A practical guide to deep-learning light-field microscopy for 3D imaging of biological dynamics

Area of Science:

• Biophysics
• Microscopy
• Computational Biology

Background:

• Capturing rapid biological processes in 3D requires advanced imaging techniques.
• Light-field microscopy (LFM) offers potential for dynamic imaging but often requires complex reconstruction.
• Deep learning presents opportunities to enhance LFM capabilities.

Purpose of the Study:

• To present a detailed protocol for deep-learning-enhanced light-field microscopy (VCD-LFM).
• To enable 3D imaging of instantaneous biological processes with high temporal resolution.
• To demonstrate quantitative analysis of biological dynamics using the developed method.

Main Methods:

• Construction of a custom light-field microscope for capturing dynamic optical signals.
• Development and training of a view-channel-depth (VCD) neural network for 3D reconstruction.
• Application of VCD-LFM to image biological samples and analyze neural and cardio-hemodynamic activities.

Main Results:

• Successful implementation of a step-by-step protocol for VCD-LFM.
• Demonstration of instant 3D image reconstruction from single 2D light-field snapshots.
• Quantitative analysis of neural activities and cardio-hemodynamics in model organisms.

Conclusions:

• Deep-learning enhanced LFM provides a powerful tool for high-speed 3D biological imaging.
• The VCD-LFM protocol enables efficient reconstruction and quantitative analysis of dynamic biological events.
• This method advances the study of fast biological processes in vivo.