Related Experiment Video
Updated: Jan 18, 2026
Deep-Tissue Three-Photon Fluorescence Microscopy in Intact Mouse and Zebrafish Brain
Published on: January 13, 2022
Hundred-Nanosecond Equivalent Pixel Dwell Time for Deep-Tissue 3D Three-Photon Fluorescence Microscopy via Sparse
Yifei Li1, Runnan Zhang2, Keying Li3
1State Key Laboratory of Extreme Photonics and Instrumentation, Centre For Optical and Electromagnetic Research, College of Optical Science and Engineering, International Research Center For Advanced Photonics, Zhejiang University, Hangzhou, China.
DeepR-SXYZ accelerates deep tissue imaging using three-photon fluorescence microscopy (3PFM) by reconstructing sparse data. This deep learning framework significantly boosts imaging speed and reduces phototoxicity, enabling new biological discoveries.
Area of Science:
- Biophotonics and Imaging Science
- Computational Biology and Machine Learning
Background:
- Three-photon fluorescence microscopy (3PFM) offers deep tissue imaging but faces speed-resolution trade-offs due to low photon flux and prolonged laser exposure.
- Current 3PFM methods are limited by slow imaging speeds and potential phototoxicity, hindering the study of dynamic biological processes in vivo.
Purpose of the Study:
- To develop a deep learning framework, DeepR-SXYZ, for accelerated and high-resolution volumetric imaging in deep tissues using 3PFM.
- To overcome the inherent limitations of 3PFM by enabling sparse data acquisition and reconstruction for improved imaging speed and reduced phototoxicity.
Main Methods:
- DeepR-SXYZ integrates convolutional neural networks (CNNs) with a structure-dynamic attention (SDA)-enhanced transformer for sparse X-Y-Z reconstruction.
- The framework is trained on paired datasets of sparsely acquired low-resolution scans and densely sampled high-resolution counterparts.
- It synergistically captures intra-layer morphological features for X-Y plane reconstruction and inter-layer dynamic variations for Z-axis interpolation.
Main Results:
- DeepR-SXYZ achieves 8.8× acceleration in X-Y plane imaging speed compared to conventional dense sampling.
- The framework demonstrates >60% Z-axis layer recovery, significantly improving imaging throughput.
- Experimental validation on cerebral vasculature and muscle macrophages shows accurate 3D volume reconstruction from sparse data.
Conclusions:
- DeepR-SXYZ establishes a computational paradigm for high-speed, low-phototoxicity 3PFM via sparse X-Y-Z reconstruction.
- The framework effectively balances imaging speed and spatial resolution, enabling large-field 3D imaging and dynamic volumetric tracking of cellular behaviors.
- This approach reveals previously inaccessible spatiotemporal biological processes, advancing in vivo microscopy capabilities.
