Phantom studies
In an anthropomorphic torso phantom (model ECT/TOR/P, DATA Spectrum, Hillsborough, NC, USA) a static cardiac insert was used (Data spectrum cardiac phantom model ECT/CAR/UM, DATA Spectrum, Hillsborough, NC, USA; Fig. 1), consisting of a left ventricle with separate compartments for blood pool (with a volume of 61 mL) and myocardium (with wall thickness of 10 mm and volume of 100 mL). The insert compartments were filled with a solution of water and 99mTc-pertechnetate. Three fillable defects (small, 2.6 mL; medium, 5.6 mL; and large, 11.8 mL) were evaluated to simulate a myocardial perfusion defect. The defects were positioned in the anterior or inferior wall of the myocardium, or a combination of both. The cardiac insert was positioned in an anthropomorphic torso phantom containing lung, liver and spine inserts. The myocardium was filled with 64 kBq/mL. Moreover, the liver was filled with 15.9 kBq/mL according to a myocardium-to-liver ratio of 4:1. The myocardium-to-background ratio was approximately 12:1. In addition, the myocardium-to-blood pool ratio was 10:1 and the defect-to-myocardium ratio was 1:2, based on literature [9,10,11]. The lungs were filled with polystyrene spheres and water to achieve a representative physiologic tissue density.
A total of 36 measurements were acquired and evaluated: six configurations with six acquisitions. The first three single configurations are the described fillable defects measured separately on the mid anterior part of the myocardium, further referred to as small (S-configuration), medium (M-configuration) and large (L-configuration). The fourth and fifth configurations are two combinations of two defects and were positioned on the mid-inferior and mid-anterior part of the myocardium. These double defect configurations consist of two small defects (SS-configuration) and a combination of a small and medium defect (SM-configuration). Finally, the phantom was also scanned without defects, further referred to as the ground truth (GT-configuration). The measurements were repeated six times with a time per view that was adjusted per repetition to compensate for the radioactive decay and to have similar count statistics as the first experiment.
Data acquisition and reconstruction were based on the clinical protocol using one bed position according to the EANM guidelines [12]. All studies were performed with a SPECT/CT dual head system (Discovery NM/CT 670 Pro, GE Healthcare, Milwaukee, Wisconsin, USA). The SPECT measurements are acquired using a low-energy, high-resolution (LEHR) collimator that was positioned in L mode, noncircular orbit, step-and shoot mode and 60 (2 × 30) views. The technetium energy window (photopeak) was set on 140.5 keV (window ±10%) for emission and on 120 keV (window ±5%) for scatter. The camera sensitivity was determined as recommended by the vendor [6] as described in more detail in Collarino et al. [7]. After the SPECT acquisition, a low-dose CT scan (120 kV, 20 mAs, pitch 0.938, collimation 16 × 1.25) was acquired for attenuation correction purposes.
All SPECT data were reconstructed using Evolution with compensation for collimator-detector response, resolution recovery, attenuation and scatter on a matrix of 64 × 64 voxels with 1.5 zoom, resulting in a voxel size of 5.89 × 5.89 × 5.89 mm3. Moreover, the CT data were reconstructed using adaptive statistical iterative reconstruction (ASIR, GE healthcare) with a voxel size of 0.98 × 0.98 × 5.00 mm3.
Based on the recommendations of the vendor, no post-reconstruction filter should be applied for quantification [6]. However, a Butterworth filter (cut-off frequency of 0.52 cycles/cm and an order of 5) was recommended by the vendor for clinical SPECT/CT cardiac studies without quantification. Therefore, SPECT data were reconstructed both with and without a Butterworth filter to investigate the impact of post-filtering on quantification. After reconstruction, the Q.Metrix package resampled both the CT and SPECT images to an equivalent and isotropic voxel size (1.47 × 1.47 × 1.47 mm3) that was used for delineation.
Data analysis
All SPECT/CT images were converted from counts to Bq/mL using Q.Metrix as detailed in the paper of Collarino et al. [7]. To determine the number of iterations that are required for Evolution to converge, volumes of interest (VOI) were drawn on CT images for the background, liver, small defect and myocardium compartments. Evolution was considered to have converged when for each VOI the relative difference in the activity concentration of the iteration was less than 1% with respect to the previous iteration. In order to determine the noise level, the coefficient of variation (COV) in the background compartment was calculated by the standard deviation divided by the mean activity concentration.
Interobserver reliability of delineation on CT images was assessed by three physicians that draw VOIs on the small defect and the myocardium for the S-configuration. The interobserver reliability was computed by an intraclass correlation coefficient using a two-way mixed effects model for absolute agreement (ICCA) (SPSS statistics; version 25; IBM Statistics, Armonk, USA). ICC scores range from 0 to 1, representing a level of agreement: ≤ 0.40 poor to fair, 0.41–0.60 moderate, 0.61–0.80 substantial and 0.81–1.00 almost perfect [13].
Subsequently, recovery coefficients (RCs), representing the ratio between the reconstructed activity concentration (in Bq/mL) and the true activity concentration as measured with a dose calibrator (VIK-202, Comecer, the Netherlands), were computed for VOIs drawn on CT images for the myocardium and defect compartments for all configurations. Here, an RC of 1 indicates that injected activity concentration was in accordance with the measured activity concentration. Furthermore, for all six configurations separately, the precision was expressed in average repeatability deviation (RD) calculated by the standard deviation (SD) of the myocardium RCs as a percentage of the mean myocardium RC. Moreover, Bull’s eye plots with 17 segments were generated using 4DM (INVIA-Ann Arbor, MI, USA) for the S- and GT-configurations.
Patient cases
Patient cases were evaluated retrospectively to illustrate the clinical applicability of the phantom study. For patients with CAD and ischemia without long-term treatment options, intramyocardial injection of autologous bone marrow cells (BMC) has emerged as an alternative treatment strategy at Leiden University Medical Center (LUMC) [14]. SPECT/CT data were composed before and after BMC injection. The collected patient data were anonymized and recorded in a database. Performance of this retrospective study was approved by the medical ethical review board and the requirement to obtain informed consent was waived.
Comparing the MPS at rest and after stress, areas of under-perfusion and resultant stress-induced ischemia were identified. According to a two-day protocol, as described in the EANM guidelines, patients under and over 100 kg received 500 and 750 MBq 99mTc-tetrofosmin, respectively, for both the stress and the rest examination [12]. All patients underwent a pharmacologic adenosine stress examination. SPECT/CT was acquired 30–45 min post-injection after stress and 45–60 min post-injection during rest.
All stress and rest studies were converted into lean-body mass standardized uptake values (SUVLBM) in g/mL and calculated as described in Kim et al. [15]. Reversibility Bull’s eye plots were composed for the studies before and after BMC injection as used in clinical practice [12]. Subsequently, the difference in tracer uptake (SUVLBM) prior to and after BMC injection was calculated by the reversibility Bull’s eye plots after BMC injection minus before BMC injection.
Statistical analysis
Statistical analysis was performed using SPSS statistics software and Excel (version 2017; Microsoft, Redmond, USA). A Shapiro-Wilk test was performed in order to evaluate the (log)normality of the data. Paired data for both the phantom and the patient study were statistically performed with the paired T test or Wilcoxon Signed Rank test depending on the (log)normality. The statistical analyses of the non-paired data were performed with an independent T test or Mann-Whitney U depending on the (log)normality. Continuous data were expressed as mean(SD).