Aim

Cardiovascular diseases represent the first cause of death in Europe and worldwide.1 Myocardial perfusion imaging (MPI) with single-photon emission computed tomography (SPECT) plays a major role in the screening and follow-up of patients suffering from ischemic heart disease (IHD). This imaging modality is well validated and a negative stress MPI identifies subjects at low risk of future cardiovascular events.2

In the last decade, dedicated cadmium-zinc-telluride (CZT) cameras brought a new advance in nuclear cardiology, offering higher sensitivity and better spatial and energy resolution than Anger cameras3 while holding the same negative predictive value.4 These dedicated cameras allow for a reduction in acquisition time and/or in injected radiopharmaceutical activity without compromising image quality.5,6,7,–8 Recently, CZT detector technology has been expanded to dual head SPECT cameras for whole-body imaging. These cameras with parallel collimators can be used for a wide range of scintigraphic examinations with faster acquisitions,9,10,11,12,13,–14 which could lead to their development. The feasibility of myocardial perfusion SPECT imaging using the new whole-body General Electric Discovery NM 670 CZT camera was reported with technetium-99m labeled tracers15,16 but there is still a need for standardization of acquisition and reconstruction parameters.

As with dedicated CZT camera,17 we hypothesized that dose or time reduction is also achievable with whole-body SPECT cameras without compromising image quality. This study aimed to determine the equivalence between a conventional protocol of MPI performed with 201Thallium on a CZT SPECT/CT camera, and a reduced acquisition time protocol simulated by list-mode edition.

Materials and Methods

Design

This monocentric trial prospectively included consecutive patients performing a MPI in our center from December 2017 to March 2018. The study was approved by the local Ethical Committee and conformed to the Declaration of Helsinki on human research. Written informed consent was obtained from every patient after complete explanation of the protocol.

Stress Protocol and Image Acquisition

Exercise stress testing was performed on a bicycle ergometer up to 85% of age-predicted maximal heart rate before intravenous infusion of 1.5 MBq kg−1 of 201Thallium. For patients who could not sustain physical exercise, a pharmacologic stress was performed with an intravenous infusion of 0.7 mg kg−1 of dipyridamole (with a maximum dose of 63 mg). All stress tests were performed under 12-lead electrocardiogram monitoring with serial blood pressure measurement. All images were recorded in list mode on a DISCOVERY NM 670 CZT camera (GE Healthcare) within a maximum of 10 minutes after the injection of 201Thallium (50 s/projection, 32 projections).

Post-processing

First, native data were anonymized with “Anonymized Export”, a dedicated program available on Xeleris 4.0 stations (GE Healthcare). Anonymization was performed by a Nuclear Medicine physician subsequently not involved in images analysis (CA). After anonymization, data were post-processed by the same physician to decrease the total recorded counts by 25% or 50% (simulating a reduction of 25% or 50% in the acquisition time), thus generating D75 and D50 datasets, respectively.

Images were reconstructed by filtered back projection (FBP) with QGS/QPS software on a Xeleris 4.0 station (GE Healthcare), using a Butterworth filter with cutting frequency set to 0.4 and power set to 10 and a quantitative ramp filter. Images were manually reoriented by each reader during analysis and a motion correction was applied if necessary. Correction attenuation was not performed.

Stress images were analyzed with QGS/QPS software by three different readers (PBB, LJ, RAV) blinded for post-processing and clinical information. Each reader had access to all functions of the software to study perfusion (including multiplanar images with various color scales and normalized polar maps with a 17-segment LV model) and function on gated images. A qualitative and semi-quantitative analysis was performed. Qualitative interpretation was based on the experience of the three readers to determine the number of segments considered to be pathological. For example, in case of perfusion abnormalities, data obtained from ECG-gated images helped to distinguish true perfusion defects from artifacts. Semi-quantitative analysis expressed segmental tracer uptake on stress scans as a percentage of maximum myocardial tracer uptake. This analysis aimed to identify segments displaying a tracer uptake lower than 70% of the maximum. LV function analysis was performed from 16 frames ECG-gated images with an automatic calculation of end-diastolic volume (EDV), end-systolic volume (ESV) and ejection fraction (EF).

Statistical Analysis

The main outcome was to assess the equivalence between datasets (full dataset, D75 and D50) concerning the number of segments displaying an uptake lower than 70% of the maximum myocardial uptake.

Secondary outcomes included the assessment of equivalence between datasets concerning the number of segments considered to be pathological and the values of functional parameters (EF, EDV and ESV) derived from ECG-gated images. Finally, we also evaluated inter-observer agreement in interpretation, based on the number of segments considered as pathological and for functional parameters.

Equivalence was demonstrated if the CI95 of the difference fell within equivalence margins, set to 1 segment. Multiple testing was addressed using a hierarchical gatekeeping procedure. Agreement between the 3 readers was assessed by the Lin’s CCC (Concordance Correlation Coefficient).

Results

During the study period, 103 consecutive patients with complete stress MPI were included. The description of the population is summarized in Table 1.

Table 1 Clinical characteristics of the population
Full size table

Segmental perfusion and functional parameters have shown a good concordance between the three datasets both for normal and pathological cases, as illustrated in Figures 1 and 2. The average number of segments identified per patient as displaying a tracer uptake lower than 70% of the maximum myocardial uptake was 6.0, 6.0 and 6.5 for the full, D75 and D50 datasets respectively (range was 0-14 for the 3 datasets). The mean number of segments considered pathological per patient was 0.8, 0.8 and 0.9 (range 0-8), respectively. Mean FE was 68.5%, 68.9% and 67.8% respectively (Table 2).

Figure 1
figure 1

Clinical illustration of a normal post-stress myocardial perfusion SPECT (bullseye and short axis). No pathological segment was found, irrespective of the dataset used: full set (a) reduced to 75% (b) and reduced to 50% of total counts (c). EF value and myocardial volumes show limited variations. EF, ejection fraction; ESV, end-systolic volume; EDV, end-diastolic volume

Full size image
Figure 2
figure 2

Clinical illustration of pathological post-stress myocardial perfusion SPECT (bullseye and short axis). Pathological segments appear identical whatever dataset is used: full set (a) reduced to 75% (b) and reduced to 50 % of total counts (c). EF value and myocardial volumes show limited variations. We can observe an increased visualization of digestive activity after a 50% reduction of total counts in comparison with other datasets. EF, ejection fraction; ESV, end-systolic volume; EDV, end-diastolic volume

Full size image
Table 2 MPI results according to the post-processing protocol
Full size table

Considering our principal outcome, equivalence in the number of segments displaying a tracer uptake lower than 70% of the maximum myocardial uptake was demonstrated for both count-reduced datasets (Table 3A). Comparison between full dataset and D75 yielded a mean difference of 0.06 (CI 95 [− 0.15;0.27]), and with our pre-defined equivalence margin of 1 segment, the associated P-value was < 0.001. Regarding the full versus D50 datasets comparison, mean difference was 0.518 (CI 95 [0.28;0.76]) and the associated P-value < 0.001. Agreement between readers was good concerning the number of segments displaying a tracer uptake lower than 70% of the maximum myocardial uptake, with a CCC = 0.90 for full dataset, 0.89 for D75 and 0.88 for D50 (Table 4).

Table 3 Results for equivalence tests
Full size table
Table 4 Evaluation of concordance with Lin Concordance Coefficients (CCC)
Full size table

Equivalence was also demonstrated for both count-reduced datasets concerning the number of segments considered to be pathological per patient (Table 3B). Comparison between full dataset and D75 yielded a mean difference of 0.003 (CI 95 [− 0.07;0.06]) which, considering an equivalence margin of 1 segment, led to a P-value < 0.001. Regarding the full versus D50 datasets comparison, mean difference was 0.147 (CI 95 [0.03;0.27]) and the associated P-value was < 0.001. Agreement between observers was moderate relating to the number of segments identified as pathological (between 0.6 and 0.7)

Agreement between readers was excellent for EF (0.97, 0.96 and 0.97 for full dataset, D75 and D50, respectively), EDV (0.98, 0.75 and 0.98 respectively) and ESV (0.99, 0.82 and 0.97 respectively).

Discussion

Our results, obtained in a prospective cohort of consecutive patients referred for a 201Tl stress MPI, confirm that a reduction in dose or acquisition time is possible with novel whole-body CZT SPECT/CT cameras. We demonstrated that, up to a reduction of 50% in total recorded counts, the number of segments displaying a tracer uptake lower than 70% of the maximum myocardial uptake is equivalent to that obtained with our standard acquisition protocol. This equivalence remains true for the number of segments considered pathological, and for functional parameters such as EF and LV volumes.

In our study, the agreement between the 3 readers, blinded from post-processing parameters, was excellent for quantitative and semi-quantitative variables including the number of segments displaying a tracer uptake lower than 70% of the maximum myocardial uptake, EF, EDV or ESV. Inter-observer agreement for more subjective variables such as the number of segments considered pathological is lower but still remains satisfying.

These data are consistent with the literature. A moderate inter-observer variability has already been reported for MPI by Johansen et al18 They reported moderate to good inter-observer agreement (κ = 0.52-0.56) for segmental score interpretation in a cohort of 108 consecutive male patients with stable angina pectoris.

Moreover, our results fit with data obtained initially on dedicated cardiac CZT cameras. In a cohort of 10 patients undergoing stress MPI acquired with a dedicated CZT SPECT camera, Palyo et al reported no differences in the size of regional defects, LV volume, or ejection fraction with a 50% reduction protocol.17 Reduction in the acquisition time or in the injected activity is an important benefit for nuclear imaging techniques. By reducing the acquisition time, the comfort of the patient is improved, leading to reduced motion artifacts and better image quality. Reducing the injected activity lowers the dose delivered to the patient, which seems particularly interesting when using 201Tl for MPI.

Ultra-low-dose protocols are currently validated with technetium-99m labeled radiotracers on dedicated CZT SPECT cameras (injection of 1.85 MBq/kg leading to an effective dose < 1 mSv) with good sensitivity.19 The high negative predictive value of normal MPI using an ultra-low-dose protocol was also established both with technetiated radiotracers and thallium.4,20 However, artifacts are often seen with Anger and dedicated CZT cameras and lead to a reduction in the specificity of the examination.19 Attenuation artifacts were frequent in our cohort. The variable “Number of segments < 70% of maximal intensity” reported all segments with significantly altered tracer uptake, possibly due to a true perfusion defect or an artifact. The variable “Number of segments considered pathological” results from qualitative analysis including the information given by ECG-gated data and is somewhat more dependent on the experience of the reader. As attenuation correction was not applied, the average number of segments lower than 70% is high, but the number of segments considered pathological by readers was much lower (respectively 0.8, 0.8 and 0.9 for the 3 datasets), suggesting a high prevalence of attenuation artifacts. However, the ability of the readers to identify these artifacts was not impacted by the reduction in the total number of counts, as demonstrated by a stable number of segments considered pathological across the three datasets.

A recent meta-analysis reported good diagnostic performance in detecting angiographically proven coronary artery disease for both conventional SPECT (pooled sensitivity and specificity 85% and 66% respectively) and for CZT SPECT (pooled sensitivity and specificity 89% and 69% respectively), with a slightly higher accuracy for CZT SPECT.21 Computed tomography-based attenuation correction (CTAC), available with the majority of Anger cameras but only with a minority of dedicated cardiac CZT cameras, is a useful method to improve specificity. As for Anger cameras, whole-body CZT SPECT cameras are mainly equipped with CT, allowing for CTAC with a potential increase in specificity. Moreover, CT images can provide an evaluation of the coronary calcium score22 which brings a complementary appreciation of the cardiovascular risk.23

Our study has some limitations. First, our main outcome was based on protocol optimization and image quality evaluation, thus only stress acquisitions were analyzed and patient outcome was not available. A longitudinal study with clinical follow-up is needed to evaluate the performances of these new devices in terms of cardiovascular events risk prediction. Second, data were missing for three patients due to acquisition artifacts or technical issues which made post-processing impossible. Third, intra-observer variability was not evaluated in our study. We chose to focus on inter-observer variability because it is the most penalizing situation in clinical routine. Finally, this study focused only on a 201Thallium imaging protocol. Dedicated evaluations are necessary for technetium-99m labeled tracers and some data are already available in literature.15,16 Recently, Morelle et al15 reported the first comparison of performances between the new whole-body CZT SPECT/CT camera and a dedicated cardiac CZT camera for myocardial perfusion imaging using 99mTc-sestamibi. They reported a slightly poorer spatial resolution of whole-body CZT SPECT/CT camera (DNM 670CZT, GE Healthcare) compared with dedicated cardiac CZT SPECT camera (DNM 530c, GE Healthcare). Despite this difference, results obtained in patients were concordant between both cameras in terms of perfusion, EDV, ESV, EF, wall motion and thickening scores. In 19 patients that underwent sequential low-dose 99mTc-tetrofosmin both on dedicated cardiac CZT camera and whole-body CZT SPECT/CT camera, Gimelli et al16 also reported an excellent correlation for segmental myocardial uptake. To the best of our knowledge, our study is the first to address the impact of a reduction in acquisition time or injected activity when using 201Thallium for MPI on a whole-body CZT camera.

New Knowledge Gained

The current study highlights the validity of whole-body CZT SPECT/CT thallium myocardial perfusion imaging in clinical routine and demonstrates the possibility to reduce injected activity or acquisition time with equivalent image interpretation compared to a standard protocol.

Conclusion

Novel whole-body CZT SPECT/CT cameras allow to adapt protocols designed for Anger cameras in order to reduce injected activity or acquisition time by 50% while maintaining equivalent image interpretation and a good inter-observer agreement.