Three-Dimensional Autofocusing Visual Feedback for Automated Rare Cells Sorting in Fluorescence Microscopy.

Huaping Wang,Paolo Dario,Qing Shi,Qiang Huang,Kailun Bai,Juan Cui,Toshio Fukuda,Tao Sun

doi:10.3390/mi10090567

Abstract

Sorting rare cells from heterogeneous mixtures makes a significant contribution to biological research and medical treatment. However, the performances of traditional methods are limited due to the time-consuming preparation, poor purity, and recovery rate. In this paper, we proposed a cell screening method based on the automated microrobotic aspirate-and-place strategy under fluorescence microscopy. A fast autofocusing visual feedback (FAVF) method is introduced for precise and real-time three-dimensional (3D) location. In the context of this method, the scalable correlation coefficient (SCC) matching is presented for planar locating cells with regions of interest (ROI) created for autofocusing. When the overlap occurs, target cells are separated by a segmentation algorithm. To meet the shallow depth of field (DOF) limitation of the microscope, the improved multiple depth from defocus (MDFD) algorithm is used for depth detection, taking 850 ms a time with an accuracy rate of 96.79%. The neighborhood search based algorithm is applied for the tracking of the micropipette. Finally, experiments of screening NIH/3T3 (mouse embryonic fibroblast) cells verifies the feasibility and validity of this method with an average speed of 5 cells/min, 95% purity, and 80% recovery rate. Moreover, such versatile functions as cell counting and injection, for example, could be achieved by this expandable system.

Highlights

We proposed a cell screening method based on the automated microrobotic aspirate-and-place strategy under fluorescence microscopy
In the context of this method, the scalable correlation coefficient (SCC) matching is presented for planar locating cells with regions of interest (ROI) created for autofocusing
To meet the shallow depth of field (DOF) limitation of the microscope, the improved multiple depth from defocus (MDFD) algorithm is used for depth detection, taking 850 ms a time with an accuracy rate of 96.79%

Summary

Introduction

Low-abundance cells with samples containing less than 1000 target cells/ml are considered as rare cells [1]. Research on sorting and isolating a specified number of rare cells, such as circulating tumor cells (CTCs), circulating fetal cells, and stem cells from complex and heterogeneous mixtures, are vitally important in biology and medicine [2]. FACS is limited for clinical application and commercialization due to the high initial costs of system, the risk of sample contamination, and the necessity of technical expertise for operating complex machinery [10] Another commercially available sorter, the immunomagnetic-assisted cell sorting (IMACS) system, was invented later. The fast visual processing algorithm for rare cell fluorescence observation is lacking. We propose a fast autofocusing visual feedback (FAVF) method for automated sorting of rare cells under fluorescence microscope. The FAVF algorithm mainly consists of four parts: preprocessing, planar locating, depth detection, and object tracking.

Materials and Methods

Neighborhood Searching Based Micropipette Tracking

System Framework