Investigation of Statistical and Imaging Methods for Luminescence Detection of Irradiated Ingredients

Sanderson, D.C.W. , Carmichael, L., Murphy, S., Whitely, V., Scott, E.M. and Cresswell, A. (2001) Investigation of Statistical and Imaging Methods for Luminescence Detection of Irradiated Ingredients. Project Report. Food Standards Agency, London, UK.

[img]
Preview
Text
58359.pdf

1MB

Publisher's URL: http://www.food.gov.uk/science/research/choiceandstandardsresearch/foodirradiation1/a05prog/a05projlist/a05006proj/

Abstract

This project investigated two potential approaches to improving the reliability of lumines-cence methods for detecting minor irradiated ingredients in foods. Whereas in the 1980’s there were no validated methods for laboratory detection of irradiated foods, work conducted in the UK and elsewhere by the mid 1990’s had resulted in the development of a series of physical, chemical and biological methods capable of detecting a range of irradiated food classes. Of these the luminescence methods embodied in EN1788 (Thermoluminescence) and EN13751 (Photostimulated luminescence) standards have been applied to detection of a vari-ety of products including herbs and spices, and seafood. In common with the other EN stan-dard methods almost all validation work had been originally conducted using pure irradiated or unirradiated ingredients. Yet application experience had shown the presence of mixed products containing both irradiated and unirradiated ingredients. A short study was commis-sioned by MAFF to investigate the impact of blending on standard EN1788 methods, and on the provisional draft EN13751 (the standard having been published in the meantime) method. This showed the impact of dilution of irradiated material between 10% and 0.1% concentra-tions on detection rates, which unsurprisingly are reduced by extreme dilution. UK labelling regulation, both before and after adoption of the European Directive on Food Irradiation, call for labelling of all irradiated ingredients regardless of concentration or origin within the final product. This study was therefore motivated by the recognition of the long term need for im-proved methods to improve reliability at low concentrations. Two complementary approaches were investigated. The project first examined whether TL data collected using the EN1788 method could be enhanced using advanced statistical proce-dures. Data sets from the SURRC TL archive, and from project CSA4790 were used both to define the characteristics of irradiated and unirradiated end members, and to assess classifica-tion methods using the controlled blending experimental data sets of CSA 4790. Multivariate analyses, based on principal components analysis and discriminant analysis of glow curve data; kinetic deconvolution approaches coupled to PCA and DA, and neural analyses were investigated and compared with detection rates achieved using expert visual classification. To complement this experiments were undertaken to explore the potential of using focussed laser stimulation to produce spatially resolved measurements from mineral grains separated from foods. Two systems were evaluated based on IR and visible band lasers. Work was under-taken to explore sample presentation and to assess the ability of this approach to distinguish mixtures of irradiated and unirradiated grains. The statistical work was successful in developing three approaches which could be used for objective identification of irradiated materials. Pure irradiated and unirradiated data sets from 150 sample pairs were obtained having searched the SUERC archive of more than 3500 lu-minescence analyses. These were used to set up multivariate analyses based on the ap-proaches outlined above. Performance in recognising irradiated ingredients using these meth-ods was then assessed with data drawn from the MAFF blending investigation, comprising 160 permutations of irradiated and unirradiated herbs and spices at 10%, 1% and 0.1% con-centrations. It was possible to achieve good detection rates with alatistical approaches, the best approaches inigated being the use of glow curve deconvolution coupwith li discrimination, and the use of neural appros. The absolute performance achieved matched that opert visual clfication utilising the revised EN1788 criterwhich were adopted within the international standauring course of this project. The use of ad-vancedtistical methods, while not adding performance, can pde objective support to visual classifications. During performance assessment it was aloted that theformance of all methods wasficiently close to infer that detections rates are most dependent on the statistical presence or absence of irradiated grains within the extracted samples used for TL analysis. This raises practical suggestions for improving detection rates at low concentrations based on the use of larger samples and more specific mineral separation approaches. These may be worth investigating further. Laser scanning approaches were also investigated using highly focussed laser beams to stimulated luminescence sequentially from different parts of separated mineral samples. Work was conducted using a system which had been developed in earlier work at SUERC, and then followed by additional investigation using an improved instrument built during the project. Initial work confirmed the feasibility of using laser scanning approaches to obtain spatially resolved luminescence data at or near the dimensions of individual mineral grains. Practical obstacles included the recognition that laser scattering from surfaces coated with mineral grains introduced an element of cross-talk between different parts of the sample, and difficulties in accurate re-positioning of the sample using the first generation prototype in-strument. Work was conducted to investigate a series of different sample presentation media to improve the former, and to incorporate high precision mechanical and optoelectronic means of re-positioning samples between initial measurements, external irradiation, and sub-sequent re-measurement. Both IR and visible band semiconductor lasers were investigated with successful production of single grain images. The short and medium term reliability of the lasers used was acceptable. The lasers used both however eventually failed, which sug-gests that long term lifetime may be an issue for further work. Of the two lasers the IR laser in particular gave a good signal to background ratio for discriminating between irradiated and unirradiated grains. Quantitative analysis of the grain resolved images confirms the potential of this approach in identifying minor irradiated components. The overall conclusions of the work are that both statistical approaches and imaging instru-ments are able to enhance current methods. The observation that visual classification can match the performance even of deconvolution or neural approaches suggests that future effort should be directed more towards improvement of grain statistics in conventional measure-ments, and in further development and investigation of imaging approaches. In these ways it can anticipated that the performance of standard luminescence methods for detecting dilute mixtures of irradiated and unirradiated food ingredients could be significantly improved. To do so would further enhance work conducted by FSA and other bodies to ensure that regula-tions governing the use of irradiation in food processing and the labelling of imported foods are followed.

Item Type:Research Reports or Papers (Project Report)
Status:Published
Glasgow Author(s) Enlighten ID:Scott, Professor E Marian and Murphy, Mr Simon and Carmichael, Dr Lorna and Sanderson, Professor David and Cresswell, Dr Alan
Authors: Sanderson, D.C.W., Carmichael, L., Murphy, S., Whitely, V., Scott, E.M., and Cresswell, A.
College/School:College of Science and Engineering > School of Mathematics and Statistics > Statistics
College of Science and Engineering > Scottish Universities Environmental Research Centre
Publisher:Food Standards Agency
Copyright Holders:Copyright © 2001 The Authors
Publisher Policy:Reproduced with the permission of the authors

University Staff: Request a correction | Enlighten Editors: Update this record