2013 Annual Report
Once reasonable results are obtained for pure cultures, important food matrices will be spiked with pathogens and bio-threat agents at varying concentration levels and hyperspectral images will be collected on both the total contaminated food matrices and on a rinsate from the infected food matrices. Multivariate calibration models, based on the spectral libraries of the bio-threat agents and food matrices, will then be developed and the lower limit of detection for each pathogen and bio-threat agent will be determined.
In this project, five optical detection systems were evaluated for detection of foodborne pathogens including E. coli in ground beef, Salmonella in poultry carcass rinse.
They are: Visible Near Infrared (VNIR) Hyperspectral Imaging System – 400 to 1000 nm pushbroom hyperspectral imager Visible Near Infrared (VNIR) Hyperspectral Microscope Imaging System (HMI) – 400 to 850 nm acousto-optic tunable filter (AOTF) hyperspectral imager with Nixon light microscope Short-Wave Near Infrared (SWIR) Hyperspectral Imaging System – 900 to 1700 nm pushbroom hyperspectral imager Fourier-transform infrared (FTIR) micro-spectrometer – 4000 to 6000 reciprocal cm Surface Enhance Raman Spectroscopy (SERS) – 785 nm excitation confocal Raman Spectroscopy
Each system had both strengths and weaknesses when compared to the other systems. In the final analysis, no attempt was made to compare these systems to other non-optical detection methods. Therefore, all comparisons were described relative to these five systems, and their corresponding detection methods were presented with their relative strengths and weaknesses.
Early in the assessment of the systems, the Short-Wave Near Infrared (SWIR) hyperspectral imaging system was determined to be unsuitable for further evaluation as the high moisture content of the agar media produced a broad water absorption peak which effectively masked any pathogen related response. The VNIR and SWIR systems were easier to use (relatively) than the HMI, FTIR, and SERS systems. Likewise, sample preparation was easier and required no sample contact for the VNIR and SWIR systems while additional sample preparation and sample removal was required for the HMI, FTIR, and SERS systems. Thus, the VNIR and SWIR systems were also nondestructive. Although samples had to be removed for the HMI and the FTIR systems, the samples were still alive on the slide (HMI) and in the cell (FTIR). The cells could later be immersed in a broth and would grow. Even though the SERS method did not kill the bacteria, they were bound by the aptamer and thus, the SERS method was considered destructive. All systems required a significant amount of training and intimate knowledge to operate with the SWIR, FTIR, and SERS systems requiring more knowledge and training to use.
In our studies for specificity, the VNIR system was able to provide the highest classifications of bacteria, while the HMI, FTIR, and SERS systems provided good classification rates. As mentioned, the SWIR system was not suitable and had a very poor specificity. For sensitivity, both the FTIR and SERS systems had lower detection limits while the VNIR and HMI systems were slightly higher. Although capable of high signal response, the FTIR suffered from inconsistencies in sample preparation that could not be overcome and reduced the repeatability of its measurements. For imaging foodborne pathogens of moderate sample size, the VNIR and SWIR systems were faster in data acquisition than the others yet none of them were considered very fast.
In summary, comparing all the characteristics of the five systems described above, and from the data collected on the various systems within this study with the described methodologies, the VNIR system had the best feasibility for implementation and had the best commercial potential of the five systems. The VNIR system is rather expensive and requires significant knowledge to operate and process the data. A key element in using such a system is automation of all preprocessing steps. It can take a substantial amount of time to manually perform all the steps need for analysis. Therefore programs to automate the data analysis are critical.
Furthermore, it would be very advantageous if the method could be reduced to a simpler format. It is hypothesized that a color-balanced digital SLR camera might have potential for identifying the pathogenic bacteria from the non-pathogenic background microflora found in ground beef or poultry carcass rinse. The results for these preliminary tests were encouraging but would require improvement before a technology could be developed. Thus, significant additional research will be required to confirm or reject this hypothesis.