Automated lineage tree reconstruction from Caenorhabditis elegans image data using particle filtering based cell tracking

Caenorhabditis elegans is an important model organism for the study of molecular mechanisms of development and disease processes, due to its well-known genome and invariant cell lineage tree. Such studies generally produce vast amounts of image data, and require very robust and efficient algorithms to extract and characterize lineage phenotypes and to determine gene expression patterns. Previously published methods for this purpose show only mediocre performance and often require extensive manual post-processing. The challenge remains to develop more powerful and fully automated methods. In this paper we propose a new algorithm for C. elegans cell tracking and lineage reconstruction, based on a Bayesian estimation framework, implemented by means of particle filtering. The tracking is enhanced with a detection stage, based on the h-dome transform. Preliminary experiments on several image sequences demonstrate that the new tracking algorithm is able to reconstruct the lineage tree, at least until the 350-cell stage, without manual intervention, at low computational cost and with low error rates.