Figure 1. Concept of cell segmentation and tracking.

A. Top row: Artificial sequence that simulates six consecutive frames of a time-lapse video. The gray circles represent cells moving on a flat surface. Middle row: The goal of a segmentation algorithm is to accurately determine the regions of each individual cell in every frame, constructing a set of binary segmentation masks that correspond to the cells and locate them on a flat background. Bottom row: A tracking algorithm finds correspondences between the masks, i.e., the cells, in consecutive frames. If properly designed, a tracking algorithm is able to detect a moving cell (e.g., C1 or C3) while being within the field of view, determining when the cell enters and leaves the field of view. From the location of the cells in consecutive frames, it is possible to determine the trajectory of each cell and its velocity. A tracking algorithm should also be able to detect lineage changes due, for instance to a cell division event (e.g., cell C2 divides into two daughter cells, C2-1 and C2-2) or apoptosis. B. Graph-based representation of the cell tracks found by a tracking algorithm in the sequence shown at the top of panel A. Such an acyclic oriented graph contains, for each cell, the time when the cells enters and leaves the field of view, along with its division or apoptotic events. In a real case scenario, these graphs show the complete genealogy of the cells displayed in the frame of the video, all through the length of the video. Please note that the direction of the graph follows the temporal sequence starting at t=0 and moving toward t=5.