Purpose: To investigate the ability of hyperspectral imaging to detect STEC serogroups (O26, O45, O103, O111, O121, and O145) artificially inoculated in ground beef broth from background microflora plated on Rainbow agar.
Methods: Ground beef (65 + 2 g) was enriched in 585 +15 ml of modified tryptic soy broth (m-TSB, 20 mg / ml novobiocin) overnight at 42˚C. Four ten-fold dilutions of the enriched ground beef sample were prepared in sterile saline. Then for each serogroup, 1000 CFU (10 ml of a 105 CFU/ml cell suspension) of STEC was spiked into 990 ml of each of the enriched ground beef sample dilutions. The STEC spiked ground beef sample dilutions were spread onto Rainbow agar plates. The Themis Vision Systems’ hyperspectral imaging system was used to acquire images of agar plates. Regions of interest associated with STEC target colonies and competing background microflora colonies were created for validation using a previously developed Mahalanobis distance classifier.
Results: Principal component (PC) score plots revealed separability of the target STEC colonies from each other and the background microflora colonies with the exception of O111. Scores of the competing background microflora were mixed into target O111 PC space. As a result, 33% of the background flora was misclassified as serogroup O111. No other serogroups were misclassified as background microflora. A new Mahalanobis distance classifier was developed to differentiate only O111 from the background microflora which reduced the false positive rate to 5.3%. However, the classifier had a false negative rate of 13% for O111.
Significance: Hyperspectral imaging can improve the speed and accuracy of selecting well isolated STEC colonies from enriched samples plated on Rainbow agar for confirmation.