CenFind

CenFind detects centrioles in immunofluorescence images and quantifies cell-level centriole numbers for reproducible, high-throughput analysis.

Key Features:

• Automatic detection: Uses SpotNet, a multi-scale convolutional neural network, to identify sparse, minute centriole and procentriole foci in high-resolution images.
• High accuracy and robustness: Achieves an average F₁-score greater than 90% across test sets.
• Channel-intrinsic detection: Operates on single-channel immunofluorescence datasets without requiring multi-channel inputs or hard-coded parameters.
• Integration with nucleus detection: Incorporates a StarDist-based nucleus detector to associate detected centrioles and procentrioles with their respective cells for cell-level scoring.
• Modular design: Implemented as a modular pipeline component suitable for integration into computational workflows.

Scientific Applications:

• High-throughput centriole quantification: Automated scoring of centriole numbers across large image sets.
• Cell-level centriole assignment: Linking centrioles and procentrioles to nuclei to enable per-cell analyses.
• Studies of centriole biology: Quantitative analysis of centriole dynamics relevant to cell division, signaling, and structural organization.
• Disease and functional studies: Large-scale analyses of centriole function in health and disease contexts using immunofluorescence data.

Methodology:

SpotNet, a multi-scale convolutional neural network, was trained on a dataset compiled from diverse experimental settings; performance was evaluated against existing detection methods; the pipeline includes a StarDist-based nucleus detector and is configured to operate on single-channel immunofluorescence images without fixed parameters.