Hyperspectral imaging for presumptive identification of bacterial colonies on solid chromogenic culture media

Abstract

ABSTRACT BioMerieux is automating the microbiology lab oratory in order to reduce cost (less manpower and consumables), to improve performance (increased sensitivity, machine algorithms) and to gain traceability through optimization of the clinical laboratory workflow. In this study, we evaluat e the potential of Hyperspectral imaging (HSI) as a substitute to human visual observation when performing the task of microbiological culture interpretation. Microbial colonies from 19 strains subcategorized in 6 chromogenic classes were analyzed after a 24 h-growth on a chromog enic culture medi um (chromID® CPS E lite, bioMerieux, France ). The HSI analysis was p erformed in the VNIR region (400 -900 nm) using a linescan configuration. Using algorithms relyin g on Linear Spectral Unmixing , and using exclusively Diffuse Reflectance Spectra (DRS) as input data, we report interclass classification accuracies of 100% using a fully automatable approach and no use of morphological information. In order to eventually simplify the instrument, the performance of degraded DRS was also evaluated using only the most discriminant 14 spectral channels (a model for a multi spec tral approach) or 3 channels ( m odel of a RGB image). The overall classification performance remains unchanged for our multispectral model but is degraded f or the predicted RGB model, hints that a multispectral solution might bring the answer for an improved colony recognition. Keywords: Hyperspectral Imaging Spectroscopy, Clinical diagnostics, Colonies, Microbiology, Classification, Spectroscopy, Microbial species, Chromogenic media