Nuclear Myocardial Perfusion Imaging with a Cadmium-Zinc-Telluride Detector Technique: Optimized Protocol for Scan Time Reduction

Study Protocol

In all 20 patients, a 1-d ^99mTc-tetrofosmin adenosine-stress/rest imaging protocol was performed (10). The patients were told to refrain from caffeine-containing beverages for at least 12 h before the MPI study. Pharmacologic stress was induced by infusion of adenosine at a rate of 140 μg/kg/min. A standard dose of 300–350 MBq of ^99mTc-tetrofosmin was injected 3 min into the pharmacologic stress. Ninety minutes later, the acquisition of the stress study was performed on the Ventri, immediately followed by the acquisition on the CZT camera. Thereafter, a tracer dose 2.5–3 times the stress dose was administered at rest, followed by image acquisition on both cameras with the same protocol as at stress.

SPECT MPI Image Acquisition and Reconstruction

The first scan was acquired on a Ventri dual-head camera with a low-energy, high-resolution collimator, a 20% symmetric window at 140 keV, a 64 × 64 matrix, and an elliptic orbit with step-and-shoot acquisition at 3° intervals over a 180° arc (45° right anterior oblique to 45° left posterior oblique) with 30 steps (60 views). Scan time was set to 25 s per frame for stress and rest, resulting in a total acquisition time of 14 min 52 s (including interstep rotation time) for each scan as recommended by the American Society of Nuclear Cardiology (2). Reconstruction took place on a dedicated workstation using a standard iterative reconstruction algorithm with ordered-subset expectation maximization with 2 iterations and 10 subsets. Standard short-axis images, vertical and horizontal long-axis images, and polar maps of perfusion encompassing the entire left ventricle were produced without using resolution compensation or attenuation correction. A Butterworth postprocessing filter (frequency, 0.50; order, 10) was applied to the reconstructed slices.

CZT Image Acquisition and Reconstruction

The second scan was acquired on a CZT camera with pinhole collimation (11). In this camera, the conventional sodium iodide crystals have been replaced by CZT semiconductor technology, which directly converts radiation into electric signals without any of the steps of violet light production, transport, and conversion that occur with sodium iodide crystals (12). The energy resolution and spatial resolution (radial resolution, 4.3 mm) are improved by a factor of 2, and the sensitivity (21.0 counts/s/mCi) by a factor of almost 4, compared with a dedicated cardiac γ-camera (Ventri). The CZT technology is extremely compact, and this miniaturization together with the increased resolution offers the opportunity to construct a stationary array of 19 small γ-cameras packed closely together and focused on the heart. The stationary array simultaneously acquires all the views necessary for tomographic reconstruction, saving the time required by conventional cameras for acquisitions while rotating around the subject. All views simultaneously focus on the heart to maximize the efficiency of cardiac imaging. To fit the multiple views, the image is reduced in size by means of pinhole collimation, matching the miniaturization to the improved intrinsic pixel resolution of the detectors. This allows the use of the detector surface to be maximized, increasing system efficiency. The pinhole geometry has several advantages. The reduction in pinhole sensitivity with increasing distance significantly diminishes the contribution of background organs and tissues to the cardiac data, facilitating reliable 3-dimensional iterative reconstruction. In addition, the oblique angles of incidence also improve the already superior energy resolution of the CZT. Stress (low-dose) and rest (high-dose) scans were acquired over 6 min, and data were saved as list file with the ability to obtain scan durations of 1 min, 2 min, etc., up to a maximum of 6 min. Images were reconstructed on the same workstation as for the standard SPECT acquisition using a new dedicated iterative algorithm with integrated collimator geometry modeling using maximum penalized likelihood iterative reconstruction to obtain perfusion images in standard axes. For low counts, 40 iterations have been performed, and for high counts, 50 iterations have been performed. A Butterworth postprocessing filter (frequency, 0.37; order, 7) was applied to the reconstructed slices.

Image Analysis

Image quality was graded visually on a 4-point scale as 1 (poor), 2 (fair), 3 (good), or 4 (excellent). The following parameters were considered: myocardial count density and uniformity, endocardial and epicardial edge definition, visualization, and background noise.

Quantitative analysis was performed on MPI polar maps using a 20-segment model for the left ventricle (2). Uptake was normalized to 100% peak activity, and relative percentage count uptake of γ-ray emissions was assessed for each segment from Ventri and CZT data, for each of the scan durations, and for both low and high dose.

For clinically relevant analysis, we grouped the 20 segments into the territories of the 3 main coronaries, that is, left anterior descending artery, circumflex artery, and right coronary artery, as previously suggested (2). Visual analysis was performed with regard to the presence or absence of perfusion defects in every coronary territory. Wall motion was not considered. Two experienced nuclear cardiologists made the clinical analysis by consensus.

Torso Phantom Study

An anthropomorphic torso phantom (model ECT/TOR/P; Data Spectrum Corp.) (7) was used for the phantom studies as previously established for this type of device (13). The phantom, a large, body-shaped cylinder, comprised the lungs, the liver, spine inserts, and cardiac inserts (model ECT/CAR/I), simulating the upper torso of average-to-large patients (38 × 26 cm). The torso phantom was filled with ^99mTc: 57.7 kBq/mL (1.56 μCi/mL) for the cardiac insert; 38.5 kBq/mL (1.04 μCi/mL) for the liver; 4.8 kBq/mL (0.13 μCi/mL) for low-dose background; and with 112.5 kBq/mL (3.04 μCi/mL) for the cardiac insert, 56.2 kBq/mL (1.52 μCi/mL) for the liver, and 12.6 kBq/mL (0.34 μCi/mL) for high-dose background. The heart-to-liver-to-background ratio was 12:8:1 for low dose and 9:4.5:1 for high dose. Cardiac phantom studies were performed with a 2-cm defect of 50% activity but otherwise homogeneous tracer distribution.

Statistics

SPSS software (version 15.0, SPSS Inc.) was used for statistical testing. Quantitative variables were expressed as mean ± SD, and categoric variables as frequencies, mean, or percentages. For each of the scan durations, the tracer uptake value (percentage of maximum myocardial uptake) from the CZT camera was compared segmentwise by intraclass correlation to Ventri data, whereas clinical agreement was assessed per coronary territory and significant differences were identified by McNemar and Cohen κ-test. Scan durations above which no further improvement in relevant segmental uptake correlation was found (i.e., no more than a 5% increase in correlation coefficient) were defined as minimal required scan times. Correlation coefficient and Bland–Altman limits of agreement were calculated per territories for these durations. P values of less than 0.05 were considered statistically significant, and the 95% confidence intervals (CIs) are presented.

Image Analysis

Image quality was graded good or excellent in all stress and rest MPI scans on the Ventri. On the CZT camera, the image quality for stress (low dose) was good or excellent in 25% of the 1-min scans, in 80% of the 2-min scans, and in 100% of the scans with a duration of 3 or more minutes. Comparably, for the CZT rest scans (high dose), image quality was good or excellent in 65% of the 1-min scans but reached 100% in scans with 2 or more minutes' duration.

The intraclass correlation coefficient for quantitative segmental tracer uptake of low-dose scans was 0.67 (CI, 0.45–0.76) after 1 min, 0.77 (CI, 0.72–0.80) after 2 min, 0.81 (CI, 0.76–0.83) after 3 min, 0.79 (CI, 0.74–0.82) after 4 min, 0.82 (CI, 0.77–0.84) after 5 min, and 0.80 (CI, 0.75–0.83) after 6 min. Similarly, the values for high-dose scans were 0.72 (CI, 0.60–0.80) after 1 min, 0.80 (CI, 0.72–0.81) after 2 min, 0.81 (CI, 0.77–0.84) after 3 min, 0.81 (CI, 0.78–0.85) after 4 min, 0.80 (CI, 0.78–0.85), after 5 min, and 0.82 (CI, 0.78–0.85) after 6 min (Fig. 1). This identified 3 min (low dose) and 2 min (high dose) as minimum required scan times, because further prolongation of the scan interval did not improve correlation.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

(A) Intraclass correlation is used to compare segmental uptake between standard γ-camera (Ventri; GE Healthcare) and ultrafast CZT camera (Discovery NM 530c; GE Healthcare). Correlation coefficient increases with prolonged scan time, but no further improvement appears after scan duration of 3 min is reached for low ^99mTc-tetrofosmin dose (stress scan) and 2 min for high dose (rest scan). Thus, minimally required scan times are 2 min for high dose and 3 min for low dose. (B) Diagnostic accuracy of CZT, compared with that of standard camera, increases (and significant differences according to McNemar test disappear) after scan intervals are prolonged to 3 min for low dose (stress) and 2 min for high dose (rest). This indicates solid clinical agreement for the above scan durations, further supported by excellent Cohen κ-value (low dose, 3 min: 0.90; high dose, 2 min: 0.92).

Of the 60 coronary territories (20 patients × 3 coronary territories), 24 revealed a perfusion defect in low-dose scans and 18 in high-dose scans on the Ventri. The clinical agreement for low-dose scans was 75% (CI, 62%−85%) after 1 min, 87% (CI, 75%−94%) after 2 min, 95% (CI, 86%−99%) after 3 and 4 min, and 97% (CI, 88%−100%) after 5 and 6 min. For high-dose scans, the values were 77% (CI, 64%−87%) after 1 min, 97% (CI, 88%−100%) after 2 min, 98% (CI, 91%−100%) after 3 min, 95% (CI, 86%−99%) after 4 min, and 98% (CI, 91%−100%) after 5 and 6 min (Fig. 1).

Bland–Altman limits of agreement for uptake in coronary territories were determined at 2 and 3 min—the minimal scan durations identified above—resulting in values of −11.4% to 12.2% and −7.6% to 12.9% for low- and high-dose scans, respectively (Fig. 2).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

Linear regression analysis (left) and Bland–Altman plots (right) for per territory percentage tracer uptake at minimal required scan times; that is, 3 min for low dose (stress) (A) and 2 min for high dose (rest) (B).

Torso Phantom Study

Figure 3 shows images of the torso phantom on the CZT and Ventri cameras at the minimal required scan times of 3 min for low dose and 2 min for high dose, demonstrating a sharper edge definition for the CZT images than for the Ventri images because of the improved spatial resolution of the new technology. No defect was identified for either the CZT or the Ventri images. The agreement in segmental myocardial percentage uptake between CZT and Ventri is illustrated in Figure 4.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3. 

Torso phantom reconstructions obtained from low- and high-dose acquisitions on CZT camera (3- and 2-min scan times, respectively) and on standard γ-camera (15-min scan time each) showing concordant defect in inferolateral wall. HLA = horizontal long axis; SA = short axis; VLA = vertical long axis.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4. 

Torso phantom studies demonstrating excellent concordance of repeated measurements of percentage uptake of myocardial ^99mTc activity between standard and CZT cameras for each of 20 segments for low and high doses (respective scan times as given in Fig. 3).