A graph-based cell tracking algorithm with few manually tunable parameters and automated segmentation error correction

Source: Löffler et al. (2022), doi: https://doi.org/10.1371/journal.pone.0249257.g001 License: CC BY 4.0

---

This repository contains the code to the publication:

"A graph-based cell tracking algorithm with few manually tunable parameters and automated segmentation error correction" Löffler K, Scherr T, Mikut R (2021) PLOS ONE 16(9): e0249257. https://doi.org/10.1371/journal.pone.0249257

The deep learning based segmentation approach which is used together with the graph-based tracking for submission to the Cell Tracking Challenge as team KIT-Sch-GE(2) (renamed to KIT-GE(3)) is mainted at https://github.com/TimScherr/KIT-GE-3-Cell-Segmentation-for-CTC .

The code has been tested on Windows and Linux using Python 3.8.

Setup

1) Create folder structure

• create a project directory LOCAL_DIR
• create two folders named data and results in LOCAL_DIR
• clone the code and install dependencies:

conda create --name venv_graph_tracking_kit_sch_ge_2021 python==3.8
conda activate venv_graph_tracking_kit_sch_ge_2021
git clone https://github.com/kaloeffler/KIT-GE-3-Cell-Tracking-for-CTC.git
pip install -e ./KIT-GE-3-Cell-Tracking-for-CTC

so the final structure is

LOCAL_DIR
└───data  (contains the ground truth data sets)
└───KIT-GE-3-Cell-Tracking-for-CTC (contains our tracking code)
└───results (synthetically degraded segmentation masks will be stored here)

2) Install gurobi

see help_gurobi.txt

Run tracking

• the tracking algorithm can be used with any 2D/3D image data with a segmentation which needs to be provided by the user
• it is assumed that the image data and segmentation data have a similar naming convention as used by the cell tracking challenge (http://celltrackingchallenge.net)

python -m run_tracking --image_path IMAGE_PATH --segmentation_path SEGMENTATION_PATH --results_path RESULTS_PATH

Reproduce synthetic data sets

1) Download data sets

• go to http://celltrackingchallenge.net and download the training data sets Fluo-N2DH-SIM+ and Fluo-N3DH-SIM+, unpack and save in data directory:

LOCAL_DIR
└───data  (contains the ground truth data sets)
│   └───Fluo-N2DH-SIM+
│   └───Fluo-N3DH-SIM+
└───KIT-GE-3-Cell-Tracking-for-CTC
└───results (synthetically degraded segmentation masks will be stored here)

2) Run code

• run create_synth_segm_data.py to create synthetically degraded segmentation mask images

python -m create_synth_segm_data

Citation

If you use our work in your research, please cite:

@article{loeffler2021,
    doi = {10.1371/journal.pone.0249257},
    author = {Löffler, Katharina and Scherr, Tim and Mikut, Ralf},
    journal = {PLOS ONE},
    title = {A graph-based cell tracking algorithm with few manually tunable parameters and automated segmentation error correction},
    year = {2021},
    volume = {16},
    pages = {1-28},
    number = {9},
}