An Automated Cell Tracking Approach with Multi-Bernoulli Filtering and Ant Colony Labor Division

Benlian Xu,Jian Shi,Mingli Lu,Jinliang Cong,Ling Wang,Brett Nener

DOI: https://doi.org/10.1109/tcbb.2019.2954502

2021-01-01

IEEE/ACM Transactions on Computational Biology and Bioinformatics

Abstract:In this article, we take as inspiration the labor division into scouts and workers in an ant colony and propose a novel approach for automated cell tracking in the framework of multi-Bernoulli random finite sets. To approximate the Bernoulli parameter sets, we first define an existence probability of an ant colony as well as its discrete density distribution. During foraging, the behavior of scouts is modeled as a chaotic movement to produce a set of potential candidates. Afterwards, a group of workers, i.e., a worker ant colony, is recruited for each candidate, which then embark on gathering heuristic information in a self-organized way. Finally, the pheromone field is formed by the corresponding worker ant colony, from which the Bernoulli parameter is derived and the state of the cell is estimated accordingly to be associated with the existing tracks. Performance comparisons with other previous approaches are conducted on both simulated and real cell image sequences and show the superiority of this algorithm.