Direct 4D Parametric Imaging in Dynamic Myocardial Perfusion PET

Dynamic myocardial perfusion (MP) PET imaging followed by tracer kinetic modeling allows quantification of myocardial blood flow, thus enabling computation of the coronary flow reserve, with considerable clinical potentials. Nonetheless, utilization of short dynamic frames can lead to noisy flow estimates, an issue that is further amplified in parametric imaging at the voxel level. Our purpose is to utilize an enhanced image reconstruction framework to better address this issue. We implemented a novel 4D reconstruction scheme to directly estimate MP parametric images from the measured dynamic datasets. This included formulation of a 4D log-likelihood objective function relating the kinetic parameters to the projection datasets, and implementing numerical methods to optimize the objective function. We also utilized the technique of optimization transfer to enable more convenient and reliable parametric imaging. We simulated MP Rb-82 PET projection datasets based on the XCAT phantom utilizing patient-based time activity curves for the various organs and clinically realistic noise levels, followed by noise vs. bias analysis. The proposed direct 4D methodology was shown to outperform conventional indirect parametric imaging, reducing noise by over 50% (matched bias), with further reductions of 15% in noise and a factor of five speed-up when optimization transfer was additionally utilized. Direct 4D PET image reconstruction is a viable and very promising approach towards robust parametric MP PET imaging at the individual voxel level.