Related Experiment Video
Updated: Aug 15, 2025
09:34
Automated Production of Human Induced Pluripotent Stem Cell-Derived Cortical and Dopaminergic Neurons with Integrated Live-Cell Monitoring
Published on: August 6, 2020
7.3K
Quality Control of Human Pluripotent Stem Cell Colonies by Computational Image Analysis Using Convolutional Neural
Anastasiya Mamaeva1, Olga Krasnova2, Irina Khvorova3
1Mathematical Biology and Bioinformatics Lab, Peter the Great St. Petersburg Polytechnic University, 195251 Saint Petersburg, Russia.
International Journal of Molecular Sciences
|January 8, 2023
Summary
We developed an AI classifier to monitor human pluripotent stem cells (hPSCs) in culture. This automated method accurately identifies cell colonies with optimal pluripotency and clonality, ensuring stem cell quality for research and therapy.
Area of Science:
- Stem cell biology
- Biotechnology
- Computational biology
Background:
- Human pluripotent stem cells (hPSCs) are vital for regenerative medicine and disease modeling.
- Maintaining the pluripotent and clonal status of hPSCs in culture is critical for reliable research and therapeutic applications.
- Current methods for monitoring hPSC quality rely on manual observation, which can be subjective and labor-intensive.
Purpose of the Study:
- To develop and validate a non-invasive, automated method for assessing the morphological quality of human embryonic stem cell (hESC) colonies.
- To utilize machine learning, specifically convolutional neural networks (CNNs), for the image-based classification of hESC colony phenotypes.
- To correlate morphological phenotypes with the pluripotency and clonality maintenance potential of hESCs.
Main Methods:
- Development of a CNN classifier trained on phase-contrast images of hESC colonies.
- Visual assessment of colony morphology by experts to establish ground truth for classifier training.
- Analysis of informative spatial scales for image classification, focusing on features between single-cell and colony dimensions.
- Proteomic analysis to investigate molecular differences between hESC colonies with distinct morphological phenotypes.
Main Results:
- The CNN classifier achieved 89% accuracy in predicting hESC colony phenotypes.
- The optimal spatial scale for classification was identified as approximately 144 μm.
- Proteomic analysis confirmed molecular differentiation in cells exhibiting different morphological phenotypes.
- The study demonstrated the potential for automated, non-invasive monitoring of hPSC developmental anomalies.
Conclusions:
- Automated image analysis using machine learning provides an effective, non-invasive method for quality control of hPSCs.
- The developed classifier can distinguish between hESC colonies with high and low potential for maintaining pluripotency and clonality.
- This approach facilitates the identification of undesirable developmental anomalies during hPSC culture, supporting reliable stem cell propagation.
