Light scattering imaging combined with machine learning for label-free identification of live yeast cells

Light scattering techniques are highly sensitive to structural changes in live and fixed cells, enabling label-free singlecell analysis compared to traditional fluorescence techniques. Solving the inverse problem of light scattering is interesting, as it provides the physical information from label-free single-cells. In this paper, a 532 nm laser is focused to excite a single cell to acquire light scattering images. A 10X objective lens and a CMOS camera are used to detect the elastic scattering perpendicular to the laser excitation direction. Yeast cells are cultured in a liquid containing 2% glucose and incubated at room temperature. Cell samples were obtained on day 1 and day 4, and their light scattering images were acquired, respectively. By solving the inverse light scattering problem, we found that the refractive index of yeast cells increased on day 4 of incubation compared with day 1. By combining machine learning technology with light scattering imaging, the average classification accuracy of label-free yeast cells on day 1 and day 4 is 98.3%. These findings prove the high sensitivity of light scattering imaging to the physical variations in single cells and speculate the possible applications of this technology in label-free live cell analysis.