Paltanin C.1, A. Campagnoli1, L. van Raamsdonk2 , and L. Pinotti 1
1 Dept. of Veterinary Sciences and Technology for Food Safety, University of Milan,
Italy
2 RIKILT- Institute of Food Safety, Wageningen, The Netherlands
*corresponding author: luciano.pinotti@unimi.it
The use of microscopic methods in association with computer image analysis to identify the source species of these feedstuff contaminants has been proposed. Image processing, integrated with morphometric measurements can provide accurate and reliable results and can be a very useful aid to the analyst in the characterisation, analysis and control of feedstuffs. The use of microscopic methods in association with computer image analysis to identify the source species of these feedstuff contaminants has been proposed.
Image processing, integrated with morphometric measurements can provide accurate and reliable results and can be a very useful aid to the analyst in the characterisation, analysis and control of feedstuffs. Reference samples containing poultry, bovine, swine, or ovine meat and bone meals (Agricultural Research Centre of Gembloux, Belgium, and SAFEED-PAP Project) were analysed by microscopic method.
In the sediment fractions of each sample lacunae images at high magnifications (40x) were obtained (Olympus BX41, Germany).
Through a digital camera (CoolSnapPro Color, Media Cybernetics; Retiga2000R Fast, Qimaging) and an image analysis software (Image-Pro Plus 6.3, Media Cybernetics Inc., Silver Springs, USA), lacunae images have been processed with different filtering procedures (HiGauss, Sculpt, etc.) of the software in order to reduce possible noises and to facilitate the count of canaliculae number in each lacuna. Lacunae (574 equally distribute between animal sources) were than elaborated to obtain for each one a monochrome mask on which area, perimeter, axis major, and axis minor were measured.
Data obtained were analyzed by both multivariate and variance analysis.
Results obtained indicated that the cumulative percent of variability accounted for by the first three components (>92%). Component 1 has large loading values on the variables: area, perimeter, and axis max; Component 2 has large loading values on the variables: axis minor followed by axis max (opposite in sign); Component 3 has large loading values on the variable number of canalicula (83%). We note that part of ovine material was well discriminated from other species when comp. 1 and 2 were considered.
Analysis of variance confirmed as previously reported the difficulty in distinguish between bovine vs. swine materials, even through number of canalicula can provide further information.
Referring to the results herein presented it can be concluded that ;
- inside the mammalian class (bovine, ovine and pig), swine material represents the major source of misclassification;
- Canaliculae number when visible in the microscope in association to other dimensional lacunae variables can be useful in species characterisation.