A Monte Carlo investigation of dual-planar circular-orbit cone-beam SPECT

Abstract

We use Monte Carlo simulations to study the imaging properties of a design for a dual-planar cone-beam (DPCB) single-photon emission computed tomography (SPECT) system. A dual-planar system uses a dual-camera SPECT system and two cone-beam collimators with foci in different axial planes to increase the effective axial field of view (FOV). We simulated nearly noise-free projection data from a computerized brain phantom and a phantom consisting of a series of points. Four configurations were simulated: parallel-beam low-energy high-resolution (LEHR) as a standard for comparison and DPCB at three radii of rotation (ROR) corresponding to the smallest, average and largest ROR required to clear patients' shoulders based on ergonomic data. We compared global measures of average resolution and total acquired counts for the four configurations. We also estimated local spatial frequency response for reconstructions of point sources. Finally, we estimated a local noise power spectrum by simulating 1000 noise realizations of the brain phantom and estimating a local noise covariance at selected points. The noise power spectra were used to estimate spectral signal to noise ratio (SNR) for each configuration. The resolution in the reconstructed image space ranges from 7.2 mm full-width at half-maximum (FWHM) at the minimum ROR to 9.4 mm FWHM at the maximum ROR. The efficiency is inversely related, ranging from 1.5 times that of parallel LEHR at minimum ROR to 2.5 times that of LEHR at maximum ROR. Estimates of system frequency response roughly correspond to the global resolution estimates, but the cone-beam techniques exhibit an unusual secondary peak in the axial-direction response. Estimates of spectral SNR show that the cone-beam configurations almost always result in higher SNR at all spatial frequencies regardless of ROR. The very largest ROR may be an exception. A larger ROR results in significantly higher SNR for low spatial frequencies with small reductions in SNR for mid-range frequencies. We conclude that the DPCB design allows significant improvements in both resolution and noise as compared to conventional parallel designs and that optimizing the ROR for the cone-beam system may improve the performance of certain imaging tasks.