Cancer is a worldwide disease and, in the UK, breast cancer is the most common. Compared to healthy cells, cancer cells migrate abnormally, associated with alterations in cell motility and morphology. The development of biomedical imaging techniques result in the production of large amounts of data. The analysis of such large data, the variety of cancer cell shapes and the potential links between cell motility and morphology present a challenge for cell migration study: how to analyse cell motility and morphology simultaneously.
This thesis proposes a computational framework to address integrated cancer cell migration analysis. Firstly, automated tracking of cell boundaries is undertaken by a DWNA kinematic model of cell boundaries, described by B-spline active contours. The tracked cell states intrinsically links cell morphology to motility features. As a result, cell centroid and boundary dynamics are successfully tracked, followed by quantitative motility analysis.
A module to quantitatively analyse cell morphology is proposed after tracking. Cell shapes are described by a 2D descriptor. Accordingly, cell morphodynamics are modelled as a hidden Markov process, along with three shape states: round, elongated and teardrop. In order to explore the potential interactions between cell shapes and motility, cell centroid motility characteristics are associated to the identified shape states. When the analysis was applied to breast cancer control cells, the identified shape states showed distinct motility characteristics.
Finally, the proposed framework is adapted to the comparison of MDA-MB-231 cell behaviours with regulating migration-associated proteins: i) Blebbistatin and Y-27632, which are chemical inhibitors of two different proteins working on the same pathway, showed identical, but different degrees of effects on the motility and morphology characteristics of MDA-MB-231 cells. ii) The absence of FA-associated genes, including FAK, RhoE and beta-PIX, respectively showed distinct effects on cell migrations.
