Unsupervised deep learning enables blur-free resolution enhancement in two-photon microscopy

Area of Science:

• Neuroscience
• Biomedical Imaging
• Computational Biology

Background:

• Two-photon microscopy allows deep tissue imaging but suffers from blur and low resolution.
• Quantitative 3D analysis is challenging due to these limitations.

Purpose of the Study:

• Introduce the Two-photon microscopy Image Enhancement Network (TENET).
• Improve image fidelity and segmentation for quantitative analysis of two-photon microscopy data.

Main Methods:

• TENET is a fully unsupervised framework for simultaneous deblurring, resolution enhancement (up to 12x), and semantic segmentation.
• It uses a physics-informed blur-generation module with a trainable neural implicit point spread function (PSF).

Main Results:

• TENET outperforms RLTV, CARE, and Neuroclear on synthetic, bead, and in vivo microglia datasets.
• Achieved superior image fidelity and segmentation accuracy.
• Enabled automated 3D morphometry of microglia-tumor interactions in time-lapse data.

Conclusions:

• TENET effectively enhances two-photon microscopy volumes, producing high-fidelity reconstructions.
• Streamlines downstream analysis and expands capabilities for deep-tissue live imaging.