Back to Search
Start Over
Looking for Aflatoxin B contamination with a low cost optical apparatus and machine learning approach
- Source :
- 9th International Symposium on Recent Advances in Food Analysis (RAFA 2019), Prague, Czech Republic, 5--8 November 2019, info:cnr-pdr/source/autori:Francesca Romana Bertani, Annamaria Gerardino, Luca Businaro, Eugenio Martinelli, Arianna Mencattini, Davide Di Giuseppe, Michele Solfrizzo, Lucia Gambacorta/congresso_nome:9th International Symposium on Recent Advances in Food Analysis (RAFA 2019)/congresso_luogo:Prague, Czech Republic/congresso_data:5--8 November 2019/anno:2019/pagina_da:/pagina_a:/intervallo_pagine
- Publication Year :
- 2019
-
Abstract
- Aflatoxin detection currently relies mainly on chemical methods usually based on chromatography approaches, and recently developed immunochemical based assays that are fairly accurate, however, they are time-consuming, expensive and destructive. Non-destructive, optical approaches are recently being developed in order to assess the presence of contamination in a cost and time-effective way, maintaining high levels of accuracy and reproducibility, but are usually based on the benchtop and expensive instruments. Here we will present the evolution of results of the analysis of fluorescence spectra of contaminated almond samples during the development of an optical multi-sensor device in the framework of PhasmaFOOD project. The aim of the project was to develop a low cost and portable instrument comprising multispectral and imaging capabilities, conjugated with a cloud reference database and analysis toolbox for food features analysis. One of the use cases of the project is the fast, reliable and non-destructive detection of mycotoxin (in particular aflatoxin) contamination in food products. For this use case, we used in particular fluorescence spectroscopy and different approaches to data analysis. After the first feasibility tests in the range of mg/g contamination range with a simple and effective analysis that led to highly reliable results, the work was focused on the detection limits with samples (almond) in the range of 0-291 ng/g acquired with a simple portable device and excitation light at 365 nm wavelength. An ad hoc processing strategy based on a feature selection steps coupled with a nonlinear classifier has been developed and test with two different datasets collected one month from the other another. The system performances have been evaluated training the classification model with one dataset and testing its with the other. The results have shown an accuracy higher than 80% with a threshold lower than 10ppb as contamination level.
- Subjects :
- none
Subjects
Details
- Language :
- English
- Database :
- OpenAIRE
- Journal :
- 9th International Symposium on Recent Advances in Food Analysis (RAFA 2019), Prague, Czech Republic, 5--8 November 2019, info:cnr-pdr/source/autori:Francesca Romana Bertani, Annamaria Gerardino, Luca Businaro, Eugenio Martinelli, Arianna Mencattini, Davide Di Giuseppe, Michele Solfrizzo, Lucia Gambacorta/congresso_nome:9th International Symposium on Recent Advances in Food Analysis (RAFA 2019)/congresso_luogo:Prague, Czech Republic/congresso_data:5--8 November 2019/anno:2019/pagina_da:/pagina_a:/intervallo_pagine
- Accession number :
- edsair.cnr...........b5ba5c1b2eee4b98e575f54ec9b167f4