Deep Learning Model for Predicting Airway Organoid Differentiation.

Area of Science:

• Biotechnology
• Medical Imaging
• Artificial Intelligence

Background:

• Organoids offer advantages over traditional in vitro and in vivo models but vary in self-organization.
• Confirming organoid tissue similarity typically requires destructive methods like immunofluorescence staining.

Purpose of the Study:

• To develop a non-destructive method for selecting organoids with high tissue-specific similarity.
• To utilize deep learning and bright-field imaging for organoid selection, eliminating the need for staining.

Main Methods:

• Identified four key biomarkers from airway organoid RNA.
• Employed a deep learning approach, specifically a convolutional neural network, for image-based biomarker expression prediction.
• Acquired non-destructive bright-field images of organoids.

Main Results:

• Successfully predicted airway organoid-specific marker expression using only bright-field images.
• Confirmed organoid differentiation through immunofluorescence staining post-prediction, validating the imaging-based approach.
• Demonstrated the ability to select organoids with high tissue similarity without staining.

Conclusions:

• Deep learning combined with non-destructive imaging can effectively distinguish organoids with high human tissue similarity.
• This approach holds significant potential for advancing disease research and drug screening applications.
• The method offers a more efficient and less destructive alternative to traditional organoid validation techniques.