JSNC Technologist Award
Simulation Study of High-sensitivity Cardiac-dedicated PET Systems with Different Geometries
2020 年 6 巻 1 号 p. 95-98
詳細
2020 年 6 巻 1 号 p. 95-98
Noninvasive quantification of myocardial blood flow with PET is a vital tool for detecting and monitoring of coronary artery disease. However, current standard cylindrical PET scanners are not optimized for cardiac imaging because they are designed mainly for whole-body imaging. In this study, we proposed two compact geometries, the elliptical geometry and the D-shape geometry, for cardiac-dedicated PET systems. We then evaluated their performance compared with a whole-body-size cylindrical geometry by using the Geant4 Monte Carlo simulation toolkit. In the simulation, an elliptical water phantom was scanned for 10-sec, and we calculated the sensitivity and the noise-equivalent count rate (NECR). Subsequently, a digital chest phantom was scanned for 30-sec and the coincidence data were reconstructed by in-house image reconstruction software. We evaluated the image noise in the liver region and the contrast recoveries in the heart region. Even with the limited number of detectors, the proposed compact geometries showed higher sensitivity than the whole-body geometry. The Dshape geometry achieved 47% higher NECR and 44% lower image noise compared with the whole-body cylindrical geometry. However, the contrasts in the hot area obtained by the proposed compact geometries were not as good as that obtained by the whole-body cylindrical geometry. There was no considerable difference in image quality between the elliptical geometry and the D-shape geometry. In conclusion, the compact geometries we have proposed are promising designs for a high-sensitivity and low-cost cardiac-dedicated PET system. A further study using a defect phantom model is required to evaluate the contrast of cold areas.
Coronary artery disease (CAD) is one of the leading causes of death world-wide. Noninvasive quantification of myocardial blood flow with SPECT or PET is a vital tool for detecting and monitoring CAD. Compared with SPECT, PET can provide better image quality and more accurate quantification, and allow dynamic first-pass imaging (1). A novel promising tracer, 18F-flurpiridaz, has been developed and its phase III trial is going on (2). It is expected that the demand for cardiac PET imaging will increase hereafter. However, current standard PET scanners are not optimized for cardiac imaging because they are designed mainly for whole-body imaging. A more compact geometry can increase the sensitivity and would be suitable for cardiac imaging.
In this study, we proposed two compact geometries for cardiac-dedicated PET systems, and evaluated their performance compared with a whole-body-size cylindrical geometry by using a simulation.
Using the Geant4 simulation toolkit (3, 4), we modeled three different PET systems:a whole-body-size cylindrical geometry, an elliptical geometry and a D-shape geometry (Figure 1). Although the D-shape geometry itself has been already studied (5), the D-shape proposed in this study has a smaller diameter to fit a torso closely. Specifications of the simulated systems are shown in Table 1. These specifications are based on realizable performance (6). For the whole-bodysize geometry, the larger coincidence time window of 6 ns was used in consideration of collecting all true coincidence counts. In each detector block, a paralyzable dead time of 1 μs was applied to a single event.
Detector arrangements of the simulated PET systems with the whole-body-size cylindrical geometry (a), the elliptical geometry (b), and the D-shape geometry (c).
The elliptical water phantom was placed on the central field-of-view (Figure 1). The long axis was 30 cm, the short axis was 20 cm and the length was 15 cm. The size and shape of the phantom mimicked a human chest. The water phantom was uniformly filled with an 18F activity of 10 MBq. Sensitivity was measured as the ratio of true coincidence count and decay count. At the fixed activity of 10 MBq, we calculated noise-equivalent count ratio (NECR) using the following equation:
 |
The digital chest phantom provided by the Japanese Society of Nuclear Medicine was used for image quality evaluation (Figure 2a) (7). Activity of 18F was 10 MBq in total. The activity ratio of heart, liver, and background was 20:10:1 (Figure 2b). The attenuation coefficient map included bone, lung, and soft tissue (Figure 2c). Scan duration was 30 seconds, assuming dynamic imaging. Coincidence data were reconstructed by the 3-dimensional (3D) ordered-subsets expectation-maximization (OSEM) algorithm with 3 iterations and 8 subsets. Normalization, attenuation, scatter and singles-based random corrections were included in the reconstruction process. The image matrix size was 128×128×64 with a 3.0 mm isotropic voxel. A 5-cm-diameter circular region-of-interest (ROI) was placed on the liver in three axial slices (Figure 2d). For image noise evaluation, we measured a coefficient of variation (CV) in the ROIs as follows:
 |
 |
Simulation setup (D-shape geometry) (a) and four representative slices of the activity map (b) and attenuation map (c), and ROI settings (d) for the digital chest phantom. PET images and their CRheart-to-liver, CRheart-to-ventricle and CVliver, obtained by the three different geometries (e).
The sensitivities of the whole-body cylindrical geometry, the elliptical geometry, and the D-shape geometry were 0.56%, 0.94% and 0.96%, respectively. The corresponding NECRs were 33.1, 47.7, and 48.6 kcps. The D-shape geometry showed the highest sensitivity and the highest NECR.
PET images of the digital chest phantom and those CRheart-to-liver, CRheart-to-ventricle and CVliver are shown in Figure 2e. The compact elliptical and D-shape geometries achieved a superior CVliver compared to the whole-body-size cylindrical geometry. On the other hand, the CRheart-to-liver and CRheart-to-ventricle, obtained by the compact geometries were not as good as that obtained by the whole-body-size cylindrical geometry.
We proposed two compact geometries for cardiac-dedicated PET systems and carried out their Monte Carlo simulation. For the compact geometries, the number of detectors was 3/8 (62.5% reduction) compared to the wholebody cylindrical geometry. Production cost and system size for the former two geometries can be reduced compared with the standard whole-body cylindrical geometry.
The compact D-shape geometry achieved 47% higher NECR and 44% lower image noise (CVliver) compared with the whole-body cylindrical geometry. Even with the limited number of detectors, the proposed compact geometries showed higher sensitivity than the whole-body cylindrical geometry. These compact geometries might enable more accurate dynamic myocardial imaging. However, the contrast recoveries (CRheart-to-liver and CRheart-to-ventricle) for these compact geometries was not as good as that for the whole-body cylindrical geometry. The reason for this degradation would be parallax error. The effect of parallax error is bigger as the detector ring diameter is decreased while the sensitivity is increased (8). A further investigation using a point source is needed to clarify the effect of parallax error for these compact geometries. A detector with depth-of-interaction (DOI) measurement capability is preferable to address this issue. In addition, as we simulated the only one crystal size, further simulations using various crystal sizes and detector configurations are needed to investigate a suitable detector configuration for cardiac imaging. The combination of a compact geometry and a smaller scintillation crystal might achieve higher contrast with acceptable noise level because of its higher sensitivity.
There was no considerable difference in image quality between the elliptical geometry and the D-shape geometry. In this work, we only evaluated image noise levels and the image contrast in hot areas on reconstructed PET images. For cardiac PET imaging, the image contrast of defect regions (cold contrast) is critical to detect myocardial ischemia. Next, we need to evaluate the contrast of cold areas using a defect phantom model.
The compact geometries we have proposed are promising designs for a high-sensitivity and low-cost cardiac-dedicated PET system. A further study using a defect phantom model is required in order to evaluate the contrast of cold areas.
None.
This study was supported by QST President’ s Strategic Grant (Exploratory Research).
No conflicts of interest are disclosed.
