Spatial Resolution Properties in Penalized-Likelihood Reconstruction of Blurred Tomographic Data

The understanding of the image properties produced by a medical imaging system is critical to effective interpretation of the diagnostic images they produce. Quantitative analysis of image properties like spatial resolution can be complex in computed tomography when advanced model-based reconstruction methods like penalized-likelihood estimation are used since spatial resolution is dependent on patient anatomy, x-ray exposures, and location in the field of view. Previous work [1] has derived mathematical expressions for the local impulse response which quantifies spatial resolution and permits prospective analysis, control, and optimization of the imaging chain including tuning of the reconstruction algorithm. While previous analysis is appropriate for many diagnostic systems, it relies on an idealized system model that ignores any projection blur. Newer devices including cone-beam CT systems that utilize flat-panel detectors can experience significant system blur both due to light spread in the scintillator and due to extended x-ray focal spots. This work introduces a derivation of the local impulse response for penalized-likelihood reconstruction where projections are subject to system blur. We investigate the new local impulse response expression in both simulation studies and in physical test-bench experiments, demonstrating the accuracy of this new resolution predictor. IndexTerms–Local impulse response, penalized likelihood reconstruction, flat panel cone-beam CT