TL;DR
This paper introduces a novel "Circle Plus Partial Helical" (CH) scan scheme for Cone Beam Breast CT (CBBCT) to overcome artifacts like density drop and geometric distortion inherent in standard circular scans. By adding a sparse partial helical trajectory, the CH scheme significantly improves image quality and uniformity, offering a promising advancement for breast cancer detection without requiring complex hardware modifications.
What Problem Does This Paper Address?
Mammography, while a standard screening tool for breast cancer, suffers from the limitation of structure superimposition, making it difficult to detect small or occult tumors, especially in dense breast tissue. Cone Beam Breast CT (CBBCT) offers a solution by providing 3D density distribution, thus eliminating this superimposition. However, a common CBBCT scanning method, the single circular scan, introduces significant artifacts when using a large cone angle. These artifacts manifest as intensity drops and geometric distortions, particularly in regions further from the circular scanning plane. This makes it challenging to accurately visualize and diagnose subtle abnormalities. The paper aims to address this by proposing a new scanning strategy that enhances image quality by filling in the "missing" data in the reconstruction process.
Methods
The core of this paper is the proposal and evaluation of a new scanning scheme called "Circle Plus Partial Helical" (CH). This scheme builds upon the standard circular scan by adding a supplementary, sparse partial helical trajectory. The rationale is that a single circular scan doesn't provide complete information for exact reconstruction, leaving gaps in the 3D Radon domain. The partial helical scan is designed to sparsely fill these gaps, thereby correcting artifacts without significantly increasing radiation exposure or scan time.
The reconstruction process for the CH scheme is a composite one, combining:
- FDK algorithm: For the initial circular scan data.
- Hui's term: Another component for circular scan data, likely addressing specific aspects of reconstruction.
- Shift-variant FBP: For the partial helical scan data.
For comparison, the authors also investigate a "Circle Plus Straight Line" (CL) scheme, which uses a straight line trajectory to fill the missing Radon data. This CL scheme is designed to achieve exact reconstruction, as per Katsevich's algorithm. The paper details the mathematical formulations for each component of these reconstruction algorithms.
Datasets
The authors used a mathematical breast phantom for their simulations. This phantom was specifically designed to be a large, half-ellipsoid shape with simulated skin and internal structures including:
- Carcinoma spheres of various sizes (1-8 mm).
- Glandular spheres of various sizes (1-8 mm).
- Two low-contrast disk-type objects to specifically test for geometrical distortion around the nipple area.
The phantom was designed to address artifacts arising from single circular scans. No real patient data was used for training or evaluation in this study.
Key Insights and Ideas
- The "missing volume" in the Radon domain is key: The paper elegantly explains how a single circular scan leaves gaps in the 3D Radon transform of the object, leading to artifacts. The size of this "missing volume" is directly related to the severity of artifacts.
- Partial filling is sufficient: A crucial insight is that complete filling of the missing Radon volume might not be necessary for significant artifact reduction. A sparse, supplementary scan can achieve a "certain degree" of correction, balancing image quality with patient dose.
- CH scheme offers a practical compromise: The CH scheme provides a good balance between image quality improvement and practical implementation. It avoids the need for complex mechanical modifications to existing CBBCT systems, unlike some other advanced scanning techniques.
- Helical vs. Straight Line: While the CL scheme can achieve exact reconstruction, the CH scheme offers comparable artifact correction with significantly fewer X-ray shots, making it more practical for clinical use due to reduced patient exposure.
Key Results
- Artifact Reduction: Computer simulations demonstrated that the CH scheme significantly reduces artifacts such as density drop and geometrical distortion compared to a single circular scan (FDK).
- Quantitative Improvement: The Reconstruction Error (RE) for the CH scheme (with 64 helical shots) was 0.70%, comparable to the CL scheme (with 556 line shots) at 0.70%, and a substantial improvement over the single circular scan (MFDK) at 2.1%.
- Image Uniformity: The CH scheme resulted in more uniform reconstructed images, particularly in areas further from the circular scanning plane.
- Sampling Interval Impact: For the partial helical scan, using 32 or 64 X-ray shots provided acceptable reconstructed images. A smaller sampling interval (more shots) led to fewer streak artifacts, but these were still faintly visible.
- Quantum Noise Performance: The CH scheme performed as well as the CL scheme when projection data was contaminated with simulated quantum noise.
Limitations and Future Work
- Not Exact Reconstruction: The paper explicitly states that the CH scheme is not intended to conduct an exact reconstruction. While it significantly corrects artifacts, it doesn't achieve the theoretical completeness of a full scan.
- Streak Artifacts: While reduced, some streak artifacts were still faintly visible in the CH scheme, especially with fewer helical shots.
- Sampling Interval Optimization: The optimal sampling interval for the helical scan needs careful consideration to balance image quality and radiation dose.
- Mechanical Simplicity vs. Exactness: The authors highlight the practical advantage of the CH scheme's ease of operation and lack of complex mechanical requirements compared to the CL scheme. Future work could explore further optimization of the partial helical trajectory and its integration with existing CBBCT systems. The paper suggests that the quality of reconstructed images could improve without too much additional radiation exposure to the patient.
Summary updated