A Model of PET/CT Scanner Performance as a Function of Axial Extent

Paul Kinahan; Osama Mawlawi; Jorge Uribe; Charles Stearns; John Sunderland

Objectives: There are still basic questions about the impact of the axial extent of the field of view (AFOV) on PET/CT scanner performance metrics. While there are PET/CT systems with different AFOVs, these also have other fundamental system design differences (e.g. variations in scintillator type or thickness, TOF resolution etc.) that confound comparisons between systems. The recent GE Discovery MI (DMI) scanner is uniquely available with 15, 20, and 25 cm AFOV versions, with all other system parameters being essentially constant. We used all three scanner versions to model and evaluate the impact of varying only the AFOV. Methods: We developed a generalized model of scanner performance based on (1) assumptions of constant spatial, energy, and timing resolutionand (2) a mathematical coincidence sensitivity model that predicts an approximately quadratic behavior. To evaluate the generalized model we used measurements to compare performance metrics from the 15, 20, and 25 cm AFOV versions of the DMI scanner, starting with the NEMA NU2 specifications, to determine both basic performance metrics (e.g. prompt coincidence rates, resolution characteristics) and summary metrics (e.g. noise equivalent count rates, contrast-to-noise ratio (CNR)). Results: The assumptions of constant spatial, energy, and timing resolution were confirmed by measured data. In general, spatial resolution metrics did not measurably vary as a function of AFOV. This included the resolution of point sources (reconstructed with filtered back projection) at six locations in the field of view (FOV). In addition, the contrast recovery of spheres in the image quality (IQ) phantom did not vary with AFOV. The only observable difference was a slight increase (±4%) in thevolumetricfull-width-at-tenth-max (FWTM)as a function of AFOV. For the volumetric FWHM, this variation was under ±2%. The sensitivity model was well matched by measured true coincidence sensitivity and measured noise equivalent count (NEC) rate values at both peak and clinically relevant NEC values(0.998 < R2 < 1.000). In addition, the CNR for the IQ phantom increased with AFOV as expected. Using the constraint of equivalent image quality for an example patient scan length of 760 mm, and using a typical bed position overlap of approximately 28% of the AFOV, the 15, 20 and 25 cm versions of the DMI scanner require 28, 16, and 9.6 min total scan times for equivalent image quality. Conclusions: Our findings are consistent with prior simulation studies [1,2], but are based on a simplified analytical modelthat matched the measured results. Both the model and measurements show that the peak and clinically relevant NEC values vary as the square of the axial extent of the scanner. Since other determinants of image quality are constant, for a fixed patient scan extent the total scan time varies inversely with the square of the axial extentof the scannerfor equivalent image quality.

A Model of PET/CT Scanner Performance as a Function of Axial Extent

View Online

Abstract

Details

Metrics