3D cell morphology detection by association for embryo heart morphogenesis.

Area of Science:

• Biomedical Engineering
• Developmental Biology
• Computational Biology

Background:

• Cardiac regenerative medicine requires understanding embryonic cardiac muscle growth.
• Accurate 3D cell segmentation is crucial for studying cardiomyocyte behavior.
• Challenges include low signal-to-noise and intensity inhomogeneity in dense tissues.

Purpose of the Study:

• To develop a robust computational tool for segmenting individual myocardial cells in 3D.
• To extract quantitative morphological parameters of cardiomyocytes from light-sheet microscopy images.
• To provide insights into cardiac tissue growth mechanisms.

Main Methods:

• A pipeline combining a neural network for 2D nuclei/membrane detection.
• Graph-based global association for 3D nuclei reconstruction.
• An optimization using network flow and alternating direction method of multipliers.
• An active mesh model initialized with 3D nuclei for cell segmentation.

Main Results:

• Successful extraction of cellular morphological parameters from 3D multifluorescence images of murine hearts.
• Demonstrated robustness in segmenting dense tissues with challenging imaging conditions.
• Qualitative and quantitative evaluations show superior efficiency compared to existing methods.

Conclusions:

• The developed segmentation tool accurately extracts cardiomyocyte morphology from 3D microscopy data.
• This method advances computational approaches for cardiac tissue engineering research.
• Provides a foundation for further investigation into cardiac development and regeneration.