Share Email Print

Proceedings Paper

Hyperspectral imaging applied to microbial categorization in an automated microbiology workflow
Author(s): Denis F. Leroux; Rony Midahuen; Guillaume Perrin; Jeremie Pescatore; Pierre Imbaud
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Hyperspectral imaging (HSI) is being evaluated as a pre-selection tool to categorize and localize populations of microbial colonies directly onto their culture medium, in order to facilitate the microbiology workflow downstream the incubation step. The categorization criteria were here limited to the diffuse radiance spectra acquired mostly in the visible region between 400 and 900 nm.

Although the diffuse radiance signal is much broader than the one acquired using vibrational techniques such as Raman and IR and limited to chromophores absorbing in the visible region, it can be acquired very quickly allowing to perform hyperspectral imaging of large objects (i.e. Petri dishes) with throughputs that are compatible with the needs of a clinical laboratory workflow. Moreover, additional cost reduction could possibly be achieved using application-specific multispectral systems. Furthermore, recent research has shown that good power of discrimination, at the species level, could be achieved at least for a low level of species.

In our work, we test different culture media, with and without a strong light absorption in the visible region, and report categorization results obtained when selecting end-member spectra according to a multi-parametric study (colonies, agar type). Results of categorization (e.g. at the species level) are presented using two types of supervised-categorization algorithms providing that they deliver subpixel fractional abundance information (Linear Spectral Unmixing type) or not such as Spectral Angle Mapping (SAM) and Euclidian Distance (ED) type. Interestingly the performance between the two classes of algorithms is dramatically different, a trend which is not always observed. An interpretation is proposed on the basis of the agar interference and the spectral purity of end-member spectra.

Paper Details

Date Published: 15 July 2015
PDF: 9 pages
Proc. SPIE 9537, Clinical and Biomedical Spectroscopy and Imaging IV, 953726 (15 July 2015); doi: 10.1117/12.2184140
Show Author Affiliations
Denis F. Leroux, bioMérieux SA (France)
Rony Midahuen, bioMérieux SA (France)
Guillaume Perrin, bioMérieux SA (France)
Jeremie Pescatore, bioMérieux SA (France)
Pierre Imbaud, bioMérieux SA (France)

Published in SPIE Proceedings Vol. 9537:
Clinical and Biomedical Spectroscopy and Imaging IV
J. Quincy Brown; Volker Deckert, Editor(s)

© SPIE. Terms of Use
Back to Top