Defocusing effects on label-free cervical cancer cell classification by two-dimensional light scattering static cytometry and machine learning

Cervical carcinoma is one of the most common gynecological malignancies in the world. Here we have measured the 2D light scattering patterns of two representative types of cervical cancer cell lineage cells (HeLa, H8) at six different defocusing distances. The light scattering patterns vary at different defocusing distances, where the longer the defocusing distance, the larger the pattern area is. The classification performance for cervical cancer cells at different defocusing distances is evaluated based on support vector machine (SVM) classification algorithm. Speckle features are extracted by histogram of oriented gradient (HOG). Under six defocusing distances, the difference between the highest and lowest accuracy is 5.09%. The study of defocusing effects on cell classification with 2D light scattering static cytometry may help for the development of high speed and high performance imaging flow cytometry, and the combination of flow cytometry and machine learning holds great promise for automating the early clinical diagnosis of cervical cancer and other diseases.