Statistical reconstruction for cone-beam CT with a post-artifact-correction noise model: application to high-quality...

Area of Science:

• Medical Imaging
• Radiology
• Computational Imaging

Background:

• Non-contrast CT is the standard for detecting intracranial hemorrhage (ICH).
• Flat-panel detector cone-beam CT (FPD-CBCT) offers a potential low-cost, mobile alternative for point-of-care diagnosis.
• Current FPD-CBCT systems struggle with low-contrast soft-tissue imaging and artifacts.

Purpose of the Study:

• To develop a novel penalized weighted least-squares (PWLS) image reconstruction method for FPD-CBCT.
• To accurately model noise characteristics after artifact correction (scatter and beam-hardening) in FPD-CBCT.
• To improve soft-tissue contrast and reduce noise for reliable ICH detection using FPD-CBCT.

Main Methods:

• Developed a novel PWLS image reconstruction algorithm with an accurate post-artifact-correction noise model.
• Incorporated modified weights to compensate for noise amplification from scatter and beam-hardening corrections.
• Validated the method using real data from an FPD-CBCT test-bench and an anthropomorphic head phantom simulating intra-parenchymal hemorrhage.

Main Results:

• The proposed PWLS method demonstrated superior noise-resolution tradeoffs compared to filtered backprojection (FBP) and conventional PWLS.
• At 0.50 mm spatial resolution, the contrast-to-noise ratio (CNR) was 11.9 for the proposed PWLS, versus 5.6 for FBP and 9.9 for conventional PWLS.
• Significantly reduced image noise was observed, particularly in challenging regions like the skull base.

Conclusions:

• The novel PWLS reconstruction method with an accurate noise model significantly enhances FPD-CBCT image quality.
• High-fidelity artifact correction combined with statistical reconstruction enables reliable detection of intracranial hemorrhage.
• FPD-CBCT holds promise for mobile, point-of-care diagnosis of acute brain injury.