NR AOWI
AU Martin,T.C.; Cawthraw,S.; Petrich,W.; Moecks,J.; Rauh,J.
TI Infrared Spectroscopy of Serum and Pattern Recognition Analysis - an Ante-Mortem Diagnostic tool for BSE?
QU International Conference - Prion diseases: from basic research to intervention concepts - TSE-Forum, 08.10.-10.10.2003, Gasteig, München - Poster session - DG-67
PT Konferenz-Poster
AB
Bovine Spongiform Encephalopathy (BSE) has had a major impact on the cattle industry; worldwide. The economic impact has been severe, but of far greater impact has been the implications for human health. This is because a new form of human disease, variant Creutzfeldt-Jakob disease (nvCJD), has been linked to the emergence of BSE in the UK.
To date, the application of ruminant feed bans and the development of post-mortem diagnostic tests based on brain sampling have been used to control and manage BSE. However, the development of an ante-mortem diagnostic test based on serum samples may provide a convenient and cost-effective method of surveillance.
Pattern recognition is increasingly being applied to the diagnosis of various human disease such as rheumatoid arthritis, diabetes mellitus and cancers. It has also been applied to experimental and natural prion infections of animals. In this study we explore the application of infrared spectroscopy and pattern recognition tools to the analysis of dried films of bovine serum which, were obtained from cattle naturally infected with BSE. Utilising a "Diagnostic Pattern Recognition" research system, we have investigated critical parameters such as confounding factors and reproducibility. The pattern recognition tools (classifiers) developed for this study were based on supervised learning and included the use of artificial neural networks (ANN). In total, the sera from 220 BSE suspect positive animals and 202 suspect negative animals were analysed. The results show that the DPR system has the capability to differentiate between BSE-positive and BSE-negative animals well beyond a random assignment. Given a p-value of p>0.001 the differentiation is highly unlikely to be due to chance.
The significance of these results, together with any further improvements to the accuracy of the system resulting from classifier modifications will be presented.
AD Trevor C. Martin, Saira Cawthraw, Veterinary Laboratory Agency, UK; Wolfgang Petrich, Joachim Moecks, Roche, Germany; Jurgen Rauh, Roche, Switzerland
SP englisch
PO Deutschland