Adaptive noise reduction for dual-energy x-ray imaging based on spatial variations in beam attenuation

Area of Science:

• Medical Imaging
• Image Processing
• Radiology

Background:

• Dual-energy (DE) imaging offers improved tissue differentiation compared to single-energy imaging.
• Existing DE algorithms can be limited by image noise, affecting diagnostic accuracy.
• Patient-specific pixel-based dual-energy (PP-DE) algorithms aim to personalize image reconstruction.

Purpose of the Study:

• To enhance the previously developed patient-specific pixel-based dual-energy (PP-DE) algorithm.
• To introduce an adaptive anti-correlated noise reduction (ACNR) method for DE imaging.
• To reduce image noise and improve contrast-to-noise (CNR) and signal-to-noise (SNR) ratios in DE images.

Main Methods:

• Developed theoretical models for CNR and SNR as functions of weighting factors for DE bone and soft tissue cancellation.
• Acquired high-energy (HE) and low-energy (LE) images using a clinical ExacTrac imaging system with step and anthropomorphic phantoms.
• Optimized weighting factors for material cancellation and adaptive ACNR by maximizing SNR and validating with experimental data.

Main Results:

• Theoretical models accurately predicted CNR and SNR characteristics, validating experimental findings.
• The adaptive ACNR algorithm demonstrated improved image quality compared to conventional ACNR, especially in regions with high attenuation (e.g., bone).
• Optimized weighting factors for material cancellation were found to be independent of noise cancellation weighting factors.

Conclusions:

• The adaptive ACNR weighting factors are dependent on material thicknesses and varying beam attenuation.
• The developed adaptive ACNR DE algorithm spatially varies weighting factors to enhance image quality.
• This adaptive ACNR method complements existing PP-DE algorithms for superior DE image quality.