Real space iterative reconstruction for vector tomography (RESIRE-V)

Minh Phạm, Xingyuan Lu, Arjun Rana, Stanley Osher, Jianwei Miao

TL;DR

This paper introduces RESIRE-V, a novel iterative algorithm for reconstructing 3D vector fields (like magnetic fields) from tomographic data. By leveraging a 3D support constraint and a refined iterative process, RESIRE-V achieves accurate vector tomography reconstructions, even with limited projection data and missing wedges, outperforming previous methods.

What Problem Does This Paper Address?

Traditional tomography excels at reconstructing scalar quantities (e.g., density, electron potential) in 3D. However, many critical scientific applications, particularly in materials science and physics, require the reconstruction of vector fields – quantities that have both magnitude and direction, such as electric and magnetic fields. Existing vector tomography methods often struggle with incomplete data, noise, and the inherent coupling between vector components, leading to less accurate or robust reconstructions. This paper aims to bridge this gap by developing a more accurate and robust method for 3D vector tomography.

Methods

The core of RESIRE-V is an iterative, real-space reconstruction algorithm. It builds upon the established RESIRE (REal Space Iterative REconstruction) framework for scalar tomography. Key methodological advancements include:

  • Vector Field Formulation: The algorithm is specifically designed to handle the three components (Mx, My, Mz) of a vector field simultaneously. It formulates the problem as minimizing an error metric based on the difference between measured and calculated projections, incorporating the vector nature of the field.
  • Forward and Backward Projection: Each iteration involves a forward projection step using the Radon transform and a backward projection step using its transpose. This process is applied to the vector field components.
  • 3D Support Constraint: A crucial innovation is the incorporation of a 3D support constraint. This constraint defines the spatial region where the vector field is expected to exist (e.g., the boundaries of a magnetic material), effectively separating the magnetic signal from non-magnetic background and significantly improving reconstruction accuracy, especially in noisy or incomplete data scenarios.
  • Iterative Refinement: The algorithm iteratively refines the reconstruction by updating the vector field based on the calculated gradients. It also includes an optional iterative angular refinement procedure to improve the accuracy of tilt angles, further enhancing reconstruction quality.
  • Left/Right Polarization Difference: For magnetic vector tomography, the method utilizes the difference between left and right circularly polarized projections to isolate the magnetic contrast, which is then used for reconstruction.

Datasets

The paper utilizes two types of datasets for validation:

  • Simulated Data: A 3D magnetization vector field of a meta-lattice (100x100x100 pixels) was simulated. This dataset included two orthogonal tilt series with missing wedges and Poisson noise to mimic realistic experimental conditions.
  • Experimental Data: The algorithm was applied to experimental data from a ferromagnetic meta-lattice. This dataset consisted of three tilt series, each with left and right polarized projections, acquired using X-ray magnetic circular dichroism.

The authors have made their RESIRE-V MATLAB source codes and the simulated/experimental data freely available on GitHub.

Key Insights and Ideas

  • Support is Crucial for Vector Tomography: The paper strongly emphasizes that a 3D support constraint is not just beneficial but essential for achieving accurate and robust vector tomography, particularly when dealing with limited data or noisy projections. It acts as a powerful regularizer.
  • Z-Component is Easier to Reconstruct: Mathematical analysis and experimental results show that the z-component of the magnetization vector field is generally reconstructed with higher quality than the x and y components when using in-plane rotations. This is due to the way the components are coupled by the rotation matrices.
  • In-plane Rotations Alone Are Insufficient: While multiple in-plane rotations are necessary to decouple the Fourier coefficients of the vector components, they are not sufficient on their own for a unique and accurate reconstruction. Additional constraints (like support) are vital.
  • Topological Defects are Resolved: The reconstructed vector fields clearly reveal the presence of topological magnetic defects (e.g., monopoles with positive and negative charges), demonstrating the method's ability to capture fine structural details.

Key Results

  • High Accuracy in Simulations: For simulated data, RESIRE-V achieved normalized cross-correlations of 94.1% for Mx, 93.8% for My, and 99.1% for Mz, indicating excellent agreement with the ground truth. Fourier Shell Correlation (FSC) plots confirmed high-quality reconstructions across a significant range of spatial frequencies.
  • Experimental Validation: The algorithm successfully reconstructed the 3D magnetization vector field of a ferromagnetic meta-lattice from experimental data, revealing topological defects and showing good agreement with expectations.
  • Robustness to Missing Wedge: The method demonstrates good performance even with missing wedge artifacts, a common challenge in tomography.
  • Improved Reconstruction Quality: RESIRE-V offers a significant improvement over previous vector tomography methods, particularly in scenarios with limited projection data.

Limitations and Future Work

  • Regularizer Choice: The paper discusses the impact of different regularizers (L2 norm, squared L2 norm, squared gradient norm) on reconstruction quality. It notes that the choice of regularizer is critical and depends on the specific data and prior knowledge. The L2 norm regularizer was found to be unsuitable for this specific case.
  • Experimental Feasibility: While side rotations can improve robustness, they are experimentally challenging. The current approach relies primarily on in-plane rotations and the support constraint.
  • Future Directions: The authors suggest that RESIRE-V can be incorporated into various imaging modalities as a general vector tomography method. Further exploration of different regularizers and optimization strategies could lead to even more robust and accurate reconstructions. The availability of the code and data encourages further research and application.

Summary updated

Abstract

Tomography has had an important impact on the physical, biological, and medical sciences. To date, most tomographic applications have been focused on 3D scalar reconstructions. However, in some crucial applications, vector tomography is required to reconstruct 3D vector fields such as the electric and magnetic fields. Over the years, several vector tomography methods have been developed. Here, we present the mathematical foundation and algorithmic implementation of REal Space Iterative REconstruction for Vector tomography, termed RESIRE-V. RESIRE-V uses multiple tilt series of projections and iterates between the projections and a 3D reconstruction. Each iteration consists of a forward step using the Radon transform and a backward step using its transpose, then updates the object via gradient descent. Incorporating with a 3D support constraint, the algorithm iteratively minimizes an error metric, defined as the difference between the measured and calculated projections. The algorithm can also be used to refine the tilt angles and further improve the 3D reconstruction. To validate RESIRE-V, we first apply it to a simulated data set of the 3D magnetization vector field, consisting of two orthogonal tilt series, each with a missing wedge. Our quantitative analysis shows that the three components of the reconstructed magnetization vector field agree well with the ground-truth counterparts. We then use RESIRE-V to reconstruct the 3D magnetization vector field of a ferromagnetic meta-lattice consisting of three tilt series. Our 3D vector reconstruction reveals the existence of topological magnetic defects with positive and negative charges. We expect that RESIRE-V can be incorporated into different imaging modalities as a general vector tomography method. To make the algorithm accessible to a broad user community, we have made our RESIRE-V MATLAB source codes and the data freely available at https://github.com/minhpham0309/RESIRE-V .
BibTeX
@inproceedings{pham2024real,
  author = {Minh Phạm and Xingyuan Lu and Arjun Rana and Stanley Osher and Jianwei Miao},
  title = {{Real space iterative reconstruction for vector tomography (RESIRE-V)}},
  booktitle = {Scientific},
  year = {2024},
  doi = {10.1038/s41598-024-59140-1},
}
Figure