Generative adversarial network model to classify human induced pluripotent stem cell-cardiomyocytes based on...

Area of Science:

• Biomedical Engineering
• Computational Biology
• Cell Biology

Background:

• Accurate classification of human cardiomyocytes is crucial for understanding cellular functions.
• Limited scale and diversity of experimental image data hinder computational analysis.
• Generative Adversarial Networks (GANs) show potential for data augmentation.

Purpose of the Study:

• To develop a GAN-based method for generating synthetic human cardiomyocyte image data.
• To enhance the classification accuracy of cells at various maturation stages.
• To improve the throughput of computational analysis for cellular structure and function.

Main Methods:

• Human induced pluripotent stem cell derived cardiomyocytes (hiPSC-CMs) were cultured on micropatterned hydrogels and glass plates.
• Optical measurements were performed for structural and functional analyses of hiPSC-CMs.
• A GAN model was trained using real image recordings of hiPSC-CMs.

Main Results:

• The GAN model successfully replicated true features from real cardiomyocyte image data.
• Incorporating synthetic data significantly improved cell classification accuracy compared to using real data alone.
• The proposed GAN-based method outperformed four conventional machine learning algorithms in data generalization and classification.

Conclusions:

• Synthetic data generated by GANs can effectively address challenges posed by limited sample sizes in biological studies.
• Integrating synthetic data enhances the reliability and scale of computational analysis of cellular images.
• This approach offers a valuable tool for advancing research in cardiomyocyte biology and related fields.