Cell segmentation in phase contrast microscopy images via semi-supervised classification over optics-related features

Area of Science:

• Biomedical Imaging
• Cell Biology
• Computational Microscopy

Background:

• Phase-contrast microscopy is crucial for observing long-term multicellular processes but suffers from image artifacts.
• Computer-aided analysis of cell behavior in phase-contrast microscopy is hindered by poor image quality.

Purpose of the Study:

• To develop an image restoration method for phase contrast microscopy.
• To create a semi-supervised learning algorithm for accurate cell segmentation.

Main Methods:

• A computational model of phase contrast microscopy image formation using diffraction patterns to extract phase retardation features.
• Clustering of pixels based on phase retardation features to form phase-homogeneous atoms.
• Semi-supervised classification of these atoms for cell segmentation.

Main Results:

• The proposed method effectively restores phase retardation features from phase contrast microscopy images.
• The semi-supervised learning algorithm achieves high-quality segmentation of individual cells.
• The approach demonstrates superior performance compared to existing methods.

Conclusions:

• The developed image restoration and cell segmentation techniques significantly enhance the analysis of cell behavior from phase contrast microscopy data.
• This work addresses key challenges in computational phase contrast microscopy, paving the way for more reliable cell behavior studies.