Recursive Deep Prior Video: A super resolution algorithm for time-lapse microscopy of organ-on-chip experiments

Area of Science:

• Biomedical Engineering
• Cell Biology
• Microscopy

Background:

• Organ-on-chips (OOCs) utilize Time-Lapse Microscopy (TLM) to observe cell movement.
• High spatial resolution is crucial for analyzing cell dynamics in TLM videos.
• Limitations in physical resources and cost restrict the acquisition of high-resolution videos.

Purpose of the Study:

• To develop a deep learning algorithm for Time-Lapse Microscopy (TLM) video super-resolution.
• To enhance the spatial resolution of videos from organ-on-chip experiments without requiring training data.

Main Methods:

• A novel deep learning algorithm based on Deep Image Prior (DIP) is proposed.
• The method employs a recursive updating rule for DIP network weights and an early stopping criterion.
• The DIP loss function incorporates two Total Variation-based regularization terms.

Main Results:

• The algorithm was validated on both synthetic and real organ-on-chip experimental videos.
• The Recursive Deep Prior Video method demonstrated outstanding performance compared to state-of-the-art trained super-resolution algorithms.

Conclusions:

• The developed algorithm effectively addresses the challenge of low-resolution videos in OOC experiments.
• This method offers a powerful tool for detailed analysis of cell dynamics and interactions in biological research.