Content-aware frame interpolation (CAFI): deep learning-based temporal super-resolution for fast bioimaging.

Area of Science:

• Biophysics
• Cell Biology
• Microscopy

Background:

• High-resolution microscopy enables 3D and time-lapse imaging of cellular processes.
• Observing rapid cellular dynamics is limited by photobleaching and phototoxicity.

Purpose of the Study:

• To implement and evaluate deep learning-based content-aware frame interpolation (CAFI) networks for improving temporal resolution in microscopy.
• To assess CAFI's ability to predict intermediate frames accurately and understand biological motion context.

Main Methods:

• Implementation of two CAFI deep learning networks: Zooming SlowMo and Depth-Aware Video Frame Interpolation.
• Benchmarking CAFI performance across 12 diverse datasets from four microscopy modalities.
• Application of CAFI for single-particle tracking and nuclear segmentation.

Main Results:

• CAFI accurately predicts intermediate images, enhancing temporal resolution of image series post-acquisition.
• CAFI demonstrates superior performance compared to standard interpolation methods by understanding motion context.
• Validated CAFI's utility in single-particle tracking and nuclear segmentation tasks.

Conclusions:

• CAFI significantly improves temporal resolution in microscopy imaging, facilitating the study of fast cellular dynamics.
• CAFI offers a potential solution to reduce light exposure and phototoxicity, enabling improved long-term live-cell imaging.
• The developed CAFI models and data are accessible via the ZeroCostDL4Mic platform for broader research application.