Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: Optical Detection of Food Safety and Food Defense Hazards

Location: Quality and Safety Assessment Research Unit

Title: The potential for early and rapid pathogen detection within poultry processing through hyperspectral microscopy

Authors
item Eady, Matthew
item Park, Bosoon

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: November 5, 2013
Publication Date: January 28, 2014
Citation: Eady, M.B., Park, B. 2014. The potential for early and rapid pathogen detection within poultry processing through hyperspectral microscopy. Meeting Abstract [abstract].

Technical Abstract: The acquisition of hyperspectral microscopic images containing both spatial and spectral data has shown potential for the early and rapid optical classification of foodborne pathogens. A hyperspectral microscope with a metal halide light source and acousto-optical tunable filter (AOTF) collects 89 contiguous wavelength measurements every 4 nm between 450 and 800 nm. Results have found both bacterial species and serotypes have a unique spectral signature that can be detected with a hyperspectral microscope to differentiate between various samples. Image data collected from samples incubated at times of 6, 8, 10, 12, 13, 15, and 24 hours were analyzed for early detection. The rapid methodology presented here focuses on a glass slide preparation that takes approximately 10-15 minutes with a scanning time of approximately 45 seconds without binning. Samples demonstrated high spectral correlation values, above 0.9. Chemometric data analysis was applied to differentiate the spectral variance. Data were collected from each image based on a Region of Interest (ROI) consisting of a randomized 30% pixel sample of the average cell size. A Savitzky-Golay derivative based global pretreatment algorithm was applied to all of the ROI spectral data, and followed by a principle component analysis (PCA). Model calibration was performed with partial least squares regression (PLSR) on data training sets, while soft independent class analogy (SIMCA) was implemented to predict sample clustering. A minimalist model analysis was developed for practical application of this spectral classification method, in order to reduce the amount of data processing and storage space required for continuous use in a poultry processing facility. Reducing the number of necessary wavelengths from 89 to fewer than 12 will retain cluster separation.

Last Modified: 10/1/2014
Footer Content Back to Top of Page