Compensating CBCT Motion Artifacts with Any 2D Generative Model
This paper presents a novel approach to mitigate motion artifacts in industrial Cone-Beam Computed Tomography (CBCT) caused by detector or X-ray source jitter due to mechanical vibration. Leveraging two-dimensional (2D) generative models while ensuring consistency between adjacent CBCT slices, our...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | deu |
| Published: |
NDT.net
2025-02-01
|
| Series: | e-Journal of Nondestructive Testing |
| Online Access: | https://www.ndt.net/search/docs.php3?id=30726 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This paper presents a novel approach to mitigate motion artifacts in industrial Cone-Beam Computed Tomography (CBCT) caused by detector or X-ray source jitter due to mechanical vibration. Leveraging two-dimensional (2D) generative models while ensuring consistency between adjacent CBCT slices, our method addresses the limitations of traditional deep learning approaches that process each 2D slice of a three-dimensional (3D) volume independently. While traditional deep learning approaches may adequately handle artifacts in the axial view, they often struggle with consistency problems in sagittal and coronal views, resulting in insufficient 3D coherence across the entire volume. Our approach integrates a 2D generative model for artifact reduction with a structure-aware regularization strategy, specifically employing a gradient-based smoothness constraint along the through-plane direction to maintain smooth transitions between slices and preserve the overall 3D structure. This model-independent framework accommodates various 2D architectures while effectively addressing the limitations of treating 3D volumes as independent 2D slices. Experimental results on simulated CBCT datasets demonstrate significant improvements in image quality metrics and artifact reduction, offering a promising solution for enhancing 3D CT reconstruction in industrial applications.
|
|---|---|
| ISSN: | 1435-4934 |