Advanced 3D Tomographic Reconstruction Techniques in Electron Microscopy
Event details
| Date | 26.06.2023 |
| Hour | 13:30 › 15:30 |
| Speaker | Alexandre De Skowronski |
| Location | |
| Category | Conferences - Seminars |
| Event Language | English |
EDIC candidacy exam
Exam president: Prof. Pascal Frossard
Thesis advisor: Prof. Pascal Fua
Co-examiner: Prof. Cécile Hébert
Abstract
3D reconstruction from electron microscopy (EM) images is a central topic in computational imaging. Current reconstruction methods, borrowed from other imaging modalities, are limited in handling low image counts and noise corruption. Recent attempts using neural implicit representations and learning-based techniques are discussed. The research proposal focuses on incorporating neural implicit functions for memory-effective reconstructions and exploring learning-based regularization to enhance denoising. Specific priors tailored to different sample types are considered to reduce the number of required images. The abstract emphasizes the need for contrast-agnostic models and addresses the limitations of existing systems.
Background papers
1) J. R. Ramirez, P. Rautek, C. Bohak, O. Strnad, Z. Zhang, S. Li, I. Viola, and W. Heidrich, “Gpu accelerated 3d tomographic reconstruction and visualization from noisy electron microscopy tilt-series” IEEE Transactions on Visualization and Computer Graphics, pp. 1–15, 2022. (preprint) [LINK]
2) H. Kniesel, T. Ropinski, T. Bergner, K. Shaga Devan, C. Read, P. Walther, T. Ritschel, and P. Hermosilla, “Clean implicit 3d structure from noisy 2d stem images” in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2022. [LINK]
3) A. Levy, F. Poitevin, J. Martel, Y. Nashed, A. Peck, N. Miolane, D. Ratner, M. Dunne, and G. Wetzstein, “Cryoai: Amortized inference of poses for ab initio reconstruction of 3d molecular volumes from real cryo-em images” ECCV, 2022. [LINK]
Exam president: Prof. Pascal Frossard
Thesis advisor: Prof. Pascal Fua
Co-examiner: Prof. Cécile Hébert
Abstract
3D reconstruction from electron microscopy (EM) images is a central topic in computational imaging. Current reconstruction methods, borrowed from other imaging modalities, are limited in handling low image counts and noise corruption. Recent attempts using neural implicit representations and learning-based techniques are discussed. The research proposal focuses on incorporating neural implicit functions for memory-effective reconstructions and exploring learning-based regularization to enhance denoising. Specific priors tailored to different sample types are considered to reduce the number of required images. The abstract emphasizes the need for contrast-agnostic models and addresses the limitations of existing systems.
Background papers
1) J. R. Ramirez, P. Rautek, C. Bohak, O. Strnad, Z. Zhang, S. Li, I. Viola, and W. Heidrich, “Gpu accelerated 3d tomographic reconstruction and visualization from noisy electron microscopy tilt-series” IEEE Transactions on Visualization and Computer Graphics, pp. 1–15, 2022. (preprint) [LINK]
2) H. Kniesel, T. Ropinski, T. Bergner, K. Shaga Devan, C. Read, P. Walther, T. Ritschel, and P. Hermosilla, “Clean implicit 3d structure from noisy 2d stem images” in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2022. [LINK]
3) A. Levy, F. Poitevin, J. Martel, Y. Nashed, A. Peck, N. Miolane, D. Ratner, M. Dunne, and G. Wetzstein, “Cryoai: Amortized inference of poses for ab initio reconstruction of 3d molecular volumes from real cryo-em images” ECCV, 2022. [LINK]
Practical information
- General public
- Free
Contact
- edic@epfl.ch