Skip to content
/ btrack Public
forked from quantumjot/btrack

Bayesian multi-object tracking

License

Notifications You must be signed in to change notification settings

nthndy/btrack

 
 

Repository files navigation

PyPI Downloads Black Tests pre-commit Documentation codecov

logo

Bayesian Tracker (btrack) 🔬💻

btrack is a Python library for multi object tracking, used to reconstruct trajectories in crowded fields. Here, we use a probabilistic network of information to perform the trajectory linking. This method uses spatial information as well as appearance information for track linking.

The tracking algorithm assembles reliable sections of track that do not contain splitting events (tracklets). Each new tracklet initiates a probabilistic model, and utilises this to predict future states (and error in states) of each of the objects in the field of view. We assign new observations to the growing tracklets (linking) by evaluating the posterior probability of each potential linkage from a Bayesian belief matrix for all possible linkages.

The tracklets are then assembled into tracks by using multiple hypothesis testing and integer programming to identify a globally optimal solution. The likelihood of each hypothesis is calculated for some or all of the tracklets based on heuristics. The global solution identifies a sequence of high-likelihood hypotheses that accounts for all observations.

We developed btrack for cell tracking in time-lapse microscopy data.

Installation

btrack has been tested with Python on x86_64 macos>=11, ubuntu>=20.04 and windows>=10.0.17763. Note that btrack<=0.5.0 was built against earlier version of Eigen which used C++=11, as of btrack==0.5.1 it is now built against C++=17.

Installing the latest stable version

pip install btrack

Usage examples

Visit btrack documentation to learn how to use it and see other examples.

Cell tracking in time-lapse imaging data

We provide integration with Napari, including a plugin for graph visualization, arboretum.

CellTracking
Video of tracking, showing automatic lineage determination


Development

The tracker and hypothesis engine are mostly written in C++ with a Python wrapper. If you would like to contribute to btrack, you will need to install the latest version from GitHub. Follow the instructions on our developer guide.


Citation

More details of how this type of tracking approach can be applied to tracking cells in time-lapse microscopy data can be found in the following publications:

Automated deep lineage tree analysis using a Bayesian single cell tracking approach
Ulicna K, Vallardi G, Charras G and Lowe AR.
Front in Comp Sci (2021)
doi:10.3389/fcomp.2021.734559

Local cellular neighbourhood controls proliferation in cell competition
Bove A, Gradeci D, Fujita Y, Banerjee S, Charras G and Lowe AR.
Mol. Biol. Cell (2017)
doi:10.1091/mbc.E17-06-0368

@ARTICLE {10.3389/fcomp.2021.734559,
   AUTHOR = {Ulicna, Kristina and Vallardi, Giulia and Charras, Guillaume and Lowe, Alan R.},
   TITLE = {Automated Deep Lineage Tree Analysis Using a Bayesian Single Cell Tracking Approach},
   JOURNAL = {Frontiers in Computer Science},
   VOLUME = {3},
   PAGES = {92},
   YEAR = {2021},
   URL = {https://www.frontiersin.org/article/10.3389/fcomp.2021.734559},
   DOI = {10.3389/fcomp.2021.734559},
   ISSN = {2624-9898}
}
@ARTICLE {Bove07112017,
  author = {Bove, Anna and Gradeci, Daniel and Fujita, Yasuyuki and Banerjee,
    Shiladitya and Charras, Guillaume and Lowe, Alan R.},
  title = {Local cellular neighborhood controls proliferation in cell competition},
  volume = {28},
  number = {23},
  pages = {3215-3228},
  year = {2017},
  doi = {10.1091/mbc.E17-06-0368},
  URL = {http://www.molbiolcell.org/content/28/23/3215.abstract},
  eprint = {http://www.molbiolcell.org/content/28/23/3215.full.pdf+html},
  journal = {Molecular Biology of the Cell}
}

About

Bayesian multi-object tracking

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 57.0%
  • C++ 41.1%
  • Cuda 1.3%
  • Other 0.6%