Direct parametric estimation of blood flow in abdominal PET/CT within an EM reconstruction framework

Parametric estimation of perfusion using [15O]-H2O is an important biomarker for assessing treatment response in oncology clinical trials involving anti-vascular and anti-angiogenic agents. Traditionally the set of dynamic images are reconstructed independently, followed by post-reconstruction kinetic modelling. This methodology results in sub-optimal and often noisy end point parameters if voxel-by-voxel kinetic modelling is used. In [15O]-H2O scans, the extremely short temporal frames and the rapid decay of the tracer results in further signal-to-noise ratio (SNR) reduction. Direct 4-D reconstruction methods have the potential to reduce noise and improve parameter estimates by combining image reconstruction and kinetic modelling in a unified process. In this work we implement a direct parametric reconstruction using an EM framework and apply it on a real [ 15O]-H2O PET/CT dataset to derive parametric images of perfusion (f), clearance rate (k2) and fractional blood volume (Va). Results show substantial variance reduction both in perfusion and k2 images compared to the post-reconstruction kinetic analysis, especially in areas of increased perfusion. This results in increased tumour-to-background contrast and improved delineation of organ boundaries. Although further analysis in needed, direct reconstruction methods have the potential to be applied in a wide variety of oncology PET/CT studies with the improvements highly depended on the tracer and the system under study. © 2010 IEEE.