Low-dose cone-beam CT via raw counts domain low-signal correction schemes: Performance assessment and task-based...

Area of Science:

• Medical imaging physics
• Image processing
• Computational imaging

Background:

• Low-signal correction (LSC) methods reduce noise in CT images for low-dose scans.
• These methods introduce shift-variant and anisotropic spatial resolution and noise, complicating parameter optimization.
• Developing an effective parameter optimization framework is crucial for LSC methods.

Purpose of the Study:

• To develop a local task-based parameter optimization framework for LSC methods.
• To optimize filter parameter selection for LSC methods, considering spatial resolution and noise properties.
• To enhance CT imaging performance through tailored LSC parameter selection.

Main Methods:

• Utilized adaptive trimmed mean (ATM) and anisotropic diffusion (AD) filters as examples.
• Employed the detectability index (d') with the non-prewhitening (NPW) observer model for optimization.
• Defined high-contrast-high-frequency discrimination tasks and analyzed performance across different spatial locations.

Main Results:

• Optimal LSC parameters depend on the interplay between imaging tasks and image properties.
• Performance was most influenced when task frequencies aligned with spatial resolution loss or noise patterns.
• Parameter optimization showed strong spatial location dependence for both LSC methods.

Conclusions:

• A local task-based detectability framework was successfully developed for LSC parameter optimization.
• The framework accounts for shift-variant and anisotropic properties to maximize CT system performance.
• Optimal LSC parameters are significantly influenced by the spatial location within the image object.