Microbial identification through multispectral infrared imaging of colonies: combining chemical and morphotype analysis

Abstract

Abstract Most of optical methods for the identification of microorganisms rely either on vibrational spectroscopies, such as Raman and infrared spectroscopy, or on morphotype imaging within the visible range. With the most recent developments of quantum cascade lasers, a fast imaging system able to collect both of these information, vibrational (composition-based) and morphotype, is now within reach. However, the potential of such system to be more precise than conventional imaging or spectroscopy has yet to be demonstrated. Here, we report a practical protocol for microbial colonies identification by means of QCL-based multispectral infrared imaging, as well as the acquisition of a first database for performance evaluation. To this end, optical fingerprints of 2253 colonies belonging to 8 microbial species were recorded at 9 different wavelengths in the mid-infrared range. Image descriptors and machine learning approaches allowed for classification with a 94.4% correct identification rate. Strain on three strains of S. epidermidis appeared feasible and as effective as species-level classification. Therefore, multispectral imaging in the mid-infrared has the potential to offer a powerful label-free and non-destructive technology for microbial identification, and could be used for strain typing.