Effects of Detector Thickness on Geometric Sensitivity and Event Positioning Errors in the Rectangular PET/X Scanner

Abstract

We used simulations to investigate the relationship between sensitivity and spatial resolution as a function of crystal thickness in a rectangular PET scanner intended for quantitative assessment of breast cancers. The system had two 20 × 15-cm2 and two 10 × 15-cm2 flat detectors forming a box, with the larger detectors separated by 4 or 8 cm. Depth-of-interaction (DOI) resolution was modeled as a function of crystal thickness based on prior measurements. Spatial resolution was evaluated independent of image reconstruction by deriving and validating a surrogate metric from list-mode data (dFWHM). When increasing crystal thickness from 5 to 40 mm, and without using DOI information, the dFWHM for a centered point source increased from 0.72 to 1.6 mm. Including DOI information improved dFWHM by 12% and 27% for 5- and 40-mm-thick crystals, respectively. For a point source in the corner of the FOV, use of DOI information improved dFWHM by 20% (5-mm crystal) and 44% (40-mm crystal). Sensitivity was 7.7% for 10-mm-thick crystals (8-cm object). Increasing crystal thickness on the smaller side detectors from 10 to 20 mm (keeping 10-mm crystals on the larger detectors) boosted sensitivity by 24% (relative) and degraded dFWHM by only ~3%/8% with/without DOI information. The benefits of measuring DOI must be evaluated in terms of the intended clinical task of assessing tracer uptake in small lesions. Increasing crystal thickness on the smaller side detectors provides substantial sensitivity increase with minimal accompanying loss in resolution.

Simulation study of quantitative precision of the PET/X dedicated breast PET scanner

“…This paper extends our previous analysis of trade-offs in crystal thickness selection 28,29 by (i) including the use of two independent fully three-dimensional (3D) image reconstruction methods, 26,27 (ii) adding a study of reproducibility through multiple simulation realizations, and (iii) performing a receiver-operating characteristic (ROC) analysis of the ability of the PET/X scanner to accurately measure changes in lesion contrast. In this first report of performance metrics from reconstructed PET/X images, we use idealized data models to investigate fundamental upper bound performance limits of the PET/X geometry and image reconstruction approaches.…”

“…This data filtering process was validated in a prior study. 28 The total energy threshold was set at 510 keV to reject scattered events. The remaining interactions were then used to compute the line of response (LOR) as detailed below.…”

“…We modeled event positioning within the detectors in the following three ways: 28 1. The first vertex interaction location of the 511-keV photon.…”

“…We also calculated a spatial-resolution surrogate, d FWHM , using the distance of closest approach (DCA) method, first introduced by MacDonald et al 28 The appeal of the DCA method is that it is independent of the image reconstruction method; here, we used list-mode data, but it can also be estimated from sinogram data. The DCA is an event-by-event metric defined as the shortest distance from each positron annihilation position to the corresponding LOR determined by the detectors.…”

Image reconstruction in rectangular PET systems using distance-driven projections

“…Recently, a design of a box-shaped PET scanner was proposed to attach to an x-ray mammography gantry to form a PET/X system as illustrated in Fig. 1 [3]. The scanner will enable development and evaluation of new and existing breast cancer therapies using PET images as prognostic biomarkers.…”