Related Experiment Video
Updated: Jul 10, 2026
11:13
Electrophysiological Analysis of human Pluripotent Stem Cell-derived Cardiomyocytes (hPSC-CMs) Using Multi-electrode Arrays (MEAs)
Published on: May 12, 2017
Unsupervised machine learning-assisted multimodal characterization of cardiomyocytes on a thin-film-transistor
Xingzhou Hu1, Junichi Sugita2, Katsuhito Fujiu2,3,4
1Institute of Industrial Science, The University of Tokyo, 4-6-1 Komaba, Meguro-ku, Tokyo 153-8505, Tokyo, Japan.
Biomedical Physics & Engineering Express
|July 8, 2026
Summary
This study introduces a machine learning framework for analyzing cardiomyocyte activity using multimodal data from transparent thin-film-transistor microelectrode arrays (TFT-MEAs). The system enables label-free, automated detection of cardiac network changes for drug discovery and toxicity screening.
Area of Science:
- Cardiology
- Biomedical Engineering
- Computational Biology
Background:
- Microelectrode arrays (MEAs) are crucial for in vitro cardiac tissue analysis.
- Simultaneous electrophysiological and optical data acquisition offers comprehensive insights.
- Existing methods may lack scalability or automated analysis for complex cardiac network dynamics.
Purpose of the Study:
- To develop a machine learning-assisted framework for multimodal characterization of cardiomyocyte activity.
- To enable label-free, automated detection of cardiac network changes using integrated electrophysiological and optical data.
- To validate the framework's reliability and biological relevance for drug discovery and toxicity screening.
Main Methods:
- Utilized a transparent thin-film-transistor microelectrode array (TFT-MEA) for simultaneous data acquisition.
- Integrated multimodal features (electrophysiological and optical) for analysis.
- Employed unsupervised learning for pattern identification and supervised learning with SHapley Additive Explanations for validation and feature importance assessment.
Main Results:
- Unsupervised learning successfully identified cardiomyocyte contraction patterns without prior labels.
- Pharmacological validation with isoprenaline demonstrated automated detection of increased beating frequency.
- Interpretable machine learning confirmed feature importance and biological relevance of emergent labels.
Conclusions:
- The developed framework enables scalable, label-free, and automated assessment of cardiomyocyte networks.
- This approach facilitates the extraction of biologically meaningful insights from complex multimodal datasets.
- The framework is applicable for detecting diverse drug- or disease-induced cardiac network changes, supporting drug discovery and mechanistic studies.
