yeaz

yeaz segments Saccharomyces cerevisiae microscopy images using a U-Net-based convolutional neural network to provide label-free, high-accuracy cell segmentation and temporal tracking for quantitative analysis of yeast cell biology.

Key Features:

• High-quality segmented dataset: A publicly available dataset of over 10,000 segmented yeast cells covering mutants, stressed cells, and time-course studies.
• U-Net-based CNN architecture: Uses a U-Net convolutional neural network architecture optimized for image segmentation of yeast cells.
• Label-free segmentation capability: Achieves accurate cell border identification without requiring fluorescent channels.
• Cell-cell boundary test: Implements a boundary test to improve segmentation accuracy in crowded or irregularly shaped cells without fluorescence input.
• Bipartite graph matching algorithm: Applies bipartite graph matching for reliable tracking of cell identities across sequential images.
• Early budding event detection: Detects subtle budding events with high reliability to support studies of budding dynamics.

Scientific Applications:

• Precise yeast segmentation: Segmentation of Saccharomyces cerevisiae cells across diverse experimental conditions for quantitative image analysis.
• Cell cycle and morphogenesis studies: Analysis of cell cycle dynamics and morphogenetic processes in budding yeast.
• Budding dynamics detection: Early detection and monitoring of budding events in time-course experiments.
• Comparative mutant analysis: Comparative analysis of morphogenesis control in cyclin mutant versus wild-type cells.

Methodology:

Training a U-Net-based CNN on the annotated dataset of segmented yeast images; applying a cell-cell boundary test for label-free border detection; and using a bipartite graph matching algorithm for temporal tracking.