YeastNet

YeastNet performs semantic segmentation of budding yeast cells from bright-field microscopy images to generate segmentation masks for single-cell analysis.

Key Features:

• U-Net convolutional network: Employs a U-Net architecture trained for semantic segmentation of bright-field images of budding yeast cells to produce pixel-accurate segmentation masks.
• Implementation: Implemented as a Python3 library using PyTorch 1.0.
• Input data support: Designed to handle bright-field microscopy and high-resolution live-cell microscopy image datasets typical of yeast experiments.
• Segmentation outputs: Generates segmentation masks suitable for downstream cell labelling, tracking, and morphological measurements.
• Performance improvement: Demonstrates superior accuracy compared to non-trainable classic algorithms and existing state-of-the-art yeast cell segmentation tools.
• Automation and throughput: Automates segmentation to accelerate analysis of large datasets in high-throughput microscopy studies.

Scientific Applications:

• Single-cell analysis: Enables extraction of per-cell measurements and population heterogeneity studies in budding yeast.
• Cell labelling and tracking: Provides segmentation masks that support automated cell labelling and time-resolved tracking of yeast cells.
• Morphology and dynamics studies: Supports analysis of cellular morphology and behavior over time in live-cell microscopy experiments.

Methodology:

Uses a trained U-Net convolutional network for semantic segmentation of bright-field microscopy images, implemented in Python3 with PyTorch 1.0 to produce segmentation masks for yeast cells.