Automated single cardiomyocyte characterization by nucleus extraction from dynamic holographic images using a fully...

Area of Science:

• Biomedical Engineering
• Cell Biology
• Microscopy

Background:

• Human-induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) are crucial for cardiac research.
• Characterizing hiPSC-CM beating patterns is vital for drug screening and disease modeling.
• Current methods for analyzing hiPSC-CM beating can be limited by noise and lack of single-cell resolution.

Purpose of the Study:

• To develop an automated method for accurate cardiomyocyte nucleus extraction from quantitative phase images (QPIs).
• To enable precise characterization and quantification of individual hiPSC-CM beating patterns.
• To improve the reliability and informativeness of hiPSC-CM beating analysis.

Main Methods:

• Utilized digital holographic microscopy to obtain time-lapse quantitative phase imaging (QPIs) of hiPSC-CMs.
• Developed a novel fully convolutional neural network (FCN)-based architecture for automated nucleus extraction from QPIs.
• Employed pixel classification techniques for accurate segmentation of cardiomyocyte nuclei.

Main Results:

• The FCN-based method achieved accurate extraction of cardiomyocyte nuclei, overcoming challenges related to variations in shape, size, and orientation.
• Reconstructed and quantified beating patterns of individual hiPSC-CMs with reduced noise compared to whole-image analysis.
• Enabled efficient measurement of multiple beating parameters at the single-cell level.

Conclusions:

• The proposed automated FCN-based nucleus extraction method significantly enhances the characterization of hiPSC-CM beating patterns.
• This approach provides a more informative and less noisy analysis at the single-cell level.
• Facilitates robust assessment of hiPSC-CM function for various research applications.