Abstract

Using Video Bioinformatics Quality Control Software to Evaluate the Health of Stem Cell Populations

Atena Zahedi, Researcher, University of California-Riverside

Human embryonic stem cells (hESC) have an immense potential for use in basic research and clinical applications such as disease modeling or cell replacement therapies. In clinical applications, it is crucial to consider the state of hESC health throughout passaging, expansion, and culture and to apply quality control methods to stem cells produced for patient transfer. We are developing a suite of software tools that correlate time-lapse phase-contrast video analysis of dynamic hESC behavior with predictions on their health over time. We envision this software, StemCellQC, being used as a quality control technology in clinics performing stem cell therapy. To evaluate morphological changes in growing colonies, we devised 22 user-defined features to analyze a set of hESC videos. Predictions were made using five computationally extracted features and compared against visual observations made by experts experienced in working with hESC. The automated video analysis algorithm predicted the health of hESC colonies with an 85% accuracy by non-invasively measuring and tracking five morphological parameters over a period of 48 hours. Morphological features such as area and number of protrusions are good indicators of colony growth. The bright area ratio is useful for monitoring the percentage of dead cells. Moreover, the change in centroid feature can track the movement of colonies and provide additional information on cytoskeletal restructuring. Lastly, solidity can serve as an early predictor of dissociating, dying cells. A combination of multiple features increases the chance of early detection of adverse effects during the culturing period. This protocol represents a novel resource-saving quality control method for determining hESC health. It can significantly reduce the time and money required to manually track large numbers of hESC colonies and  eliminate false classification due to human bias.