Purpose of the visit
The aim of this visit was the collaboration between scientists in order to produce academic publications and increase the partnership between the two institutions in the frame of the Cost Action FA0802-WG2 feed safety. In particular the STSM activities have been focused on data treatment and statistical analysis methods applied to the measurements obtained from the combination of microscopy and computer image analysis of constituents of animal origin in feedstuffs. The STSM represented also an opportunity for capitalising on the results of the SAFEED-PAP [FP6-2005-FOOD-4-B], focused on feed safety. This issue is linked with the BSE European policy, which in the TSE road map 2, addressed the possibly for a lifting of the ban on the use of processed animal proteins from non-ruminants in non-ruminant feed. The starting point when revising the current feed ban provisions should be risk-based but at the same time should take into account the control tools in place, i.e. the availability of a reliable test to identify the class/species of traces of meat and bone meal (MBM).
Description of the work carried out during the visit
The STSM Activities focuses on the following objectives: 1- Obtaining a classification model for classifying bone fragments. 2- Evaluating the accuracy of the particular model. 3- Finding a subset of variables that achieve at least the same level of classification as obtained by the full set of variables. 4- Exploring whether classification accuracy can be improved by grouping the data (sample level vs lacunae level).
The first objective has been achieved through employing two distinct classification methods: linear discriminant analysis and classification trees. The second objective is achieved through cross-validation. As observed from the preliminary inspection of the data, high correlations have been observed among the variables of the dataset. This suggested that the data can be fitted by using a lesser number of variables, which corresponds to the third objective. Alternative datasets were created from the original dataset for achieving the fourth objective.
As expected, The preliminary inspection of the data also indicate that the features of mammalian and avian origins highly overlap, but the t-tests for comparing the means of variables for mammals and avian species give statistically significant differences. This suggests that grouping the data and taking the means (or similar descriptors) to represent the grouped data, can reduce the variability in the data and give higher levels of accuracy. Most of the activities have been focused on land animals, even though a discussion and study on sea mammals data has been started.
Description of the main results obtained
A manuscript will be finalised and submitted for publication after the end of the short term scientific mission. Financial support from the Feed for Health COST Action will be acknowledged in the final submitted manuscripts.
Future collaboration with the host institution
The collaboration with the host institute was considered very successful. Based on the successful collaboration during this short term scientific visit, the host and guest institutes are willing to continue the collaboration. The continuation of this collaboration will be achieved by means of finalising and following the publication of the scientific articles and by means of further collaborative research and publication in the future (on sea mammals).
Projected publications/articles resulting or to result from the STSM
The articles to result from this short term scientific mission are currently titled (tentative title):
“Microscopic methods in association with computer image analysis for the identification of processed animal proteins in feed”. The will be submitted to the international journal Food Additives & Contaminants: Part A (Impact Factor 2.131; Five-Year IF: 2.527 ). The possibility for further publications will be explored.
During the STSM there was also the opportunity for a scientific visit, which included a presentation about the action and its progress at the University of Reading, Reading (UK). This side activity has been organised by Prof. Ian Givens (WG1 leader).