Calcium Spark Detection and Event-Based Classification of Single Cardiomyocyte Using Deep Learning.

Area of Science:

• Cardiology
• Biophysics
• Computational Biology

Background:

• Calcium (Ca2+) sparks are crucial for cardiomyocyte function, and their dysregulation contributes to heart disease.
• Traditional Ca2+ spark analysis methods struggle to detect weak signals and classify cell states effectively.
• Machine learning applications in biological data analysis are growing, but Ca2+ spark data remains underexplored.

Purpose of the Study:

• To develop a novel deep learning method for enhanced Ca2+ spark detection in cardiomyocytes.
• To create a machine learning model for classifying cardiomyocyte states based on Ca2+ spark characteristics.
• To provide a powerful tool for studying calcium signaling in cardiac pathology.

Main Methods:

• Developed a deep residual convolutional neural network for Ca2+ spark detection.
• Implemented an event-based logistic regression and binary classification model for cardiomyocyte classification.
• Validated the models using wild type vs. RyR2-R2474S± and vehicle vs. isoprenaline-insulted cardiomyocytes.

Main Results:

• The deep learning method detected more Ca2+ sparks, including weak events often missed by traditional methods.
• The classification model achieved 100% accuracy in distinguishing WT vs. RyR2-R2474S± cardiomyocytes.
• The model demonstrated 95.6% accuracy in differentiating vehicle vs. isoprenaline-insulted WT cardiomyocytes.

Conclusions:

• The novel deep learning approach significantly improves Ca2+ spark detection and analysis.
• The developed classification model accurately distinguishes between different cardiomyocyte states and disease models.
• This approach offers a powerful new avenue for research into calcium signaling and cardiac diseases.