SyMBac

SyMBac generates synthetic phase contrast and fluorescence images of bacteria to produce training data for deep-learning-based cell segmentation and tracking in high-throughput timelapse imaging.

Key Features:

• Image modalities: Produces synthetic phase contrast and fluorescence micrographs matching experimental imaging modalities.
• Physics and optics simulation: Integrates cell growth models, physical interactions, and microscope optics simulations to produce realistic synthetic images.
• Simulation environments: Simulates bacterial growth in mother machine geometries and monolayer microcolonies.
• Instantaneous data generation: Synthetic training data can be produced virtually instantly to adapt to experimental conditions.
• Accurate ground truth annotations: Each synthetic image is accompanied by precise ground-truth cell positions and masks.
• Robust model training: Data generated supports training deep-learning models that are robust to variations in cell size and morphology and can outperform models trained on human-annotated data.
• Versatile experimental parameters: Can emulate a range of biological conditions, imaging platforms, and imaging modalities for dataset generation.
• Support for biological analysis: Enables generation of datasets used to study physiological dynamics such as bacterial growth and size regulation during dormancy.

Scientific Applications:

• Deep-learning segmentation and tracking: Generating training datasets for cell segmentation and object tracking models in timelapse microscopy.
• Algorithm benchmarking and validation: Providing precise ground-truth images and masks for benchmarking segmentation and tracking algorithms.
• Imaging-condition transfer: Simulating varied imaging platforms and modalities to test model robustness across experimental setups.
• Quantitative physiological studies: Enabling analysis of bacterial growth dynamics and size regulation under different growth conditions, including dormancy.

Methodology:

SyMBac integrates cell growth models, physical interactions, and microscope optics simulations to generate synthetic phase contrast and fluorescence images with paired ground-truth cell positions for mother machine and monolayer microcolony scenarios.