Automated platform for high throughput, deep learning-based sorting of spherical 3D cell models for liver tissue engineering.

Not scheduled
ICE Krakow

ICE Krakow

ul. Marii Konopnickiej 17 30-302 Kraków


Sampaio da Silva, Claudia (CSEM / ETH Zurich)


"Liver disease is a major healthcare challenge, accounting for around 2 million deaths yearly worldwide. Although liver transplantation is the best way to re-establish normal liver function, less than 10% of transplantation needs are currently met. The EU Horizon 2020 OrganTrans consortium aims to address this issue by developing a liver tissue printing platform for transplantation. In this approach, spherical 3D liver models are used as building blocks for the bioprinted liver construct. One critical need is, therefore, a reliable source for large quantities of homogenous and healthy liver spheroids. Major limitations of current existing solutions are throughput efficiency and classification of complex objects where standard criteria, such as size and circularity, do not represent a sufficient measure of quality.
To answer this need we developed an automated deep learning-based sorting platform which allows for the screening of spheroids after formation. The platform includes imaging, classification, and individual sorting of spheroid. The imaging module consists of a brightfield microscope and is compatible with numerous microwell plates to image the spheroids directly in the culturing plate. The deep learning network is trained by experts and allows for the classification of healthy and unhealthy spheroids. A capillary needle enables removal of individual unhealthy spheroids, and the remaining healthy spheroids are harvested and concentrated in a Falcon tube. The platform is compatible with biosafety cabinets to ensure sterility. The system proved to successfully sorts 20 000 spheroids within 1.5 hours and with a classification accuracy of > 98% for both HepG2 monoculture spheroids and tri-culture liver spheroids.
The developed platform performs reliable sorting of complex 3D models, such as monoculture and co-culture spheroids, for which standard sorting criteria lack relevancy. This system represents a flexible solution for applications in regenerative medicine where large quantities of standardized spheroids are needed."

Presentation materials

There are no materials yet.