Multisoftware Reproducibility Study of Stress and Rest Myocardial Blood Flow Assessed with 3D Dynamic PET/CT and a 1-Tissue-Compartment Model of ⁸²Rb Kinetics

Abstract

Routine quantification of myocardial blood flow (MBF) requires robust and reproducible processing of dynamic image series. The goal of this study was to evaluate the reproducibility of 3 highly automated software programs commonly used for absolute MBF and flow reserve (stress/rest MBF) assessment with 82 Rb PET imaging. Methods: Dynamic rest and stress 82 Rb PET scans were selected in 30 sequential patient studies performed at 3 separate institutions using 3 different 3-dimensional PET/CT scanners. All 90 scans were processed with 3 different MBF quantifi-cation programs, using the same 1-tissue-compartment model. Global (left ventricle) and regional (left anterior descending, left circumflex, and right coronary arteries) MBF and flow reserve were compared among programs using correlation and Bland-Altman analyses. Results: All scans were processed successfully by the 3 programs, with minimal operator interactions. Global and regional correlations of MBF and flow reserve all had an R 2 of at least 0.92. There was no significant difference in flow values at rest (P 5 0.68), stress (P 5 0.14), or reserve (P 5 0.35) among the 3 programs. Bland-Altman coefficients of reproducibility (1.96 · SD) averaged 0.26 for MBF and 0.29 for flow reserve differences among programs. Average pairwise differences were all less than 10%, indicating good reproducibility for MBF quantification. Global and regional SD from the line of perfect agreement averaged 0.15 and 0.17 mL/min/g, respectively, for MBF, compared with 0.22 and 0.26, respectively, for flow reserve. Conclusion: The 1-tissue-compartment model of 82 Rb tracer kinetics is a reproducible method for quantification of MBF and flow reserve with 3-dimensional PET/CT imaging. Absol ute quantification of myocardial blood flow (MBF) at stress and rest with dynamic PET imaging is an important tool for clinicians and provides information complementary to relative myocardial perfusion imaging (1-3). With standard list-mode acquisition and fast image reconstruction , dynamic, gated, and standard static perfusion images can be obtained with a single injection of the radiopharmaceutical and without additional imaging time. Automated image analysis tools are required for reliable and robust clinical use of dynamic data for MBF quantification (4). The performance of several such software programs for MBF quan-tification has been reported recently (5-9), each of which uses different tracers and methods of segmenting and sampling the left ventricular myocardium and blood-pool activity to obtain input curves. Although each of these tools greatly simplifies MBF quantification, uses the same tracer kinetic model for 82 Rb-rubidium (10), and has been validated individually , the effect of different model implementations has not been characterized. Previous studies have compared different tracer kinetic models and implementations for 13 N-ammonia (11,12) but not for 82 Rb-rubidium and in particular not for 3-dimensional (3D)-mode PET, which is the current standard technology. We aimed to comp...

Cardiac 82Rb PET/CT for fast and non-invasive assessment of microvascular function and structure in asymptomatic patients with type 2 diabetes

“…Non-gated images were presented in the horizontal and vertical long axis and in the short axis, and displayed in a polar map format divided into 17 segments, as suggested by the American Heart Association [13], and normalised to peak myocardial activity (100%) [14]. Myocardial blood flow was calculated automatically using the Siemens Syngo MBF 2.3 (Siemens Medical Solutions, Malvern, PA, USA) with one-compartment tracer kinetic models for 82 Rb, including regional uptake and clearance variables, blood to myocardium spill-over and partial volume corrections [15], and the extraction curve from Lortie et al [16]. Maximal hyperaemia was induced with adenosine infused at 140 μg kg −1 min −1 for 6 min.…”

PET myocardial perfusion quantification: anatomy of a spreading functional technique

“…There is evidence of good agreement between software packages for blood flow quantification with 82 Rb [17] and 13 N-ammonia [18] studies and it has been stated that reconstruction using a wide range of software, under the assumption of a 1-tissue compartment mode, has sufficient agreement to be used interchangeably. Nevertheless, it has also been demonstrated that agreement is not optimal between estimations derived from different kinetic models [19].…”

Moving towards a better understanding of potential pitfalls in quantitative PET myocardial blood flow

“…6 Similarly, when point spread function modeling and time-of-flight reconstructions are compared to standard PET reconstructions, CFR results are consistent, but marked variations in rest and stress MBF have been reported. 7 Fortunately, when the same data set is analyzed with different automated software programs, MBF results are comparable, provided the analysis is performed with the same one-compartment model of 82 Rb kinetics. 8 Regarding variations in data acquisition, the vasodilator and timing of radionuclide injection both influence maximal hyperemia.…”