Electron Tomography
Electron tomography reconstructs 3D structure from a tilt series of 2D projections acquired as the specimen is rotated (+/-60-70 deg, 1-2 deg increments). The missing wedge of angular coverage causes elongation artifacts along the beam direction. Alignment of the tilt series (using fiducial gold markers or cross-correlation) is critical. Reconstruction uses WBP, SIRT, or compressed sensing methods with TV priors to mitigate missing-wedge artifacts.
Projection Tilt Series
Poisson
sirt
DIRECT_ELECTRON_DETECTOR
Forward-Model Signal Chain
Each primitive represents a physical operation in the measurement process. Arrows show signal flow left to right.
R(θ) → P(e⁻) → Π(proj) → D(g, η₁)
Benchmark Variants & Leaderboards
Electron Tomo
Electron Tomography
R(θ) → P(e⁻) → Π(proj) → D(g, η₁)
Standard Leaderboard (Top 10)
| # | Method | Score | PSNR (dB) | SSIM | Trust | Source |
|---|---|---|---|---|---|---|
| 🥇 | DiffET | 0.878 | 39.1 | 0.952 | ✓ Certified | Gao et al. 2024 |
| 🥈 | PhysET | 0.848 | 37.7 | 0.940 | ✓ Certified | Chen et al. 2024 |
| 🥉 | SwinET | 0.821 | 36.4 | 0.929 | ✓ Certified | Wang et al. 2023 |
| 4 | TransET | 0.785 | 34.8 | 0.910 | ✓ Certified | Li et al. 2022 |
| 5 | IsoNet | 0.721 | 32.1 | 0.871 | ✓ Certified | Liu et al. 2021 |
| 6 | DnCNN-ET | 0.653 | 29.3 | 0.829 | ✓ Certified | Buchholz et al. 2019 |
| 7 | CS-ET | 0.575 | 26.4 | 0.769 | ✓ Certified | Leary et al. 2013 |
| 8 | SIRT-ET | 0.505 | 23.6 | 0.724 | ✓ Certified | Gilbert 1972 |
| 9 | WBP-ET | 0.437 | 20.9 | 0.678 | ✓ Certified | Radermacher et al. 1987 |
Mismatch Parameters (3) click to expand
| Name | Symbol | Description | Nominal | Perturbed |
|---|---|---|---|---|
| tilt_angle | Δθ | Tilt angle error (deg) | 0 | 0.5 |
| tilt_axis | Δφ | Tilt axis misalignment (deg) | 0 | 0.3 |
| defocus_gradient | Δf' | Defocus gradient (nm/μm) | 0 | 10 |
Reconstruction Triad Diagnostics
The three diagnostic gates (G1, G2, G3) characterize how reconstruction quality degrades under different error sources. Each bar shows the relative attribution.
Model: projection tilt series — Mismatch modes: missing wedge, alignment error, specimen shrinkage, beam damage, focus gradient
Noise: poisson — Typical SNR: 3.0–20.0 dB
Requires: tilt angles, tilt axis orientation, fiducial positions, magnification per tilt
Modality Deep Dive
Principle
Electron tomography reconstructs a 3-D volume from a tilt series of 2-D TEM or STEM projections acquired at different specimen tilts (typically ±60-70°). The Radon transform (or its generalization) relates the projections to the 3-D structure. The limited tilt range causes a 'missing wedge' artifact — elongation in the beam direction — which must be addressed by regularization or dual-axis acquisition.
How to Build the System
Use a TEM/STEM with a high-tilt specimen holder (±70-80°). Acquire images at tilt increments of 1-2° across the full range. For STEM tomography, HAADF signal provides monotonic contrast (no CTF complications). Include gold nanoparticles as fiducial markers for alignment. Automated acquisition software (SerialEM, Tomography by Thermo Fisher) controls stage tilt, focus tracking, and image acquisition.
Common Reconstruction Algorithms
- Weighted back-projection (WBP)
- SIRT / SART (Simultaneous Iterative Reconstruction Techniques)
- GENFIRE (GENeralized Fourier Iterative REconstruction)
- Compressed sensing tomography for missing-wedge artifact reduction
- Deep-learning tomographic reconstruction (TomoGAN, DeepRecon)
Common Mistakes
- Poor tilt-series alignment causing blurring in the reconstruction
- Missing wedge artifacts not addressed, distorting features along the beam axis
- Specimen drift or deformation during the tilt series (especially for biological specimens)
- Dose damage accumulating through the tilt series degrading later images
- Inaccurate tilt angles due to stage mechanical backlash
How to Avoid Mistakes
- Align tilt series carefully using fiducial markers; refine with cross-correlation
- Use dual-axis tomography or compressed-sensing reconstruction to fill the missing wedge
- Apply autofocus and drift tracking at each tilt; use cryo-conditions for biology
- Distribute dose evenly; start at high tilts where damage impact is greatest
- Calibrate stage tilt angle accuracy; use Saxton scheme (non-linear tilt increments)
Forward-Model Mismatch Cases
- The widefield fallback processes only 2D (64,64) images, but electron tomography acquires a tilt series — projections at multiple angles through the 3D specimen volume, with output shape (n_tilts, H, W)
- The missing wedge problem (limited tilt range, typically +/- 70 degrees) is specific to electron tomography and cannot be modeled by the widefield operator — reconstructions without accounting for missing data have severe elongation artifacts
How to Correct the Mismatch
- Use the electron tomography operator that generates projection images at each tilt angle via the Radon transform applied to the 3D specimen density, including the limited tilt range constraint
- Reconstruct using weighted back-projection (WBP), SIRT, or compressed-sensing methods that account for the missing wedge and alignment errors between tilt images
Experimental Setup
Thermo Fisher Titan Krios G4 / JEOL JEM-2200FS
200
[-70, 70]
2.0
71
HAADF-STEM / Gatan K3
0.71
39000
SIRT / WBP
Signal Chain Diagram
Key References
- Frank, 'Electron Tomography', Springer (2006)
- Midgley & Dunin-Borkowski, 'Electron tomography and holography in materials science', Nature Materials 8, 271 (2009)
Canonical Datasets
- EMPIAR cryo-ET tilt series (e.g., EMPIAR-10045)
- ETDB (Electron Tomography Database, Caltech)