Characterizing cell viability is critical for many life science research and development studies. Viability is typically assessed using fluorescent markers such as propidium iodine (PI) or 7-AAD, that exploit the loss of membrane integrity in dead cells which allows these markers to permeate into nuclei, bind double stranded DNA, and thus label cells in the late stages of death. In earlier stages of apoptosis where membrane integrity has not yet been lost, other fluorescent markers such as Annexin V, Apotracker, JC-1, Caspase-3/7 are commonly used to label
activation of the apoptotic pathway. However, the use of fluorescent labels to identify cell viability states can hinder downstream research efforts and therefore label-free approaches for characterizing cell viability is a major unmet research need. Here we used VisionSort to acquire morphological profiles for three cell viability phenotypes (live, dead, and apoptotic). These profiles were used to build AI-based classifiers via supervised machine learning in VisionSort and then used to identify cell viability phenotypes without the need for conventional fluorescent markers.