The application of evolutionary computation towards the characterization and classification of urothelium cell cultures

This thesis presents a novel method for classifying and
characterizing urothelial cell cultures. A system of cell
tracking employing computer vision techniques was applied
to a one day long time-lapse videos of replicate normal human uroepithelial cell cultures exposed to different concentrations of adenosine triphosphate (ATP) and a selective purinergic P2X antagonist (PPADS) as inhibitor. Subsequent analysis following feature extraction on both cell culture and single-cell demonstrated the ability of the approach to successfully classify the modulated classes of cells using evolutionary algorithms. Specifically, a Cartesian Genetic Program (CGP) network was evolved that identified average migration speed, in-contact angular velocity, cohesivity and average cell clump size as the principal features contributing to the cell class separation. This approach provides a non-biased insight into modulated cell class behaviours.