AI-Driven, Label-Free Quantification of Cell Viability in Live-Cell Imaging

Drug discovery is moving towards the use of more relevant and more precious cell types, to improve the translational relevance of in vitro assays. These cell types often have increased sensitivity to external perturbance including the presence of fluorescent reagents and reporter dyes, thus there is a vital need to improve our ability to analyze complex images of cells without reliance on fluorescence readouts. The use of artificial intelligence for image analysis has also expanded hugely in recent years, and our new software module applies this technology to live-cell analysis.

The Incucyte® Live-Cell Analysis system uses bespoke, integrated workflows to analyze images of cells from within a cell culture incubator, providing a physiologically relevant, non-perturbing environment for long-term live-cell assays. The new Incucyte® AI Cell Health Analysis module uses neural networks, which have been trained to recognize diverse cell types, to perform automated cell segmentation and live/dead classification. 

Cell segmentation is robust, accurate, and highly adaptable to a wide variety of live and dead adherent and non-adherent cells. The live/dead classification provides label-free information on cell viability, producing comparable quantification to standard fluorescence methods. In addition to this information, fluorescence analysis can optionally be performed using the total cell population or within only the live or dead subpopulations.

Here we describe the development and validation of the Incucyte® AI Cell Health Analysis module and demonstrate its use in label-free cytotoxicity screening assays. We also investigate the detailed mechanistic insight which can be gained by combining this advanced label-free analysis with fluorescent reporters.