods, but with questionable reliability because food systems are much more complex (Gullie´n-Sans & Guzma´n-Chozas, 1998;Jo & Ahn, 1999). Fluorescence spectroscopy, measured directly on intact samples, has been demonstrated to yield good indices of lipid oxidation in biological materials such as fish and both raw and heat treated meat (Aubourg, Sotelo, & Perez-Martin, 1998;Wold & Kvaal, 2000;Wold & Mielnik, 2000;Wold, Mielnik, Pettersen, Aaby, & Baardseth, 2002). It has also shown to be a sensitive technique with regard to lipid oxidation, comparable with sensory analysis and gas chromatography (Wold et al, 2002).…”

Denoising and multivariate analysis of time‐of‐flight SIMS images

“…14,15 However, standard methods of ToF-SIMS image analysis, including creating images by the selection of a single mass fragment, the sum of selected mass fragments or the ratio of selected mass fragments, frequently require a priori knowledge of the sample chemistry and ignore much of the chemical state information contained in the datasets. Multivariate image analysis (MVIA) has been applied to many types of multispectral image types, including remote sensing by Landsat images, 16 food quality inspection by autofluorescence microscopy 17,18 and polymer blend analysis by XPS 19 and ToF-SIMS 20 -23 imaging. The ability of MVIA techniques, specifically principal component analysis (PCA), to reduce the dimensionality of hyperspectral datasets while retaining much of the original spectral information has been demonstrated.…”

Utilizing three‐dimensional wavelet transforms for accelerated evaluation of hyperspectral image cubes

“…The spatial information retained after compression is analyzed by means of red-green-blue score images belonging to PC1-PC3 (RGB (1,2,3)) and PC4-PC6 (RGB (4,5,6)). The less blurring occurs and the closer the colors are to the conventional PCA results, the more spatial information has been retained in compressed cubes.…”

“…Comparing red-green-blue score images RGB(1,2,3) and RGB (4,5,6) derived from data set #1 by means of conventional PCA and by compression-based PCA using compression method #1 (Table I) with different wavelet types.…”