ThinkCyte Publishes Paper on “Ghost Cytometry” Method in Science

Tokyo, Japan | June 15, 2018

In response to the growing demand for advanced single-cell identification and sorting system, ThinkCyte, Inc. has developed Ghost Cytometry™.

The system combines a novel imaging technique with machine learning to identify and sort cells with unprecedented high-throughput speed.

Until now, conventional methods to sort cells have been based on total fluorescence intensity. But with Ghost Cytometry single-cells are sorted based on images which means richer information content and more discerning algorithms. As Dr. Yoko Kawamura, ThinkCyte’s scientist says succinctly “A picture is worth a thousand numbers”.

This week ThinkCyte announces a high-profile publication in Science validating the use of ThinkCyte’s novel Ghost Cytometry method to image and sorting cells in experiments designed to mimic circulating tumor cells
in peripheral blood.

By integrating compressive imaging, flow cytometry, microfluidics, and machine learning algorithms, ThinkCyte has created a system that achieves the trifecta of single-cell analysis: high content, ultrahigh throughput, and selective, fluorescence image-based cell sorting in one instrument.

Conventional approaches to measure cells using flow cytometry and imaging systems have been unable to meet the simultaneous needs of high sensitivity, polychromaticity, high shutter speed, high frame rates, and continuous image-acquisition necessary to identify and sort cells in real time. Previous attempts to identify and sort single cells with high-throughput fluorescent image-based identification have been unsuccessful.

Data published in the journal Science show that the proprietary
Ghost Cytometry method successfully addresses the prior limitations of traditional flow cytometry and imaging systems.

In a series of experiments, lead authors Sadao Ota*, Yoko Kawamura,
Masashi Ugawa from ThinkCyte, Inc. describe Ghost Cytometry, a ultrafast fluorescence-imaging flow cytometry that can capture cell images and sort targeted cells based on the images. The “Ghost Cytometry” name refers to how the technique analyzes minimal light wave data without transforming any of that light data into a picture; it is image-free imaging technology.

The authors validated that Ghost Cytometry can capture fast, continuous, multi-color imaging of flowing cells by creating a flow-cell assembly and measuring double-stained MCF-7 cells (cells derived from human breast adenocarcinoma). The method generated multicolor fluorescent images with clearly resolved cellular morphological features even though the cells were flowing at over 10,000 cells per second a rate at which conventional imagers (CCD or CMOS-based cameras) only create unreadable motion-blurred images.

A critical component in correctly classifying a cell’s images at a such a throughput is the use of supervised machine learning tools. “Humans cannot classify objects based on waveforms, but machine learning methods can.” says Dr. Ota, “We apply machine learning algorithms directly to the temporal waveforms of light signals. This reduces the size of the imaging data and
we are skipping the time-consuming process of reconstructing each image from waveforms before identifying it. However, actual images of the cells can
still be produced for human validation.”


1) PROOF OF CONCEPT the study authors provide proof-of-concept for
Ghost Cytometry by training the machine-learning algorithm with a set of waveforms collected from two different but morphologically similar cell types (MCF-7 and MIA PaCa-2 cells) with fluorescently stained cytoplasm and/or membranes. They then analyzed samples containing mixtures of both cell types and found that waveform-based Ghost Cytometry can accurately classify these cells despite the fact that they are similar in size, total fluorescence intensity, and morphological features. Conventional flow cytometers using molecule-specific cell-labeling cannot distinguish between these cells as the total intensity of each cell type is similar.
2) REAL-LIFE CHALLENGE the authors then used the Ghost Cytometry system to identify rare cells, with an experiment to mimic circulating tumor cells in peripheral blood. They trained the algorithm with fluorescently stained PBMCs and MCF-7 cells and then spiked PBMCs with MCF-7 cells. The system was able to accurately detect MCF-7 cells inside a complex mixture of PBMCs from waveforms at ultrahigh throughput without using specific surface biomarkers. Then when authors added a microfluidic cell sorting system they were also able to sort the cells in real-time separating the cancerous cells with high accuracy and throughput without the use of specific biomarkers for cell identification.

Imaging flow analyzers are a valuable tool used to detect and analyze single cells in fields as diverse as oncology, immunology, and drug screening.
Ghost Cytometry is the first solution that enables the selective isolation of cells based on their morphology as captured in images (i.e. shape and size of nucleus, cytoplasm and membranes, protein co-localization) rather than labeled surface markers or clustered signal intensity.

As Dr. Ota explains, “This novel approach will enable the integration of
image-based analysis with comprehensive, downstream -omics analyses at the single cell level.”

The paper describing the experiments is entitled “Ghost Cytometry” and appeared in the journal Science. (Science 15 June 2018: Vol.360, Issue 6394, DOI: 10.1126 / science.aan0096)

* Sadao Ota is CTO & co-founder of ThinkCyte and Associate Professor of the University of Tokyo.


ThinkCyte, Inc. is a biotechnology company that enables innovative life science research, diagnostics and treatments, with integrated multidisciplinary technologies. ThinkCyte’s technologies combine advanced optics, machine learning and biotechnology to improve lives.

The company developed the Ghost Cytometry™ technology, the world’s first high-throughput fluorescent image-based cell sorting system. ThinkCyte was founded by leading experts in applied physics, optics, bioengineering and machine learning with a mission to advance single-cell analysis for research, diagnostics, and therapeutics with an integrated multidisciplinary approach.

ThinkCyte’s novel Ghost Cytometry method integrates advanced imaging technology, machine learning tools, microfluidics and flow cytometry.
Ghost Cytometry enables researchers to identify cells based on their images and isolate target cells, including rare cells in real time. This innovative approach brings ultrahigh throughput, content-rich analysis and sorting on a single cell level to disciplines such as drug discovery, cell therapy and clinical diagnostics. To learn more, please visit