3-D Visual Feedback For Automated Sorting Of Cells With Ultra-Low Proportion Under Dark Field

Jieyu Tan,Huaping Wang,Qing Shi,Zhiqiang Zheng,Juan Cui,Tao Sun,Qiang Huang,Toshio Fukuda

DOI: https://doi.org/10.1109/ROMAN.2018.8525529

2018-01-01

Abstract:Study of cellular behaviors, especially the ultra-rare cell type, can aid in the accuracy of clinic diagnoses as well as the development of bioresearch engineering, thus the importance of isolating them from heterogeneous mixtures. However, current methods may fail in purity, versatility or cause contamination to cell targets, which is fatal drawback to rare cells. To address this issue, we propose a versatile method to automatically select and capture fluorescent stained target cells with high purity and recovery rate, through developing a novel 3D image processing algorithm under dark field. With the automated pick-and-place strategies, the micro-robotic system achieves cell screening even in an environment with ultra-sparse cells. In the proposed visual method, Markov Random Field (MRF) separation is adapted into the fluorescent environment to attain real-time planar location of micropipette and target cells. A reformative method derived from Depth from Defocus (DFD) is brought up to acquire 3D information. The basic system for this method mainly consists of a camera mounted on motorized fluorescent microscope and a micromanipulator for cell capture. The fluorescent label help to screen out most of the undesired cells while also bring extra constraints and requisition to our visual method. Finally, experiments of collecting 3T3 cells are performed to verify the feasibility and validity of the designed method, achieving average 98% purity and 80% recovery rate within the time limits. This study indicates that proposed visual processing method can not only provides reliable location feedback for micro-manipulation in rare cell sorting, but also can be easily extended to satisfy other automated micro-robotics manipulation.