Deep Learning Resolves Myovascular Dynamics in the Failing Human Heart

Area of Science:

• Cardiovascular biology
• Computational pathology
• Biomedical imaging analysis

Background:

• The adult mammalian heart has limited cardiomyocyte (CM) proliferation.
• Quantifying CM cell division requires extensive microscopic image analysis.
• Existing methods are labor-intensive and may lack scalability.

Purpose of the Study:

• To introduce CardioCount, a deep learning pipeline for automated nuclei scoring in microscopy images.
• To investigate cardiomyocyte and cardiac endothelial cell interactions in the adult human heart.
• To explore the relationship between vascular rarefaction and CM hypertrophy in heart failure.

Main Methods:

• Development of a deep learning-based image analysis pipeline named CardioCount.
• Application of CardioCount to a large dataset of 368,434 human microscopic cardiac images.
• Statistical analysis to correlate cellular changes with disease states.

Main Results:

• Evidence of coupled growth between cardiomyocytes and cardiac endothelial cells in the adult human heart.
• Demonstration of an interrelationship between vascular rarefaction and CM hypertrophy in end-stage heart failure.
• CardioCount provides a scalable and rigorous method for analyzing cardiac cell populations.

Conclusions:

• CardioCount facilitates the quantitative analysis of cellular dynamics in cardiac tissue.
• The findings highlight coordinated growth mechanisms in the healthy adult heart.
• The study elucidates pathological interactions in heart failure, offering potential therapeutic targets.