ThinkCyte Inc. (Tokyo, Japan) launched to develop an artificial intelligence-enabled, image-free visual cell sorting system that can sort cells by morphology at higher speeds and with greater accuracy than current methods. The company believes its technology has potential research, clinical and therapeutic applications, including identifying circulating cancer cells in peripheral blood and ramping up the discovery and quality of cell-based therapies.
Current cell sorting methods rely on microscope images of cells classified using a computer image recognition program, which cannot accurately distinguish between cells with similar sizes and structures, or human observation, which cannot perform high-throughput sorting.
ThinkCyte's technology analyzes light wave data without transforming them into a picture, allowing more rapid processing via machine learning. ThinkCyte co-founders Sadao Ota and Issei Sato and researchers from Japanese universities described the technology, dubbed Ghost Cytometry, in a paper published in Science.
The researchers used a microfluidic system to direct a stream of fluorescently labeled cells one at a time through a static optical structure with randomly patterned spots that excite the cell's fluorophores as it passes by. A single-pixel detector camera continuously records the emitted fluorescent light wave intensities, which are compressed as a single temporal waveform. An electrical circuit attached to the camera uses machine learning algorithms to learn a cell's type based on its unique light wave pattern to identify the cell within 10 microseconds and emit an electrical current to push it into the correct sorting pathway at a throughput of about 3,000 cells per second. The compressive sensing allows high-throughput classification of a cell's morphology because it reduces the size of imaging data while retaining enough information for image reconstruction.
The researchers applied the technique to accurately isolate pancreatic cancer cells from breast cancer cells with similar sizes, fluorescent intensities and apparent morphologies, and from a mixture of peripheral blood mononuclear cells.
ThinkCyte plans to collaborate with research institutes this year on cancer and regenerative medicine clinical projects using Ghost Cytometry. It also plans to commercialize a beta prototype of the equipment for research use in 2019.
ThinkCyte raised $3.2 million in a seed round Thursday from new investors Real Tech Fund, Japan Science & Technology Agency (JST) and Osaka University Venture Capital. The company has raised over $5 million to date, including angel investments and grants.
Oto is ThinkCyte’s chief technology officer and an associate professor at University of Tokyo. Sato is a scientific adviser to the company and a computer science lecturer at the same university. ThinkCyte CEO Waichiro Katsuda is also a co-founder of the company.