Some strains of bacteria are known to fluoresce when exposed to UV-light, and are therefore detectable using spectroscopy techniques. If we could differentiate bacteria only based on an image, we open for a range of new methods within the field of microbial diagnostics. This is the underlying idea for our project. As we acquire hyperspectral images of a series of different bacteria, we aim to map their spectral “fingerprint” in a spatial manner. These first investigations make up a feasibility study which is continued this spring to allow for a more complete study.
The selected bacteria for imaging this fall wereStaphylococcus aureus, Staphylococcus epidermidis,
Streptococcus mutans and Streptococcus pneumoniae. We made use of a HySpex VNIR-1600 camera S/N: 0002 from Norsk Elektro Optikk (NEO), Skedsmokorset, held stationary above the bacterial samples which were placed on a translational stage, Standa Ltd. High Torque Stepping Motor FL42STH47-1204B-EH-7 10 cm, Vilnius, Lithuania. All samples were illuminated with a UV-LED 250 mW at 365 CLASS 3B, Hamamatsu C10559, Japan. This illumination was used to excite any autofluorescence present in the bacteria.
By making use of spectral analysis techniques such as Spectral Angle Mapping (SAM), Sequential Maximum Angle Convex Cone (SMACC) and k-means, fluorescence emission spectra from different bacteria were compared. Comparisons were made both for bacteria deposited on pore filter on agar, as well as bacteria seeded directly on TSB agar. For the bacteria deposited on filter paper, the supervised technique of SAM seemed to be most successful, but the opposite was true for the comparison of bacteria on agar, where the SMACC-extracted endmembers seemed to be able to differentiate the bacteria better.
As this is a phenomenological study, some preliminary findings were further investigated, such as
a two-peak emission in the red region detected from S. epidermidisin PBS solution, and mapping of a collected spectrum from S. mutans, a bacteria often associated with periodontitis, on a tooth diagnosed with a bacterial lesion. Hyperspectral imaging has so far proved to be a promising tool for characterisation of bacteria.

Vevang I.M¹, Bruzell E², Rukke HV² and Randeberg LL¹
1 Department of Electronics and Telecommunications NTNU, Trondheim
2 NIOM, Sognsveien 70, Oslo