Related Experiment Video
Updated: Jul 26, 2025
07:20
A Tissue Clearing Method for Neuronal Imaging from Mesoscopic to Microscopic Scales
Published on: May 10, 2022
3.0K
De-scattering Deep Neural Network Enables Fast Imaging of Spines through Scattering Media by Temporal Focusing
Zhun Wei1,2, Josiah R Boivin3, Yi Xue4
1Center for Advanced Imaging, Faculty of Arts and Sciences, Harvard University, Cambridge, MA 02138, USA.
Research Square
|June 19, 2023
Summary
Researchers developed DeScatterNet, a deep neural network, to enhance temporal focusing microscopy (TFM) images. This AI approach significantly improves deep-tissue imaging quality for neuroscience research.
Area of Science:
- Neuroscience
- Biomedical Imaging
- Artificial Intelligence
Background:
- Point-scanning two-photon microscopy (PSTPM) is the gold standard for in vivo deep-tissue imaging but is slow.
- Temporal focusing microscopy (TFM) offers faster imaging but suffers from poor image quality due to scattered photons.
- Existing methods struggle to resolve fine structures like dendritic spines in TFM images.
Approach:
- Developed DeScatterNet, a 3D convolutional neural network (CNN), to map TFM to PSTPM modalities.
- Trained the network to de-scatter fluorescence emission photons, improving TFM image quality.
- Applied the approach to in vivo imaging of dendritic spines in the mouse visual cortex.
Key Points:
- DeScatterNet significantly enhances TFM image quality, recovering details obscured by scattering.
- The AI-powered TFM achieves imaging speeds one to two orders of magnitude faster than PSTPM.
- The method successfully visualizes dendritic spines in vivo with high fidelity.
Conclusions:
- DeScatterNet enables fast, high-quality deep-tissue imaging by improving TFM performance.
- This approach overcomes TFM's limitations, offering a powerful tool for neuroscience.
- The technology has potential applications in other speed-demanding deep-tissue imaging, including in vivo voltage imaging.
