The rapid identification of bacterial species is crucial in biomedical applications such as a cure for sepsis. However, conventional techniques based on culturing and successive biochemical treatments are too time consuming. Thus, studies during the last few decades have adopted approaches such as direct sequencing or optical scattering for rapid bacterial identification. However, most of these schemes have technique limitations; they are not fast enough or require intensive labeling.

Our findings demonstrate single-shot measurements of holographic light scattering from individual bacteria, by which species of bacteria can be identified without using labeling agents or additional sample treatments. A research team led by Prof. Yong-Keun Park in the Department of Physics and KI Center for Optics for Health Science at KAIST developed a new technology by combining optical holography and machine-learning techniques.

A custom-based digital holographic microscopy was employed to measure the two-dimensional optical-field map scattered from a sample. The optical field contains both intensity and phase information, whereas conventional bright-field microscopy only addresses the intensity information. The measurements of optical-field information enable one to access the light-scattering information with unprecedented precision and sensitivity. Employing these digital holographic microscopy techniques, 2-D light scattering patterns from individual bacteria have been measured for the first time.

The researchers also established a method to maximally extract the species-specific light-scattering information in the measured light-scattering patterns to be exploited for single-bacterial identification via machine-learning algorithms. Measured optical-field maps were first decomposed into fundamental basis patterns for systematic recognition and then analyzed employing the machine-learning algorithms. Although the morphology of different species of rod-shaped bacteria look similar in conventional optical micrographs, the measured 2-D light scattering patterns from individual bacterial contain subtle information corresponding to the distribution and molecular composition of subcellular structures in each species of bacteria. To identify bacterial species based on the 2-D light-scattering patterns, we performed the supervised machine-learning algorithm in order to decompose the light-scattering patterns into basis patterns for effective discrimination. The results show the accuracy (both the sensitivity and specificity) of identification is higher than 95%.

The team expects the new method to enable ultrafast bacterial-species identification, reducing the time cost from days to seconds when combined with high-speed flow cytometry. Utilizing other optical modalities, such as spectral or polarimetric measurements, might further enhance accuracy and generalize the technique for the identification of more general pathogens.

The first part this study on the holographic measurements of light scattering from individual bacteria was published in Scientific Reports in May 2014. The second part on bacterial identification based on light-scattering measurements is now under review for publication.

This work was supported by the Korean Ministry of Education, Science and Technology (MEST), and the National Research Foundation (2012R1A1A1009082, 2013K1A3A1A09076135, 2013M3C1A3063046, 2009-0087691, 2012-M3C1A1-048860, 2013R1A1A3011886), the KAIST Undergraduate Research Participation program, the KAIST Presidential Fellowship, the SPIE Optics & Photonics Education Scholarship, and the Posco ChungAm Fellowship.