“Framework for automated analysis of live cell migration behaviors“
Diego Calzolari – Grad Student, University of Modena and Reggio Emilia, Italy Sanford-Burnham Medical Research Institute, La Jolla, CA
Angiogenesis is the physiological process of new blood vessel formation. It is a normal and vital process in growth, development and represents an excellent therapeutic target for the treatment of cardiovascular disease. Development of new vessels in vivo and vessel-like structures in vitro involves massive cell movements and, therefore, the ability to effectively visualize and analyze large numbers of cells during their migration is crucial for a better understanding of the underlying developmental mechanisms and interplay between cell motility and morphogenesis. No golden standard exists to fully track live cells in an automated fashion. Different research groups and companies are developing various techniques and methods, none of which, however, are able to track cells in a native 3D environment at the same time taking into account the undergoing differentiation process. This suggests the need to develop a new software package able to track the cells during their migration/differentiation from single cells to mature 3D vessel-like structures.
We developed a novel framework to automatically track live cells expressing fluorescent proteins. The first step consists of a supervised Neural Network to pre filter and segment the images; then through a recursive method progressively attaching pixels to an object. Each object is evaluated on morphological and image related features. To link the objects in trajectories, each object detected at time t is scored against the object classified at time t-1. The score also has a penalty factor to increase the likelihood of choosing a closer object over a remote one as the next link in the trajectory. Each trajectory is then evaluated for its main physical features such as velocity, acceleration, directionality. The individual trajectory measurements are also compiled into group and correlated behaviors of a given cell population. Comparison performances with other software are shown.