Napari NuclePhaser

Napari NuclePhaser enables cell nuclei detection and counting in phase contrast images using YOLO-based deep learning models, supporting cell counting and tracking without staining.

Key Features:

• YOLO-based nuclei detection: Detects cell nuclei in phase contrast images using deep learning object detection YOLO models trained on more than 100,000 640x640 pixels images with more than 3 million nuclei from 4 cell cultures (CHO, HEK293, iPSCs, and MSCs).
• Generalized cross-platform performance: Models detect nuclei across different cell types and imaging conditions, including different microscopes, cameras, and contrast methods, using full-scale augmentations during training.
• Arbitrary image size support: Applies sliced inference methods to process images of any size, enabling studies of large quantities of cells.
• Training-free calibration: Optimizes confidence thresholds to fine-tune models for specific cell types and imaging options without retraining.
• Model accuracy evaluation: Provides testing algorithm to evaluate calibrated model accuracy for specific use cases.

Scientific Applications:

• Cell proliferation studies: Quantifies cell proliferation rates through automated nuclei detection and counting.
• Single cell tracking: Tracks individual cells over time without requiring fluorescent staining or sample fixation.
• Label-free cell analysis: Performs cell counting and monitoring using phase contrast, differential interference contrast, and other brightfield microscopy methods.

Methodology:

YOLO deep learning object detection models trained on a dataset of phase contrast images from CHO, HEK293, iPSC, and MSC cell cultures, with sliced inference for arbitrary image sizes and confidence threshold optimization for training-free calibration.