Published in

SAGE Publications, Journal of Near Infrared Spectroscopy, 4(9), p. 297-311, 2001

DOI: 10.1255/jnirs.315

Links

Tools

Export citation

Search in Google Scholar

Use of discriminant analysis on visible and near infrared reflectance spectra to detect adulteration of fishmeal with meat and bone meal

Journal article published in 2001 by Ian Murray, Lorna S. Aucott ORCID, Aucott Ls, Ian H. Pike, Pike Ih
This paper was not found in any repository, but could be made available legally by the author.
This paper was not found in any repository, but could be made available legally by the author.

Full text: Unavailable

Green circle
Preprint: archiving allowed
Green circle
Postprint: archiving allowed
Red circle
Published version: archiving forbidden
Data provided by SHERPA/RoMEO

Abstract

Since the link between feeding ruminant-derived meat and bone meal (MBM) and the occurrence of bovine spongiform encephalopathy (BSE) and its human equivalent variant Creutzfeldt–Jakob disease (vCJD) has been established, it is imperative that potentially infective material is excluded from the food chain. To this end, a Partial Least Squares (PLS) discriminant analysis, using visible and NIR reflectance spectra, was developed on a calibration set of 67 samples consisting of 22 authentic fishmeal (FM) specimens and 45 fishmeals deliberately adulterated with meat and bone meal (MBM) at 3%, 6% and 9% by weight, respectively; 15 samples were prepared at each concentration. Each material was unique in that any one fishmeal or meat and bone meal was used once only. In an independent validation set of 69 specimens prepared in exactly the same way, the discriminant successfully detected 44 out of 45 adulterated specimens with an error of one false positive among the remaining 24 pure fishmeals. Performance was tested on two independent monochromators and a canonical discriminant algorithm gave similar results. The NIR region (1100–2500 nm) or the visible and Herschel IR (400–1100 nm) alone did not perform as well as the combined visible and NIR regions. Modified PLS calibration for MBM % on the complete set of 136 specimens gave a standard error of calibration ( SEC) of 0.85% and coefficient of determination ( R2) of 0.94 based on the use of nine factors. Selection of appropriate and representative specimens for calibration and validation had a much greater effect on performance than any data treatment, scatter correction or the number of cross-validations or PLS factors used. Misclassification errors arose from specimens which were global H outliers having atypical spectra not represented in the calibration model. We believe that visible-NIR reflectance spectroscopy could routinely provide the first line of defence of the food chain against accidental contamination or fraudulent adulteration of fishmeal with meat and bone meal, which could present a health risk from transmissible spongiform encephalopathies (TSEs).