Automated Deep Learning-Based System to Identify Endothelial Cells Derived from Induced Pluripotent Stem Cells

Area of Science:

• Biotechnology
• Cell Biology
• Artificial Intelligence

Background:

• Induced pluripotent stem cells (iPSCs) are crucial for regenerative medicine.
• Identifying specific cell types, like endothelial cells, from iPSC derivatives is essential but often complex.
• Current methods like immunostaining or lineage tracing can be time-consuming and labor-intensive.

Purpose of the Study:

• To develop an automated deep learning method for identifying endothelial cells derived from iPSCs.
• To enable cell identification based solely on morphological features from phase-contrast images.
• To eliminate the need for immunostaining or lineage tracing in endothelial cell identification.

Main Methods:

• Convolutional neural networks (CNNs) were employed for image analysis.
• Networks were trained to predict endothelial cell presence based on morphology.
• Model performance was validated against CD31 immunofluorescence staining.
• Iterative optimization of method parameters enhanced prediction accuracy.

Main Results:

• The CNN model successfully identified endothelial cells using only morphological data.
• Prediction accuracy demonstrated a correlation with CNN network depth and image pixel size.
• K-fold cross-validation confirmed the high performance of the optimized CNNs.
• The automated method achieved high accuracy without requiring specific cell markers.

Conclusions:

• Optimized convolutional neural networks provide a highly effective, automated approach for identifying iPSC-derived endothelial cells.
• Morphological analysis via deep learning offers a label-free and efficient alternative to traditional identification methods.
• This technology has the potential to streamline stem cell differentiation protocols and downstream applications.