34201

DIFFERENTIATION OF FOODBORNE BACTERIA USING NIR HYPERSPECTRAL IMAGING

Terri-Lee Kammies 1 ; Marena Manley 1 ; Pieter Gouws 1 ; Paul Williams 2
Favorite this paper

Conventional microbiological and immunological methods for detection and differentiation of foodborne pathogenic bacteria are known to be time-consuming and labour intensive as identification of pathogenic strains could take several days, causing delay in product distribution. Near infrared (NIR) hyperspectral imaging provides a spatial map of spectral variation, visualised in NIR hyperspectral images. When applied to intact microbial cells, this technique provides a rapid method for detection and differentiation.
Four commonly encountered foodborne pathogens (Salmonella, Staphylococcus aureus, Bacillus cereus and Escherichia coli), as well as non-pathogenic S. epidermidis were used in this staudy. All organisms were imaged on Luria Bertani (LB) agar in glass petri dishes after 20, 40 and 60 h incubation at 37°C using the SisuChema short wave infrared (SWIR) camera, in the range 1000 to 2500 nm. The Evince v.2.7.0 hyperspectral imaging software package and MATLAB v 7.10 were used for all multivariate image analysis. Principal component analysis (PCA) was applied to mean-centered data calculating three principal components (PCs). Score plots and score images were used interactively for removal of all irrelevant pixels and to explore the data for any clustering. In addition wavelengths from 902-1097 nm and 2477-2514 nm were removed. Standard nominal variate (SNV) correction and the Savitzky-Golay was applied (2nd derivatives, 3rd order polynomial; 25 point smoothing) preprocessing to the final data set (1103-2471 nm).
Chemical differences between colonies which appeared similar in colour (B. cereus, E. coli and Salmonella) were evident in the PCA score plots. B. cereus was separated from Salmonella and E. coli along PC1 (58.1% Sum of squares (SS)) and was associated with peaks at 1480 (glucose), 1806 (cellulose) and 2130 (amino acid) nm, while E.coli and Salmonella (gram negative) were linked with peaks at 1693 (CH3) and 1974 (amino acid). It was possible to distinguish between E. coli and Salmonella in the direction of PC2 (7.75% SS). Salmonella was associated with 1530 (amino acid) and 2167 (amino acid) nm, while E. coli was associated with 1405 (ROH), 1687 (aromatic group) and 2316 (triglycerides) nm. Thus gram positive bacteria were separated from gram negative along PC1 and further separation between the gram negative bacteria was evident along PC2.
Gram positive bacteria (S. aureus, B. cereus and S. epidermidis) were also compared and clear differences were apparent. S. epidermidis was separated from B. cereus and S. aureus along PC1 (37.5% SS) and was attributed to variation in amino acid and cellulose content.
Further investigations were done on species differentiation of S. aureus and S. epidermidis. Two clusters were evident in the PCA score plot of PC1 vs. PC2 (78% SS), thus permitting distinction between species.
These results indicate that NIR hyperspectral imaging can be used to:
• distinguish between similarly coloured bacterial colonies;
• differentiate between gram positive and gram negative bacteria, and
• distinguish between pathogenic and non-pathogenic bacteria.