Impact of Noise‑Suppression Strength and Exposure Reduction on Contrast–Detail In Image‑Guided Radiation Therapy with High-Performance CBCT
Abstract
Purpose
To evaluate how varying the noise-suppression level and reducing the imaging exposure affects contrast-detail performance and other image quality metrics in image-guided radiation therapy.
Methods
A CATPhan 604 image-quality phantom was scanned on a C-arm linac equipped with high-performance CBCT using three imaging modes (head, thorax, and pelvis). For each mode, factory-default acquisition and reconstruction settings were tested as well as three reduced exposure settings and four additional iterative reconstruction noise-suppression levels. Image-quality metrics, including spatial resolution, noise, and low contrast were assessed to determine the potential for CBCT dose reduction. Contrast-detail (CD) curves were generated, and the smallest detectable object size was evaluated using the Rose criterion at 1% and 0.5% contrast.
Results
Spatial resolution, quantified by MTF (10%), was unaffected by changes to the noise-suppression level but decreased by up to 8% when exposure was lowered to 25% of the default settings. Noise decreased linearly with the strength of the noise suppression setting. The relationship between squared noise and exposure followed a power-law, showing diminishing improvements at higher exposures. The CD curves showed that reducing exposure to 50% of the default values increased the smallest detectable low-contrast object size by only 1-2 mm. Increasing the noise-suppression by one level improved detectable low-contrast object size by a comparable amount (1-3 mm).
Conclusion
This phantom study suggests that imaging dose in image-guided radiation therapy using a high-performance CBCT system may be reduced by up to 50% from factory-default settings when combined with a moderate increase in noise-suppression strength without substantially compromising low-contrast detectability. Validation with clinical image data is needed to confirm these findings.