Abstract:Flow cytometry is a well-established technique that is widely applied in numerous fields, including pathology, pharmacology, immunology, marine biology, plant biology, and molecular biology. Conventional methods for flow cytometry fail to accurately detect cellular phenotypic characteristics due to a limited number of variants and the lack of spatial metrics. Imaging-based cell analysis methods, such as high-content screening and imaging flow cytometry, are advantageous over those univariate or few-variate methods because they offer the capability of acquiring multi-dimensional information of single cells, from which cellular characteristics can be detected with high accuracy and high specificity. However, currently available imaging flow cytometry methods suffer from low throughput which is mainly limited by the imaging techniques, or specifically, the frame rate of the commercial imaging sensors, such as CCD or CMOS sensors. In order to address these problems, here we present optofluidic time-stretch microscopy with extremely high throughput which is capable of acquiring bright-field images of large populations of cells with a high spatial resolution of 780 nm and a high throughput of >1 million cells/s, which is two orders of magnitude higher than conventional univariate or few-variate flow cytometry methods and three orders of magnitude higher than other imaging flow cytometry methods. This is made possible by integrating an optical time-stretch microscope with a hydrodynamic-focusing microfluidic device. In addition, we apply machine-learning algorithms to the acquired images to extract multiple morphological features from each cellular image to identify and classify the cells in a label-free manner with accuracy higher than 90%, which is comparable with the fluorescence-based methods. Specifically, we experimentally performed optofluidic time-stretch microscopy to detect K562 cells (leukemia cell line) spiked in whole blood samples which were treated with different concentrations of anti-cancer drugs. In the experiment, more than 10,000 high-quality images of K652 cells were acquired for each concentration of the drug. With machine-learning-based image processing and analysis, 548 morphological features were extracted from each image to comprehensively evaluate its cellular phenotypes and hence dose-dependent morphological changes of the cells caused by the drug treatment. We further confirmed the dose-dependent results in different experimental trials where the cells were treated with the drugs for different time spans. This is potentially applicable for research of cellular drug responses directly with whole blood, hence, beneficial to drug discovery and drug screening. With such high throughput, high performance and good compatibility with existing techniques, we believe that optofluidic time-stretch microscopy with extreme throughput will revolutionize the flow cytometry field and play an integral role in high-throughput, high-accuracy, and label-free cell screening in the future.

Label-free Detection of Cellular Drug Responses by High-Throughput Bright-Field Imaging and Machine Learning

Label-free cell cycle analysis for high-throughput imaging flow cytometry

Analysis of cellular response to drugs with a microfluidic single-cell platform based on hyperspectral imaging

Label-free Image-Based Detection of Drug Resistance with Optofluidic Time-Stretch Microscopy (conference Presentation)

Rapid and Accurate Identification of Cell Phenotypes of Different Drug Mechanisms by Using Single-Cell Fluorescence Images Via Deep Learning

Optofluidic time-stretch microscopy with an extreme throughput of 1 million cells/s (Conference Presentation)

Deep Learning in Label-free Cell Classification

A linear classification approach to model the early cellular responses induced by drugs

High-Throughput, Label-Free, Multivariate Cell Analysis With Optofluidic Time-Stretch Microscopy

Label-Free Assessment of the Drug Resistance of Epithelial Ovarian Cancer Cells in a Microfluidic Holographic Flow Cytometer Boosted through Machine Learning

Machine-learning-guided quantitative delineation of cell morphological features and responses to nanomaterials

An Automatic Method for Robust and Fast Cell Detection in Bright Field Images from High-Throughput Microscopy.

Label-free prediction of cell painting from brightfield images

Label-Free Quantification of Molecular Interaction in Live Red Blood Cells by Tracking Nanometer Scale Membrane Fluctuations.

High-Throughput Screening Based on Label-Free Detection of Smallmolecule Microarrays

Drug effects analysis on cells using a high throughput microfluidic chip

Computational Imaging Of Label-Free Cells Using Lens-Less Digital Holography

Label-free chemical imaging flow cytometry by high-speed multicolor stimulated Raman scattering

Method of Tracking and Analysis of Fluorescent-Labeled Cells Using Automatic Thresholding and Labeling

Label-Free Chemical Nano-Imaging of Intracellular Drug Binding Sites

Computer vision meets microfluidics: a label-free method for high-throughput cell analysis