Session S51.2

Iterative EM Reconstruction of Cardiac Small Animal PET Images Using System Point Spread Function Modelling and MAP with Anatomical Priors

AE Spinelli*, G Fiacchi, D D'Ambrosio, P Cilibrizzi, C Lamberti,
G Baldazzi, S Boschi, R Franchi, M Marengo

Scuola di Specializzazione in Fisica Sanitaria
Bologna, Italy

Introduction: The spatial resolution of a small animal Positron Emission Tomography (PET) system depends on several parameters such as for example detectors dimensions, positron range etc. By using for example estimation maximization (EM) iterative algorithms it is possible to include in the reconstruction process such effects in order to obtain a better estimate of the image. The use maximum a posteriori (MAP) reconstruction based on anatomical prior provides also a smooth reconstructed PET images while preserving at the same time anatomical edges. The goal of this work was to improve the spatial resolution of small animal PET images by introducing in the reconstruction process the true system point spread function (PSF) and an anatomical image prior.
Material and Methods: In order to evaluate the proposed image reconstruction approach, simulations were carried on using a mouse heart phantom (myocardium and ventricle) and a comparison between standard EM reconstruction and EM+MAP+PSF modelling reconstruction was performed. The system PSF was assumed to be a Gaussian distribution, and its Full Width Half Maximum (FWHM) was modelled to be spatially variant in order to simulate the parallax effect due to the position of the source inside the scanner field of view (FOV). The true system PSF was measured experimentally by moving a small (0.2 mm) 18-F source along scanner y axis. The FWHM varies between 1.5 mm (at the centre of FOV) and 2.0 mm at the edge. Comparison have been made in terms of contrast improvement between the myocardium and left ventricle using 80 iterations for both standard EM and EM+MAP+PSF algorithms.
Results: An analysis of line profiles drawn across the myocardium (M) and left ventricle (LV) showed that difference between M and LV was equal to 15 MBq/cc for standard EM, while with the EM+PSF+MAP algorithm the difference in amplitude was equal to 34 MBq/cc. A visual comparison of the images reconstructed with the two approaches showed that the EM+MAP+PSF reconstructed images look like much sharper (as suggested by the line profiles) and are very close to the true ones.
Conclusions: Results shows that the proposed EM+MAP+PSF image reconstruction method provides a significant improvement in contrast and sharpness of small animals PET cardiac images.

(Abstract Control Number: 262)