E. coli Tracking (ECT) Project

A framework for bacterial cell segmentation, tracking, and growth analysis using deep learning and classical image processing.

Overview

This project provides tools and analysis pipelines for:

• Cell Segmentation: Multiple deep learning models (Omnipose, Cellpose, DeLTA2) and a classical method
• Cell Tracking: Lineage tracking and cycle analysis
• Morphological Analysis: Cell shape and size measurements
• Growth Dynamics: Population and single-cell growth analysis

Quick Access Options

Option 1: View Pre-computed Results (Recommended for Quick Review)

The complete analysis results are available as HTML reports in the docs/ folder:

• Model Comparison: docs/model_comparison_final.html
• Elongated Morphology: docs/elongated_morphology_validation.html
• Tracking Analysis: docs/tracking_cycle_analysis.html
• Thesis Document: docs/LR_Thesis.pdf

(View directly in your browser — no setup required.)

Option 2: Download Datasets and Re-run Analysis

For full reproducibility, download the complete dataset from Zenodo and follow the setup below.

Dataset Access

Download from Zenodo

Dataset DOI: 10.5281/zenodo.17351896

Citation:

Reed, L. (2025). E. coli Tracking (ECT) Dataset: Time-lapse Microscopy and Segmentation Analysis [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.17351896

Quick Setup for Reproducibility

1. Download the dataset:

   • Go to Zenodo
   • Download data.zip (~5GB)

2. Extract and organize:

   # Extract downloaded data
   unzip data.zip -d data
   
   # Ensure the data folder is in your project root
   # Your structure should be: ECT/data/

3. Verify data structure:

   # Check that you have the expected directories
   ls data/
   # Should show: timelapse_data/, processed/, examples/, models/, elongated_morphology/

Quick Start

1. Setup Environment

# Clone the repository
git clone https://github.com/LucasGeno/ECT.git
cd ECT

# Create conda environment
# The environment file installs all dependencies needed for all notebooks, tracking, and visualization
conda env create -f notebooks/01_env_setup/environment_tracking.yml
conda activate tracking_env

# Install Jupyter kernel
python -m ipykernel install --user --name tracking_env --display-name "Python (tracking_env)"

*A separate environment file (environment_napari_btrack_arboretum.yml) is included for users wishing to reproduce the Napari-based lineage reconstruction and visualization workflows.

2. Verify Data Access

# Check if data directory exists and contains expected files
python -c "
from pathlib import Path
data_dir = Path('data')
if data_dir.exists():
    print('✓ Data directory found')
    print(f'  Contents: {list(data_dir.iterdir())}')
else:
    print('❌ Data directory not found')
    print('Please download and extract data from Zenodo first')
"

Usage

Run Analysis

# Start Jupyter Lab
jupyter lab

# Open notebooks in order:
# 1. notebooks/01_env_setup/01_environment_setup.ipynb
# 2. notebooks/02_data_preprocessing/data_loader.ipynb
# 3. notebooks/03_model_comparison/model_comparison_final.ipynb
# 4. notebooks/04_elongated_morphology/elongated_morphology_validation.ipynb
# 5. notebooks/05_tracking_analysis/tracking_cycle_analysis.ipynb

Project Structure

ECT/
├── data/                        # Downloaded from Zenodo
│   ├── timelapse_data/          # LB & M9 tracking datasets
│   ├── training_data/           # Training 
│   ├── processed/               # Precomputed results and test data
│   ├── elongated_morphology/    # Elongated-cell benchmark
│   ├── examples/                # Sample datasets
│   └── models/                  # Pre-trained model files
├── docs/                        # Reports and thesis
│   ├── LR_Thesis.pdf
│   ├── model_comparison_final.html
│   ├── elongated_morphology_validation.html
│   └── tracking_cycle_analysis.html
├── notebooks/                   # Modular Jupyter notebooks
│   ├── 01_env_setup/
│   ├── 02_data_preprocessing/
│   ├── 03_model_comparison/
│   ├── 04_elongated_morphology/
│   └── 05_tracking_analysis/
├── src/                         # Utility scripts
│   ├── data/
│   └── utils/
├── setup_data.py                # Automated directory setup
└── README.md

Example Output

Cell Lineage Visualization

The tracking analysis produces interactive lineage trees showing cell division patterns across generations:

Lineage tree reconstructed from LB timelapse data using btrack + Arboretum.
Interactive visualization available in notebooks/05_tracking_analysis/tracking_reproducibility_demo/

Dataset Information

Available Datasets

Primary Datasets (from Zenodo):

1. Time-lapse Datasets (data/timelapse_data/):

   • LB Dataset: 361 frames, rich growth conditions
   • M9 Dataset: 601 frames, minimal growth conditions
   • Format: TIFF stacks with corresponding masks and tracking data

2. Training/Test Datasets (data/processed/):

   • Training Set: 576 frames (288 LB + 288 M9)
   • Test Set: 118 frames (45 LB + 73 M9)
   • Precomputed Results: 1,498 segmentation outputs from 7 models

3. Specialized Datasets:

   • Elongated Morphology: 180 frames with filamentous cells
   • Pre-trained Models: 7 segmentation models
   • Example Data: Sample datasets for testing

Data Format Specifications

Time-lapse Images:

• Format: TIFF stack (.tif)
• Dimensions: (time, height, width)
• Pixel size: 0.065 µm/px
• Frame interval: 0.5 min
• Naming: original_images.tif, bf_frames.tif

Segmentation Masks:

• Format: TIFF stack (.tif)
• Values: Integer labels (0 = background, >0 = cell ID)
• Naming: masks.tif

Tracking Data:

• Format: Parquet (.parquet) or CSV (.csv)
• Required columns: track_id, t, area, parent
• Optional columns: x, y, width, length, generation

Analysis Workflow

1. Data Preprocessing

• Load and inspect microscopy data
• Validate data quality

2. Model Comparison

• Evaluate segmentation models
• Compare performance metrics
• Generate publication figures

3. Morphological Analysis

• Analyze cell shapes and sizes
• Validate on elongated cells
• Compute width distributions

4. Tracking Analysis

• Parse cell lineages
• Compute generation times
• Validate balanced growth

Key Features

• Segmentation: Omnipose, Cellpose, DeLTA2, Random Forest Watershed
• Tracking: Cell lineage reconstruction with btrack + Arboretum
• Analysis: Cell counting, morphometrics, growth dynamics, cycle analysis

Run Analysis

# Start Jupyter Lab
jupyter lab

# Open notebooks in order:
# 1. notebooks/01_env_setup/01_environment_setup.ipynb
# 2. notebooks/02_data_preprocessing/data_loader.ipynb
# 3. notebooks/03_model_comparison/model_comparison_final.ipynb
# 4. notebooks/04_elongated_morphology/elongated_morphology_validation.ipynb
# 5. notebooks/05_tracking_analysis/tracking_cycle_analysis.ipynb

Dependencies

• Python 3.9+, NumPy, Pandas, SciPy
• PyTorch, Cellpose, Omnipose
• Napari, Jupyter Lab
• tifffile, h5py, parquet

Troubleshooting

• Data not found: Download from Zenodo first
• Missing dependencies: Use environment setup notebook
• Quick review: Check HTML reports in docs/ folder

License

This project is licensed under the MIT License - see the LICENSE file for details.

Dataset:

Reed, L. (2025). E. coli Tracking (ECT) Dataset: Time-lapse Microscopy and Segmentation Analysis [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.17351896

Thesis:

**Thesis (in preparation)**:  
Reed, L. (2025). *Deep Learning-Enhanced Segmentation and Tracking of E. coli in Time-Lapse Microscopy*. MSc Thesis, Vrije Universiteit Amsterdam.

Acknowledgments

• Omnipose and Cellpose communities
• DeLTA2 developers
• Napari project
• Scientific Python ecosystem
• O2 lab
• J. van Heerden (supervisor)
• P. Savakis (daily supervisor)