Assessment of Stem Cell Viability through Visual Analysis Coupled with Teachable Machine.

Area of Science:

• Biomedical Research
• Cell Biology
• Drug Discovery

Background:

• Cell viability assessment is crucial for drug discovery and biomedical research.
• Current methods like staining and colorimetric assays are time-consuming, costly, and prone to variability.
• Existing AI tools still require cell staining for viability determination.

Purpose of the Study:

• To develop an AI model for determining cell viability based on visual cell characteristics.
• To automate the labeling process for training data to improve efficiency.
• To enhance the speed, cost-effectiveness, and reproducibility of cell viability analysis.

Main Methods:

• Utilized Google's Teachable Machine, a web-based AI tool.
• Developed an automated labeling process using contour functions for individual cell extraction.
• Created diverse datasets to train and evaluate multiple AI models.

Main Results:

• Achieved an accuracy of over 80% with the best-performing model.
• Demonstrated the model's ability to assess cell viability directly from visual features.
• Showcased significant improvements in time efficiency and reduced manual labor.

Conclusions:

• The developed AI model offers a faster, more cost-effective, and less variable approach to cell viability assessment.
• This method enhances the efficacy and reproducibility of experiments in drug discovery and biological research.
• Potential to minimize analysis time, expenses, and individual variability in cell-based assays.