Cardiac motion-corrected quantitative dynamic Rb-82 PET imaging

Objectives In the context of dynamic cardiac imaging, gating information is not routinely utilized. Our goal is to introduce gated motion correction within dynamic Rb-82 reconstruction and to evaluate its effect on the quantitative estimation of tracer transport between blood and myocardial tissue.

Methods Making use of the list-mode acquisition capability along with gating information, we propose to first estimate motion from reconstructed single-frame, multi-gated images obtained by binning the acquired data, starting after initial blood uptake (i.e. from ~0.5min to 2min). Next, starting from time of injection, for each dynamic frame, the gated data would be combined within a 4D EM reconstruction algorithm that incorporates the estimated motion. To validate, stress time activity curves of blood pool, myocardium, and other organs, as extracted from Rb-82 PET images of 5 normal patients, were averaged and smoothed, and used to simulate a 5D NCAT phantom with both gates (8 gates) and dynamic frames (10 frames*12 sec). Analytical simulation was performed to simulate gated, dynamic noise-free PET data, which were scaled to clinical count levels before 30 Poisson noise fluctuations were created. Image reconstruction was performed, without and with motion correction for each dynamic frame. Kinetic rate constants were extracted from the noise-free and noisy reconstructed images using a two-compartment model, and compared between the conventional and proposed methods.

Results With gate-motion-corrected (4D) reconstruction for each dynamic frame, the estimated K1 values in each of the five regions (anterior, septal, inferior, lateral, and apex) were shown to be significantly increased relative to no motion-correction in both noise-free and noisy cases (p<0.0001). The average increase of K1 in both the noise-free and noisy case was 33% over the 5 regions, with the motion-corrected estimation close to the truth (5% error).

Conclusions Incorporation of cardiac motion correction within dynamic Rb-82 PET imaging was shown to improve the accuracy of estimated K1 values compared to no motion correction