High-volume, label-free imaging for quantifying single-cell dynamics in induced pluripotent stem cell colonies.

Area of Science:

• Stem Cell Biology
• Artificial Intelligence
• Biotechnology

Background:

• Characterizing induced pluripotent stem cells (iPSCs) during culture is crucial for regenerative medicine.
• Non-invasive methods are needed to monitor iPSC proliferation and health without labeling.

Purpose of the Study:

• To develop an AI pipeline for automated nuclear segmentation and mitosis detection in unlabeled iPSCs using phase contrast microscopy.
• To enable non-invasive, high-throughput tracking of individual iPSCs and assessment of cell division rates.

Main Methods:

• A 2D and 3D U-Net convolutional neural network architecture was employed for image analysis.
• Fluorescence data was used to train models for phase contrast image segmentation, avoiding manual annotation.
• Classical image processing routines were integrated for robust nucleus segmentation.

Main Results:

• The AI pipeline achieved high accuracy in segmenting nuclei and detecting mitotic events (average F1 score of 0.94).
• The model demonstrated generalizability across different cell densities and pluripotent cell lines.
• Non-invasive assessment of mitosis and cell division rates was performed on hundreds of thousands of cells over 36 hours.

Conclusions:

• The developed AI pipeline provides a reliable and non-invasive method for characterizing unlabeled iPSCs.
• This approach facilitates the assessment of cell state and health by monitoring cell division dynamics.
• The method is applicable to large-scale, long-term monitoring of iPSC cultures.