Comparison of 2-dimensional and 3-dimensional 82Rb myocardial perfusion PET imaging

We compared 2-dimensional (2D) and 3-dimensional (3D) (82)Rb PET imaging in 3 different experiments: in a realistic heart-thorax phantom, in a uniformity-resolution phantom, and in 14 healthy volunteers.

A nonuniform heart-thorax phantom was filled with 111 MBq of (82)Rb injected into the left ventricular (LV) wall. In the LV wall of the cardiac phantom, 3 inserts-1, 2, and 3 cm in diameter-were placed to simulate infarcts. A standard rest cardiac PET imaging protocol in 2D and 3D modes was used. Following the same protocol, a uniformity-resolution phantom with uniformly distributed activity of 1,998 MBq and 740 MBq of (82)Rb in water was used to obtain 2D PET images and 3D PET images, respectively. All 2D volunteer studies were performed by injecting 2,220 MBq of (82)Rb intravenously. For half the volunteers, 3D studies were performed with a high dose (HD) (2,220 MBq) of (82)Rb; for the remainder of the 3D studies, a low dose (LD) (740 MBq) of (82)Rb was used. In the 2D and LD 3D studies, there was a delay of 2 min and 3 min, respectively, followed by a 6-min acquisition. In the HD 3D volunteer studies, there was a delay of 5 min followed by a 6-min acquisition. Circumferential profiles of the short-axis slices and the contrast of the inserts were used to evaluate the cardiac phantom PET images. The transaxial slices from the uniformity-resolution phantom were evaluated by visual inspection and by measuring uniformity. The human studies were evaluated by measuring the contrast between LV wall and LV cavity, using linear profiles and visual analysis.

In the cardiac phantom study, circumferential profiles for the 2D and 3D images were similar. The contrast values for the 1-, 2-, and 3-cm inserts in the 2D study were 0.19 +/- 0.03, 0.34 +/- 0.05, and 0.61 +/- 0.03, respectively. The respective contrast values in the 3D study were 0.15 +/- 0.02, 0.36 +/- 0.04, and 0.52 +/- 0.05. In the uniformity-resolution phantom study, the coefficients of variation, calculated for a representative uniform slice, were 5.3% and 7.6% for the 2D and 3D studies, respectively. For the 7 volunteers on whom HD 3D was used, the mean 2D contrast was 0.33 +/- 0.08 and the mean HD 3D contrast was 0.35 +/- 0.08 (P = not statistically significant). For the other 7 volunteers, on whom LD 3D was used, the mean 2D contrast was 0.39 +/- 0.06 and the mean LD 3D contrast was 0.39 +/- 0.10 (P = not statistically significant). In the tomographic slices, the 2D and 3D images and polar plots were similar.

When obtained with a PET system having a high counting-rate performance, 2D and 3D (82)Rb PET cardiac images are comparable. LD 3D imaging can make (82)Rb PET cardiac imaging more affordable.