Optical bacteria recognition: cross-polarized scattering

The rapid identification of bacteria is extremely important for controlling infec- tions and enabling swift and effective action. Light scattering has proven to be a highly versatile technique for identifying bacteria, as it does not require long colony growth times. In this article, we present a study on the use of cross-polarized optical scattering (CPS). Despite a relatively low scattering efficiency (10−5 to 10−6), working with cross-polarization enhances contrast by eliminating a highly intense background of scattered light. CPS has been applied to four bacteria, with three similar in shape. Moreover, two of them are Gram+ and two Gram-. The obtained images have been reduced in size down to a 16-bit images and camera noise has been added. Although bacteria are symmetrical in principle, in reality rotations of their orientation generate asymmetries in the CPS patterns that were exploited precisely to recognize and classify the different species. The classification of bacteria by a t-SNE algorithm in a reduced-dimension space shows that their features are grouped into specific clusters. However, such classification is not completely decisive due to partial cluster overlapping.