Induced Pluripotent Stem Cells Detection via Ensemble Yolo Network

Area of Science:

• Stem Cell Biology
• Biotechnology
• Regenerative Medicine

Background:

• Induced pluripotent stem cells (iPSCs) hold significant promise for regenerative medicine and industrial applications.
• Accurate and automated identification of iPSCs is crucial for efficient production but remains challenging without cell staining.
• Developing noninvasive methods is key to streamlining iPSC research and application.

Purpose of the Study:

• To create an accurate and noninvasive method for detecting iPSCs colonies.
• To enhance the efficiency of iPSC production through automated colony identification.
• To establish a robust detection system using bright-field microscopy images.

Main Methods:

• An ensemble Yolo network was employed for iPSCs colonies detection.
• The method utilizes self-collected bright-field microscopy images, avoiding cell staining.
• Test-time augmentation (TTA) was incorporated to further optimize detection accuracy.

Main Results:

• The developed method achieved a high F1 score of 0.867.
• The mean average precision (mAP) score reached 0.898.
• The ensemble Yolo network with TTA demonstrated superior performance compared to mainstream methods.

Conclusions:

• The proposed noninvasive iPSCs colony detection method is accurate and efficient.
• This automated approach facilitates iPSC research and industrial applications.
• The ensemble Yolo network offers a promising solution for iPSC identification challenges.