Volumetric CT with sparse detector arrays (and application to Si-strip photon counters)

Area of Science:

• Medical Imaging
• Computational Imaging
• Detector Physics

Background:

• Novel x-ray medical imaging sensors, including photon counting detectors (PCDs) and large area cameras, often exhibit irregular or sparse sampling patterns.
• Applying these detectors to computed tomography (CT) results in undersampling distinct from traditional sparse angular sampling, leading to image artifacts.

Purpose of the Study:

• To investigate volumetric sampling in CT systems with sparsely sampled detectors using axial and helical scan orbits.
• To evaluate the performance of model-based image reconstruction (MBIR) with spatially varying regularization in mitigating artifacts caused by sparse detector sampling.

Main Methods:

• Introduced volumetric metrics for sampling density and uniformity.
• Employed penalized-likelihood MBIR with a spatially varying penalty to account for detector gaps and homogenize resolution.
• Tested the methodology using simulations and an imaging bench with a Si-strip PCD, employing various scanning trajectories and phantoms (spherical clutter, hand).

Main Results:

• Intermediate helical width scan trajectories (approx. 10 mm longitudinal distance per 360° rotation) offered an optimal balance between sampling density and uniformity.
• Spatially varying regularization significantly reduced sampling artifacts, improving minimum Structural Similarity Index (SSIM) by 10% and reducing SSIM dispersion by 40% compared to a constant penalty.
• Phantom images showed a 25% improvement in image uniformity and a 1.8x higher Contrast-to-Noise Ratio (CNR) with the spatially varying penalty.

Conclusions:

• Established a relationship between detector plane sampling, acquisition orbit, reconstructed volume sampling, and resultant image quality.
• Demonstrated the significant benefit of spatially varying regularization in MBIR for CT imaging scenarios with irregular detector sampling patterns.
• Findings support the integration of sparsely sampled Si-strip PCDs into CT imaging systems by effectively addressing sampling-related artifacts.