Pitfalls in quantitative myocardial PET perfusion II: Arterial input function

Abstract

Rationale: We aimed to define the impact of variable arterial input function on myocardial perfusion severity that may misguide interventional decisions and relates to limited capacity of 3D PET for high-count arterial input function of standard bolus R-82.

Methods: We used GE Discovery-ST 16 slice PET-CT, serial 2D and 3D acquisitions of variable Rb-82 dose in a dynamic circulating arterial function model, static resolution and uniformity phantoms, and in patients with dipyridamole stress to quantify per-pixel rest and stress cc·min-1·g-1, CFR and CFC with (+) and (-) 10% simulated change in arterial input.

Results: For intermediate, border zone severity of stress perfusion, CFR and CFC comprising 7% of 3987 cases, simulated arterial input variability of ± 10% may cause over or underestimation of perfusion severity altering interventional decisions. In phantom tests, current 3D PET has capacity for quantifying high activity of arterial input and high-count per-pixel values of perfusion metrics per artery or branches.

Conclusions: Accurate, reproducible arterial input function is essential for at least 7% of patients at thresholds of perfusion severity for optimally guiding interventions and providing high-activity regional per-pixel perfusion metrics by 3D PET for displaying complex quantitative perfusion readily understood ("owned") by interventionalists to guide procedures.

Keywords: Quantitative myocardial perfusion; cardiac PET.

Figures

Figure 1

Dynamic circulating model for reproducible “arterial” time activity curves

Figure 2

Circulating model arterial time activity curves A, B, C, and D shows essentially identical serial repeated total activity–time curves displayed as total counts on the scanner screen after serial separate injections of Rb-82 into the circulating arterial activity model as the input data to the scanner. For each of serial identical time–activity curves as input to the scanner, each scanner protocol was tested for acquiring absolute time activity curves as would be done for the arterial input in patients. Panel E shows the two acquisition protocols—the 2-minute protocol (red brackets) and the serial 15-second (blue brackets) PET acquisition frames with resulting scanner acquired time activity curve. The serial 15-second time–activity was the reference protocol as compared to activity of a timed sample drawn into a syringe during each run after Rb-82 injection and counted in the well counter

Figure 3

Bar graphs of scanner acquired model arterial activity compared to well counter activity for various doses of Rb-82 with 2D and 3D acquisition. White bars and ± values indicate one standard deviation. The red-highlighted values indicate P ≤ 0.005 compared to the highest value for the Rb-82 dose coded blue 1176 MBq (48 mCi), green 1147 MBq (31 mCi), or purple 740 MBq (20 mCi)

Figure 4

Effect on coronary flow reserve of rest and stress arterial activity increased or decreased by 10% compared to the standard optimal arterial activity in 30 subjects in each of three representative groups—normal or mildly reduced, intermediate, or worst severity of coronary flow capacity (CFC) that accounts for both rest–stress perfusion in cc·min−1·g−1 and CFR. Vertical bars and ± values indicate one standard deviation. Red-highlighted data indicate significance with P < 0.05

Figure 5

Cumulative fraction of LV in CFC severity ranges and Kolomogorov–Smirnov statistics for significant differences in histogram distributions with rest and stress perfusion increased or decreased by ±10% (gray lines) versus no change (black line) for 10 patients in each of the following groups: (A) intermediate CFC abnormalities; (B) normal or mildly reduced CFC; (C) worst CFC abnormalities. The KS statistic for all gray lines versus the black line is significant with P ≤ 0.0001. The blue and green arrows emphasize the changes in moderate to severe CFC abnormalities having greatest impact on potential interventions based on CFC severity

Figure 6

Fractional partial volume activity loss for 8 to 30 mm wide one-dimensional tree phantom imaged by GE DSTE, D710, and DMic PET-CT scanners

Figure 7

Three types of Regions of Interest (ROI) or regional boundaries in which quantitative metrics are measured ranging from large externally imposed, fixed whole artery distributions (A and B) to fixed external multi-segmental compartments (C) to per-pixel values (D) providing perfusion, CFR, and coronary flow capacity (CFC) maps of precise, actual perfusion arterial distributions for each individual as they actually are without assumed arbitrary fixed externally imposed regions of interest for which average perfusion is determined

Figure 8

Even angiogram anatomy does not define how regions of interest should be drawn for regional perfusion measurement