Three-dimensional correction for spillover and recovery of myocardial PET images.

UNLABELLED PET permits the quantification of myocardial blood flow, but is hampered by the limited spatial resolution of PET images. METHODS We evaluated two methods for the correction of resolution effects in PET perfusion 13NH3-ammonia images. In one model, the spillover and recovery coefficients are estimated in the kinetic modeling analysis. The new, second model uses an explicit delineation of the left ventricular wall and a convolution model for the system point spread function to compute the regional values of the spillover and recovery coefficients. RESULTS The new method is validated with phantom measurements. The two methods are evaluated on animal experiments using 13NH3-ammonia. Both two- and three- compartment models were used to compute absolute flow values. Excellent linear correlations with microsphere data were obtained. The slope of the regression line was lower for corrections based on kinetic modeling as compared to convolution-based correction. In animal experiments, recovery coefficients of 59% for the myocardial wall and 86% for the blood pool were obtained. Spillover from the blood pool into the myocardial was was 14%. CONCLUSION The new correction method strongly suppresses spillover and recovery effects due to limited resolution.