Value of SiPM PET in myocardial perfusion imaging using Rubidium-82

We performed a prospective single-center study and included 30 consecutive patients referred for MPI using PMT PET (Discovery 690, GE Healthcare; D690) with Rb-82 for the evaluation of coronary artery disease. Within three weeks after the first PET scan, patients underwent a second MPI PET scan on a SiPM PET scanner (Vereos, Philips Healthcare). The local institutional ethics committee approved the study protocol and informed consent was obtained from all individual participants included in the study.

Patients were asked to refrain from caffeine containing beverages for at least 24 h before both scans. All patients underwent a rest scan followed by a regadenoson-induced stress scan on both scanners. First, a low-dose computed tomography (CT)-scan was performed for attenuation correction purposes. The CT scan on the D690 was performed using 0.8 s rotation time, pitch of 0.97, collimation of 32 × 0.625 mm, tube voltage of 120 kV, and tube current of 10 mA. On the Vereos, the CT scan was acquired using 1.5 s rotation time, pitch of 0.83, collimation of 64 × 0.625 mm, tube voltage of 120 kV, and tube current of 22 mA. The PET acquisition protocol was similar for the D690 and Vereos. A fixed activity of 740 MBq Rb-82 was intravenously administered with a flow rate of 50 mL/min using a strontium-82/Rb-82 generator (CardioGen-82, Bracco Diagnostics Inc.) immediately followed by a seven-minute PET acquisition. Ten minutes after the first activity bolus, stress was pharmacologically induced by administering 400 µg (5 mL) regadenoson over 10 seconds. After a 5 mL saline flush (NaCl 0.9%), the second activity bolus of 740 MBq was administered followed by a seven-minute stress PET acquisition. To obtain patient’s effective radiation dose for both PET examinations we used the conversion factors of 0.00126 mSv/MBq for rest and 0.00128 mSv/MBq for stress14,15 resulting in a total dose of 1.9 mSv. To calculate the effective dose for the attenuation CT, we used a conversion factor of 0.014 mSv/(mGy·cm)16 resulting in 0.2 mSv for the D690 based on an average dose length product (DLP) of 11.8 mGy·cm and 0.8 mSv for the Vereos based on an average DLP of 60.5 mGy·cm.

CT data associated with the D690 were reconstructed using an iterative reconstruction method (70% adaptive statistical iterative reconstruction algorithm, ASIR) and a slice thickness of 5 mm. CT data associated with the Vereos were reconstructed using an iterative reconstruction method (iDose level 4) and a slice thickness of 3 mm.

We applied attenuation correction to all acquired PET data after semi-automatic registration of the CT and PET using the PET data acquired between 2:30 and 7:00 minutes.17 We reconstructed the images of the D690 with a 3D-ordered subset expectation maximization (OSEM) technique using 2 iterations and 24 subsets and a Gaussian post-smoothing filter of 12 mm, as recommended by the manufacturer. The voxel size of the D690 was 3.3 × 3.3 × 3.3 mm³. Images of the Vereos were reconstructed with 3D OSEM using 3 iterations and 15 subsets and a Gaussian post-smoothing filter of 6 mm. The voxel size of the Vereos was 4.0 × 4.0 × 4.0 mm³. These Vereos settings were determined prior to our study (see appendix), based on measurements using an anthropomorphic torso phantom with a cardiac insert (model ECT/TOR/P, Data Spectrum Corp.). Intensity profiles through the cardiac insert were collected for several reconstruction settings to compare the full width at half maximum value to that of the D690. This way we obtained reconstruction settings resulting in an equivalent image resolution. For both the D690 and the Vereos data, corrections were performed for decay, scatter and random coincidences, and dead time effects. We used data acquired from 2:30 to 7:00 minutes for both rest and stress scans to obtain static images. Dynamic data sets were reconstructed using 26 time frames (12 × 5 seconds, 6 × 10 seconds, 4 × 20 seconds and 4 × 40 seconds). All reconstructed images were post-processed using Corridor4DM software (v2016).

Each set of static rest and stress PET images, showing the relative perfusion, was analyzed by two expert readers in consensus. They scored the image quality, visibility of perfusion defects, and interpreter’s confidence. Image quality of the static images was assessed using a four-point grading scale: 1) poor, 2) fair, 3) good, and 4) excellent. Readers assessed the image quality based on myocardial count density and uniformity in well-perfused areas, signal to background noise, and shape of the left ventricle (LV). Static images were visually characterized as normal or abnormal. Abnormal scans were characterized as ischemic and/or irreversible. The interpreter’s confidence was scored as either definite or equivocal. Readers were unaware of the patient’s history or other clinical findings. Images were presented in random order and readers were blinded for the PET system.

Activity concentrations were measured in the 26 reconstructed time frames to calculate the time activity curves (TACs) for the LV, for the three vascular territories: left anterior descending (LAD), left circumflex (LCX) and right coronary (RCA) artery, and for the whole myocardium (global). The one-tissue compartment model of Lortie et al. based on a ROI methodology was used to calculate the MBF from the TACs.18 Rest MBF was calculated without rate-pressure product correction. Furthermore, myocardial flow reserve (MFR) was calculated as the ratio between the stress and rest MBF. We categorized the global MFR values into three categories: high risk on cardiac failure with MFR <1.5, intermediate risk with MFR between 1.5 and 2.0, and low risk with MFR > 2.0.2,19

Rest or stress MBF and MFR values were excluded from the comparison evaluation in case of unreliable TACs. Unreliable TACs were defined as TACs without a clear LV peak during the first-pass phase or a lack of steady state for the three vascular territories during the tissue phase, as previously described.20 Test–retest precision was calculated as the standard deviation (SD) of the relative MBF and MFR differences, as previously defined by Kitkungvan et al.21 A test–retest precision ≤ 21% was considered acceptable.21

Patient-specific parameters and characteristics were determined as mean ± SD, or as percentages using SPSS (IBM SPSS Statistics for Windows, Version 24.0). Image quality, MBF and MFR measurements were compared using the Wilcoxon signed rank test. In addition, the visibility of perfusion defects and interpreter’s confidence were compared using the McNemar test. The level of statistical significance was set to 0.05 for all statistical analyses.

The baseline characteristics of the included patients are summarized in Table 1.

Image quality of the static images improved (P = 0.03) using the Vereos as compared to the D690. Image quality of the D690 and the Vereos was graded fair in 20% (6/30) and 10% (3/30), good in 60% (18/30) and 50% (15/30), and excellent in 20% (6/30) and 40% (12/30), respectively, as illustrated in Figure 1. None of the images using either the D690 or the Vereos were scored as poor. An example of the image quality for patients with high and low BMI is shown in Figure 2.

Defect interpretation did not differ in 93% (28/30) of the patient scans between the D690 and the Vereos (P = 0.50). In the 7% (2/30) of patient scans where defect interpretation differed, the scans were scored as normal on the D690, whereas they were interpreted to show ischemia on the Vereos (Figures 3 and 4). Patient scans were scored as normal in 80% (24/30) and 73% (22/30) for the D690 and the Vereos, respectively. Furthermore, 10% (3/30) and 17% (5/30) was interpreted as showing ischemia and for both PET scanners 13% (4/30) was scored as showing an irreversible defect. There was no difference in interpreter’s confidence as all scans were scored as definite.

Of the 30 included patients, both rest and stress MBF values of one patient were excluded due to unreliable rest and stress TACs. Furthermore, rest MBF values of another patient and stress MBF values of four other patients were excluded due to unreliable TACs. The main reason for an unreliable TAC was no clear or absent LV peak, which would be most likely caused by a pinched vein.22 The remaining PET scans provided a paired comparison of 28 rest MBF values, 25 stress MBF values, and 24 MFR values.

There were no significant differences in any of the vascular territories nor in the whole myocardium regarding the rest MBF (P ≥ 0.29), stress MBF (P ≥ 0.11) or MFR (P ≥ 0.51), as shown in Table 2 and Figure 5. When categorizing the global MFR values into high, intermediate, and low risk on cardiac failure, 25% (6/24) of the patients were reclassified when using the Vereos. More specifically, one patient was reclassified from intermediate risk to high risk, three from intermediate risk to low risk, and two patients from low risk to intermediate risk. None of the patients were reclassified from low risk to high risk or vice versa. Moreover, test–retest precision of global rest MBF, stress MBF, and MFR was 18%, 16%, and 21%, respectively, and was considered to be within the previously reported test–retest precision of 21%, as shown in Figure 6.

The Vereos scanner was the first SiPM PET scanner available for clinical use6 after which two other SiPM PET scanners became available, namely, the Biograph Vision PET/CT (Siemens Healthineers) and the Discovery MI (GE healthcare).7,8 As the performance characteristics of SiPM PET are in general better that those of PMT PET, image quality is expected to improve for all three SiPM PET scanners. Moreover, flow measurements are expected to be similar or possibly more accurate as compared to using PMT PET, provided that PMT PET has a sufficient count-rate capability.7,8,26,28,29

Whereas the MFR was shown to be robust when using different advanced reconstruction settings or software packages, one should be cautious in the occurrence of possible systematic changes in MBF measurements.30,31 Furthermore, one has to be aware of a relatively large test–retest precision in MBF and MFR measurements of typically 21%. In general, a MFR<1.5 is associated with an increased risk on cardiac failure while patients with a MFR>2.0 are associated with a reduced risk on cardiac failure.2,19 Hence, it is possible that a patient classified as having an intermediate risk may be classified as a low/high-risk patient when repeating the scan and reprocessing the data, solely due to the test–retest variation.

The use of the Vereos SiPM system in PET Rb-82 MPI results in an improved image quality and no significant differences for visual interpretation or interpreter’s confidence in comparison with the conventional D690 PMT PET scanner. Furthermore, no significant differences were found in MBF and MFR quantification. We did find a change in visual defect interpretation in two patients. Defect interpretation may therefore differ and it could be possible that the Vereos SiPM PET system has a superior diagnostic performance over the conventional D690 PMT PET system. Additional studies with a larger patient population are required to confirm this.