Noise-matched total-likelihood-based bilateral filter: Experimental feasibility in a benchtop photon-counting CBCT...

Area of Science:

• Medical Imaging
• Image Processing
• Photon-Counting CT

Background:

• Material decomposition in CT imaging generates significant noise in basis and synthesized images.
• Existing likelihood-based bilateral filters struggle with image contrast sensitivity and noise texture.
• Optimal combination of filtered basis images for CT synthesis remains a challenge.

Purpose of the Study:

• To introduce a novel noise reduction technique for photon-counting CT (PCCT) using total likelihood and a noise-matched condition.
• To improve noise reduction effectiveness and image quality in synthesized CT images.
• To address limitations of existing filters in terms of contrast sensitivity and noise texture.

Main Methods:

• Demonstrated feasibility on a benchtop PCCT system.
• Employed a calibration process for forward modeling.
• Utilized maximum likelihood (ML)-based material decomposition followed by a total-likelihood-based filter for noise reduction.
• Synthesized CT images using a noise-matched condition.
• Compared the proposed method against conventional neighborhood filters and statistical iterative reconstruction.

Main Results:

• The proposed method demonstrated superior spatial resolution preservation, particularly in low-contrast regions, compared to other techniques.
• Analysis using a test phantom and chicken leg experiments showed improved fine structures and background textures in denoised images.
• The method exhibited enhanced capabilities in noise power spectrum analysis and superior noise reduction properties.

Conclusions:

• The developed method is effective and computationally efficient for noise reduction in PCCT.
• It offers potential as a replacement for conventional iterative edge-preserved regularization approaches.
• The technique shows promise for improving the quality of synthesized CT images.