TL;DR
This paper introduces FDKHSCW, a novel cone-beam CT reconstruction algorithm that improves temporal resolution by using a half-scan acquisition. It addresses the attenuation coefficient drop artifact common in existing methods by employing a new heuristic weighting scheme that accounts for cone beam geometry, leading to more accurate reconstructions without sacrificing spatial resolution.
What Problem Does This Paper Address?
Cone Beam Computed Tomography (CBCT) is a powerful imaging technique, but acquiring a full 360-degree scan can be time-consuming, limiting its use in applications requiring high temporal resolution, such as dynamic imaging. Half-scan methods aim to reduce acquisition time by scanning only a portion of the full circle. However, traditional half-scan algorithms, particularly those based on the Feldkamp-Davis-Kress (FDK) algorithm, suffer from an "attenuation coefficient drop" artifact. This artifact manifests as a decrease in reconstructed density for objects further from the central scanning plane, especially when the cone angle (the angle of the X-ray beam in the axial direction) is large. This phenomenon arises because the redundancy in projection data, which is crucial for accurate reconstruction, diminishes for data acquired away from the central scanning plane in a half-scan. Existing methods either don't fully correct this drop or compromise spatial resolution.
Methods
The core of this paper is the development of a new weighting scheme for the FDK half-scan algorithm, termed FDKHSCW. The standard FDK algorithm projects off-scanning plane data onto the scanning plane and applies a 2D fan beam reconstruction. For half-scans, a weighting function is applied to the projection data before filtering and back-projection. The traditional FDK half-scan with Parker's weighting function (FDKHSFW) applies a weighting function derived from fan beam geometry uniformly across all detector rows.
FDKHSCW introduces a heuristic cone beam geometry-dependent weighting scheme. The key idea is that the redundancy of projection data is not uniform across all detector rows when viewed in the context of the cone beam geometry. Specifically, data further away from the central scanning plane exhibits less redundancy. Therefore, FDKHSCW proposes a weighting function that is adjusted based on the cone beam geometry. This is achieved by introducing a "cone-weighting angle" ($\beta'$) that depends on the position of the projection row in the Z-direction (axial direction), implicitly incorporating the cone angle information. This new weighting function is applied to the projection data before filtering and back-projection. The authors also note that this scheme can be applied to algorithms that don't require rebinning, making it flexible.
Datasets
The authors used a 3D Shepp-Logan phantom for computer simulations. This is a standard mathematical phantom used for evaluating CT reconstruction algorithms. They also simulated quantum noise contaminated data to test the algorithm's robustness.
Key Insights and Ideas
- Redundancy is not uniform in half-scans: The paper highlights that the assumption of uniform redundancy in projection data for half-scans, which underlies some existing methods, breaks down when dealing with cone beam geometry and larger cone angles.
- Geometry-dependent weighting is key: The core insight is that a weighting function tailored to the specific cone beam geometry, rather than a generic fan beam approach, is necessary to correct for the attenuation coefficient drop.
- Balancing temporal resolution and accuracy: The proposed FDKHSCW scheme successfully improves temporal resolution (by using a half-scan) while simultaneously mitigating artifacts that plague simpler half-scan approaches, offering a better trade-off.
- Maintaining spatial resolution: A significant advantage is that this improved accuracy in attenuation correction is achieved without compromising the spatial resolution of the reconstructed images, which is a common challenge in artifact reduction techniques.
Key Results
The paper demonstrates through computer simulations on a 3D Shepp-Logan phantom that FDKHSCW effectively corrects the attenuation coefficient drop artifact observed in both FDK full-scan (FDKFS) and traditional FDK half-scan with Parker's weighting (FDKHSFW). The authors state that the largest compensation achieved was within 0.03 in terms of attenuation coefficient for low-contrast objects. They also show that FDKHSCW maintains the same spatial resolution as FDKHSFW. The simulations also showed that FDKHSCW can achieve comparable image quality with significantly lower total exposure levels compared to FDKFS, due to the reduced scan time.
Limitations and Future Work
The authors acknowledge that while FDKHSCW shows improvement, it might not be as promising for high-contrast phantoms (like a Defrise disk) as it is for low-contrast objects. They also mention that other modified FDK methods exist that also correct attenuation drop but require parallel rebinning, which FDKHSCW avoids.
Future work suggested includes combining FDKHSCW with other artifact reduction techniques, such as those that address data inconsistency between direct and conjugate rays. They also propose exploring an extended half-scan scheme (larger than 180 + 2$\Delta$) using this weighting function for improved noise characteristics, and investigating its performance on high-contrast phantoms. The potential for real-time monitoring in dynamic imaging is also highlighted as a benefit of this approach.
Summary updated