Ultrasparse View X-ray Computed Tomography for 4D Imaging.

Area of Science:

• Materials Science
• Imaging Technology
• Computational Science

Background:

• X-ray computed tomography (CT) is a vital noninvasive technique for 3D material imaging.
• Traditional 3D CT requires numerous images, limiting temporal resolution for dynamic studies (4D imaging).
• Sparse-view imaging presents a challenge in maintaining image fidelity.

Purpose of the Study:

• To introduce Tomo-NeRF, a novel sparse-view imaging method for high-fidelity 3D CT reconstruction.
• To evaluate Tomo-NeRF's performance using limited 2D radiographic images.
• To assess the potential of Tomo-NeRF for improving 4D imaging applications.

Main Methods:

• Developed Tomo-NeRF, a neural radiance field-based approach for sparse-view CT reconstruction.
• Acquired and utilized experimental 2D and 3D X-ray images for validation.
• Tested reconstruction capabilities in two-view, four-view, and six-view scenarios.

Main Results:

• Tomo-NeRF successfully reconstructed high-fidelity 3D images from <10 2D radiographic images.
• Experimental validation showed high structural similarity (0.9971-0.9975) and voxel-wise accuracy (81.83-89.59%).
• Reconstruction accuracy on experimental data was comparable to synthetic data results.

Conclusions:

• Tomo-NeRF enables high-quality 3D CT reconstruction from significantly reduced 2D image sets.
• This method offers a promising solution for enhancing temporal resolution in 4D imaging.
• Tomo-NeRF advances the field of sparse-view CT imaging for materials and engineered components.