A generalized labeled multi-Bernoulli tracker for time lapse cell migration

Abstract

Tracking is a means to accomplish the more fundamental task of extracting relevant information about cell behavior from time-lapse microscopy data. Hence, characterizing uncertainty or confidence in the information inferred from the data is as important as the tracking of the cells. In this paper, we show that in addition to being a principled Bayesian multi-object tracking approach, the Random Finite Set (RFS) framework is capable of providing consistent characterization of uncertainty for the information inferred from the data. In particular, we use an efficient implementation of the Generalized Labeled Multi-Bernoulli (GLMB) filter to track a large number of cells in a cell migration experiment and demonstrate how to characterize uncertainty on variables inferred from the data such as cell counts, survival rate, birth rate, mean position, mean velocity using standard constructs from RFS theory.