Experimental Evaluation of a 3-D CZT Imaging Spectrometer for Potential Use in Compton-Enhanced PET Imaging

Abstract

We constructed a prototype positron emission tomography (PET) system and experimentally evaluated large-volume 3-D cadmium zinc telluride (CZT) detectors for potential use in Compton-enhanced PET imaging. The CZT spectrometer offers sub-0.5-mm spatial resolution, an ultrahigh energy resolution (~1% @ 511 keV), and the capability of detecting multiple gamma-ray interactions that simultaneously occurred. The system consists of four CZT detector panels with a detection area of around 4.4 cm × 4.4 cm. The distance between the front surfaces of the two opposite CZT detector panels is ~80 mm. This system allows us to detect coincident annihilation photons and Compton interactions inside the detectors and then, exploit Compton kinematics to predict the first Compton interaction site and reject chance coincidences. We have developed a numerical integration technique to model the near-field Compton response that incorporates Doppler broadening, detector's finite resolutions, and the distance between the first and second interactions. This method was used to effectively reject random and scattered coincidence events. In the preliminary imaging studies, we have used point sources, line sources, a custom-designed resolution phantom, and a commercial image quality (IQ) phantom to demonstrate an imaging resolution of approximately 0.75 mm in PET images, and Compton-based enhancement.

Keywords: Compton scattering; cadmium zinc telluride (CZT) detector; positron emission tomography (PET).

Fig. 1.

(Left) Prototype CZT-PET system and (Right) schematic and dimensions of the CZT-based detection system.

Fig. 2.

Illustration of the calculation of the probability $\mathit{p}({\mathit{A}}_{\mathit{j}}\mid \mathit{i})$.

Fig. 3.

Illustration of the near-field Compton imaging problem and the potential parallax error in modeling the Compton responses within a realistic CZT detector.

Fig. 4.

Illustration of calculating the probability that a gamma ray originated from a given source bin interacted with a detector-element through Compton scattering.

Fig. 5.

FWHM angular uncertainty map of the proposed near-field Compton model at different ${\widehat{d}}_2$ and scattering angle.

Fig. 6.

FWHM angular uncertainty map of the near-field Compton model if ${\widehat{d}}_2$ is not considered.

Fig. 7.

Illustration of using Compton kinematics to determine the interaction sequence (in terms of which is the first Compton interaction site among the two interactions detected on Detector 1). The colored bands are the Compton cones projected from the Compton event in Detector 1 onto Detector 2, where the second coincidence gamma-ray resulted in a single interaction as marked by the red dots.

Fig. 8.

Illustration of using Compton kinematics to reject chance-coincidence events. For both possible sequences, given the Compton cones derived by the Compton event in Detector 1, the probabilities of detecting the second coincidence gamma ray at the location marked by the red dots in Detector 2 are too low. Hence, this event is rejected as chance coincidence.

Fig. 9.

Schematic of the custom-made hot-rod resolution phantom containing four groups of hot rods of 0.35, 0.5, 0.75, and 1 mm diameter. The center-to-center distances of each two adjacent rods are 0.7, 1, 1.5, and 1.6 mm, respectively.

Fig. 10.

(a) Sensitivity and (b) its profiles of the prototype PET system estimated from modeled system response function.

Fig. 11.

Energy spectra of Cu-64 acquired by CZT detectors. (a) Energy spectrum of gamma-rays detected as one interaction. (b) Spectrum of the total energy of gamma-rays detected as Compton events (two detected interactions). (c) Spectrum of the total energy of gamma-rays detected as Compton events (three detected interactions).

Fig. 12.

(4th iteration) Three views and line profiles (going through the green dashed line) of the reconstructed Na-22 point source. The object space has 66 × 66 × 80 cubic voxels of 0.1 mm × 0.1 mm × 0.1 mm in size.

Fig. 13.

(11th iteration) (a) Reconstructed image of two line sources filled with Cu-64 and (b) its 1-D cross section. The object space has 128 × 128 × 128 cubic voxels of 0.25 mm × 0.25 mm × 0.25 mm in size.

Fig. 14.

Selected four uniform regions in the hot-rod resolution phantom with a thickness of 1 mm. The total volume 14.25 mm³.

Fig. 15.

Reconstructed images of hot-rod resolution phantom filled with Cu-64 with DOI resolutions of (a) 1 mm (9th iteration), (b) 2.5 mm (8th iteration), (c) 5 mm (9th iteration), and (d) 10 mm (6th iteration). Only coincidence events with single-site interactions were used in this comparison. The object space has 66 × 66 × 80 cubic voxels of 0.25 mm × 0.25 mm × 0.25 mm in size. These transverse views are of 1-mm thick slices. The normalized SDs of the ROI are all approximately 0.1534.

Fig. 16.

Comparison of line profiles (going through the red dashed lines in (a)) with different DOI resolutions, 2.5 mm (8th iteration), and 10.0 mm 6th iteration), in different phantom regions: (b) 0.75-mm hot-rod region and (c) 1-mm hot-rod region. One-millimeter-thick slices were used.

Fig. 17.

Total intensity of reconstructed images across the process of OS-EM iteration of (a) line sources and (b) hot-rod resolution phantom with different DOI resolutions.

Fig. 18.

(2nd iteration) Reconstructed image of the IQ phantom filled with Cu-64: (a) transverse view of rod region, (b) transverse view of uniform region, (c) coronal view, and (d) sagittal view. The object space has 80×86×96 cubic voxels of 0.25 mm × 0.25 mm × 0.25 mm in size.

Fig. 19.

With the near-field Compton model proposed in this study (blue solid lines) and the conventional Compton model without considering ${\widehat{d}}_2$ (orange dashed line), we evaluated the (a) relationship between the probability of false rejection and the rejection threshold and (b) relationship between NECR′/NECR and the rejection threshold.

Fig. 20.

Comparison of reconstructed images of the hot-rod resolution phantom with (a) dataset 1 (6th iteration), (b) dataset 2 (11th iteration), (c) dataset 3 (14th iteration), (d) dataset 4 (6th Iteration), (e) dataset 5 (7th iteration), and (f) dataset 6 (6th iteration) The object space has 66 × 66 × 80 cubic voxels of 0.25 mm × 0.25 mm × 0.25 mm in size. These transverse views are of 1-mm thick slices.

Fig. 21.

Comparison of line profiles from the 750-μm hot-rods [going through the red dashed line in (a)] of the resolution phantom: (b) reconstructed using datasets 1, 2, and 3 with P/V ratios of 1.13, 1.13, and 1.07; and (c) reconstructed using datasets 1, 4, 5, and 6 with P/V ratios of 1.13, 1.08, 1.05, and 1.02. The profiles are of transverse views. of 1-mm thick slices.

Fig. 22.

Comparison of line profiles going through 1.10-mm diameter rod (the red dashed line in (a)) of selected images reconstructed by dataset 1 (2nd iteration, dataset 4 (4th iteration), dataset 5 (5th iteration), dataset 6 (6th iteration). With Gaussian fitting, the FWHMs of these peaks are ~1.20 mm. The profiles are of transverse views of 1-mm thick slices.

Fig. 23.

Comparison of reconstructed images of the IQ phantom with (a) dataset 1 (2nd iteration), (b) dataset 4 (4th iteration), (c) dataset 5 (5th iteration), and (d) dataset 6 (6th Iteration). The object space has 80×86 × 96 cubic voxels of 0.25 mm × 0.25 mm × 0.25 mm in size. These transverse views are of 1-mm thick slices.

Fig. 24.

Selected ROI in the uniform region of the IQ phantom with a size of 10.25 mm × 10.25 mm × 4 mm. The total volume is 420.25 mm³.