An Ant Colony Inspired Multi-Bernoulli Filter for Cell Tracking in Time-Lapse Microscopy Sequences

Benlian Xu, Mingli Lu, Jinliang Cong, Brett D. Nener

Research output: Contribution to journalArticle

Abstract

The analysis of the dynamic behavior of cells in time-lapse microscopy sequences requires the development of reliable and automatic tracking methods capable of estimating individual cell states and delineating the lineage trees corresponding to the tracks. In this paper, we propose a novel approach, i.e., an ant colony inspired multi-Bernoulli filter, to handle the tracking of a collection of cells within which mitosis, morphological change and erratic dynamics occur. The proposed technique treats each ant colony as an independent one in an ant society, and the existence probability of an ant colony and its density distribution approximation are derived from the individual pheromone field and the corresponding heuristic information for the approximation to the multi-Bernoulli parameters. To effectively guide ant foraging between consecutive frames, a dual prediction mechanism is proposed for the ant colony and its pheromone field. The algorithm performance is tested on challenging datasets with varying population density, frequent cell mitosis and uneven motion over time, demonstrating that the algorithm outperforms recently reported approaches.

Original languageEnglish
Article number8884723
Pages (from-to)1703-1716
Number of pages14
JournalIEEE Journal of Biomedical and Health Informatics
Volume24
Issue number6
DOIs
Publication statusPublished - Jun 2020

Fingerprint Dive into the research topics of 'An Ant Colony Inspired Multi-Bernoulli Filter for Cell Tracking in Time-Lapse Microscopy Sequences'. Together they form a unique fingerprint.

  • Cite this