Related Experiment Video
Updated: Aug 5, 2026
08:55
Cryo-Electron Tomography Remote Data Collection and Subtomogram Averaging
Published on: July 12, 2022
A general data-driven framework for scalable electron tomography
Han Li1, Wenting Cui1, Huan Lei1
1Institute of Materials Research, Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China.
National Science Review
|July 25, 2026
Summary
This study introduces a new data-driven framework for electron tomography (ET) reconstruction. It overcomes data scarcity by using random projections, enabling high-quality 3D atomic structure determination.
Area of Science:
- Materials Science
- Nanotechnology
- Imaging Physics
Background:
- Electron tomography (ET) is vital for 3D material structure determination.
- Challenges in ET include missing wedge artifacts, high radiation dose, and limited depth-of-field.
- Deep learning shows promise but is hindered by limited ET data availability.
Purpose of the Study:
- To develop a general data-driven electron tomography reconstruction framework.
- To overcome data scarcity using readily available random high-entropy projections.
- To achieve high-quality 3D reconstruction independent of material or resolution.
Main Methods:
- Constructing large-scale datasets from random high-entropy projections.
- Integrating real structural priors and depth-dependent imaging physics.
- Employing a data-driven approach for ET reconstruction.
Main Results:
- Successfully determined the 3D atomic structure of a 13-nm platinum nanoparticle (52,138 atoms).
- Achieved a root-mean-square displacement of 22.6 pm for atomic positions.
- Significantly reduced projection consistency error, expanding the depth-of-field limit.
Conclusions:
- The proposed framework enables high-quality 3D reconstruction in electron tomography.
- It effectively addresses data scarcity and reconstruction challenges.
- This method advances atomic-scale ET for materials science applications.
