🧬 3D Cell Tracking with Dask for IEEE DSAA 2019

This repository hosts the code for our paper accepted in IEEE DSAA 2019: "Lightweight and Scalable Particle Tracking and Motion Clustering of 3D Cell Trajectories". The project provides a robust pipeline for tracking and clustering cell trajectories using various implementations, including serial, local Dask, and Dask on a cluster.

---

🌟 Project Overview

This repository features three versions of the code:

• 🔗 Serial Version: A straightforward implementation for running on a single machine.
• ⚙️ Local Dask Version: A parallelized version that leverages Dask for local computation.
• ☁️ Dask on Cluster Version: A scalable implementation for running on distributed computing clusters.

Key Technologies Used

• Python 3.6
• Scientific Computing Libraries: NumPy, SciPy, scikit-learn, matplotlib
• Computer Vision Tools: OpenCV 3.1 for detection and tracking
• Parallel and Distributed Computing:
  • Dask Arrays & DataFrames for scalable processing
  • Dask-ML & Dask-NDArray for distributed machine learning
  • Joblib & multiprocessing for local parallel execution

---

📄 Paper Details

The methods and results of this project are detailed in our paper, which can be accessed through the following links:

• IEEE Download Page: Link to IEEE
• arXiv Preprint: Link to arXiv

Citation

For those interested in citing this work, please use the following format:

@INPROCEEDINGS{8964177,
  author={Sedigh Fazli, Mojtaba and Stadler, Rachel V. and Alaila, BahaaEddin and Vella, Stephen A. and Moreno, Silvia N. J. and Ward, Gary E. and Quinn, Shannon},
  booktitle={2019 IEEE International Conference on Data Science and Advanced Analytics (DSAA)}, 
  title={Lightweight and Scalable Particle Tracking and Motion Clustering of 3D Cell Trajectories}, 
  year={2019},
  volume={},
  number={},
  pages={412-421},
  keywords={Large scale 3D Cell Tracking;Motion Trajectories;Geodesic distance;Spectral Clustering;Toxoplasma gondii;Martin Distance},
  doi={10.1109/DSAA.2019.00056}
}

---

🚀 How to Run the Code

Step 1: Prepare Your Data

Ensure that all image data is stored locally on your machine. Update the following parameters in the code before execution:

• folders: Path to the directories containing image slices.
• sample_address: Path to a sample TIFF image to extract initial parameters.
• cluster_numbers: The number of clusters to use in the analysis.

Additional parameters can be modified within the script based on your specific application needs.

Step 2: Execute the Code

After setting up the parameters, run the code and measure the wall time with:

time python <The_code_filename.py>

This command will provide a performance benchmark by displaying the total execution time.

---

❓ Why a Single Python File?

We opted to include all functions in a single script to facilitate seamless wall-time computation for the entire pipeline. This structure allows us to efficiently test and compare running times across different implementations.

---

🛠️ Prerequisites

Before running the code, ensure you have the following libraries installed:

pip install numpy scipy scikit-learn matplotlib opencv-python dask dask-ml joblib

For the cluster version, additional Dask configurations may be required.

---

✨ Customization Tips

• Cluster Configuration: Adjust Dask scheduler and client settings to optimize performance on distributed systems.
• Memory Management: For large datasets, tweak Dask's memory handling parameters to prevent overflow.

---

📧 Contact

For questions, feedback, or collaboration opportunities, feel free to reach out:

• 📧 mfazli@stanford.edu
• 📧 mfazli@meei.harvard.edu

🚀 Dive into the code and explore the powerful capabilities of lightweight and scalable 3D cell tracking!