Objective : We present the system design and optimization of the ProVision prostate-dedicated time-offlight, depth-of-interaction, dual-panel PET scanner. Dual-panel PET geometries offer compact and open configuration suitable for prostate-dedicated imaging; yet, they intrinsically suffer from limitedangle sampling. We evaluated two practical strategies providing the same overall scanning time, to improve transaxial angular coverage: (i) large, static panels (3×8 detector blocks) optimized through inter-block gap patterns (8 candidate geometries), and (ii) smaller movable panels (3×4 blocks) optimized via multi-step motion protocols (4 candidate protocols).
Approach : Candidate designs were screened using two complementary frameworks. First, a fast, reconstruction-independent approach proposed to quantify voxelwise sampling characteristics, capturing both the magnitude and isotropy of angular coverage throughout the imaging volume. Second, conventional Monte Carlo simulations followed by image reconstruction and standardized image quality analysis (using a scaled NEMA NU2 image-quality phantom) to assess the resulting imaging performance. The performance was evaluated using contrast recovery (CR), background variability (Noise), contrast-to-noise ratio (CNR), plus quantification of image elongation through the full-width at half-maximum (FWHM) of line profiles passing through the smallest (4.5 mm diameter) spherical insert.
Main results : For static designs, nonuniform gap configuration (smaller at centre, and progressively larger gaps toward the panel edges) were consistently favoured. The top-performing geometry achieved the best overall balance, with +4.3% CR, -2.4% noise, and reduced elongation (7.2 mm vs 8.0 mm FWHM) relative to the minimum-gap reference geometry. For dynamic designs, more complex motion protocols led to improved performance; the best protocol achieved modest gains (+1.6% CR, -3.6% noise) and the smallest elongation among protocols (FWHM 6.3 mm vs 7.4 mm for the reference protocol). Comparing the best of static and dynamic configurations, the moving-panel design improved CR (55.9% vs 53.4%) and elongation (6.3 mm vs 7.2 mm) at the cost of substantially increased noise (8.2% vs 5.7%; +44%), highlighting a clear performance-noise-complexity trade-off. Rankings derived from the proposed reconstructionindependent framework showed overall consistency with those obtained from Monte Carlo simulations, leading to the same winning candidates.
Significance : The two evaluation frameworks should be viewed as complementary rather than directly comparable, each offering distinct and valuable insights at different stages of system design and optimization. This study provides a structured basis for examining performance trade-offs, practical design constraints, and methodological implications associated with static and dynamic dual-panel PET configurations for prostate imaging.
