Deep learning-based segmentation of subcellular organelles in high-resolution phase-contrast images

Area of Science:

• Cell Biology
• Microscopy
• Artificial Intelligence

Background:

• Quantitative analysis of biological images requires accurate cell and organelle extraction.
• Traditional methods struggle with complex features in grayscale images.
• Advancements in artificial intelligence offer solutions for image analysis challenges.

Purpose of the Study:

• To develop and validate machine learning-based segmentation models for subcellular structures.
• To enable accurate segmentation of organelles in high-resolution phase-contrast microscopy images.
• To provide a practical framework for studying cellular dynamics in unstained living cells.

Main Methods:

• Utilized a fine-tuned apodized phase-contrast microscopy system for label-free imaging.
• Employed machine learning models for segmentation of subcellular targets.
• Used fluorescent markers to generate ground truth masks for model training.

Main Results:

• Achieved accurate segmentation of organelles in high-resolution phase-contrast images.
• Demonstrated the effectiveness of deep learning-based segmentation for label-free microscopy.
• Validated the utility of the method for analyzing organelle dynamics.

Conclusions:

• Machine learning-based segmentation offers a robust solution for analyzing unstained living cells.
• The developed framework facilitates high-throughput studies of organelle dynamics.
• This approach advances label-free imaging techniques in cell biology.