Light-sheet flow cytometry for label-free classification of acute and chronic myeloid leukemic cells with machine learning

Abstract

Conventional flow cytometry has been used for leukemia characterization via fluorescence measurements. Here we measured the 2D light scattering patterns of label-free HL-60 cells (human acute myeloid leukemic cells) and K562 cells (human chronic myeloid leukemic cells) with a light-sheet illuminated flow cytometer. Approximately 70 light scattering patterns of leukemia cells were obtained in a one minute video taken by this cytometer operating at 50 frames per second. Local binary pattern (LBP) was used to extract features of the 2D light scattering images, which were then analyzed by the support vector machine (SVM) algorithm. An accuracy rate of 98.23% was obtained for the label-free classification of these two kinds of leukemia cells, with a specificity of 99.28% and a sensitivity of 97.22%. The combination of light-sheet flow cytometry with machine learning may be helpful for leukemia subtyping diagnosis in clinics.