Technical Abstract: Direct plating onto solid agar media has been widely used in microbiology laboratories for presumptive-positive pathogen detection in spite of the fact that it is often subjective, labor intensive and time consuming. Rainbow agar is a selective and chromogenic medium that helps to detect pathogenic Escherichia coli (E. coli) strains. However, it is challenging to rapidly and accurately differentiate the well-known six pathogenic non-O157 Shiga-toxin producing E. coli (STEC) serogroups (O26, O45, O103, O111, O121, and O145) by human eyes due to the variability in STEC populations and the presence of background microflora. Therefore, there is a need for objectively, rapidly and accurately performing high-throughput screening of pathogen colonies on agar plates. In this paper, we report the development of a visible and near-infrared (VNIR) hyperspectral imaging technique and prediction model for objective, rapid and accurate differentiation of the six non-O157 STEC colony identities from pure cultures individually inoculated on Rainbow agar, which adopted supervised linear classification of factor sores obtained by principal component analysis (PCA). Both PCA-MD (Mahalanobis distance) and PCA-kNN (k-nearest neighbor) classifiers were used for model development. The number of spectral samples collected from semi-automatically defined 1,421 colony regions was 51,173, which were on 24 hyperspectral images from two replicates. Chemometric preprocessing methods and other operating parameters such as scatter correction, first derivative, moving average, sample size and number of principal components (PCs) were compared using a classification and regression tree (CART) method and followed by brute-force searching from candidates selected by the CART. The number of PCs, scatter correction and moving average were selected as the most important elements to consider before selecting a set of candidate models. Cross-validation (CV) such as hold-out and k-fold CV was used to validate the prediction performance of candidate models. Serogroups O111 and O121 showed consistently over 99% accuracy regardless of the classification algorithms. However, the classification accuracies of serogroups O26, O45, O103, and O145 showed varying results from 84% up to 100%, depending on which preprocessing treatment and prediction model were adopted. The best overall mean classification accuracy of 95.06% was achieved with PCA-kNN (k = 3), 6 PCs, 5 pixel sample size defined around each colony center, standard normal variate and detrending, first derivative with 11 point gaps and moving average with 11 point gaps. Future studies will focus on automating colony segmentation, further improving detection accuracy of the developed models, expanding the spectral library to include background microflora from ground beef and developing prediction models to detect the target bacteria in the presence of these background microorganisms.