Quantification of myocardial blood flow using 13N-ammonia and PET: comparison of tracer models

Several tracer kinetic methods have been proposed for quantification of regional myocardial blood flow (MBF) with 13N-ammonia and PET. Merits and limitations specific to each approach, however, generally are not clear, because they have not been evaluated in the same experimental environment. Therefore, we compared six different commonly used methods (11 modifications) to characterize the accuracy of each approach. The methods included the two-parameter model (method 1), the modified two-parameter model (method 2), the four-parameter model (method 3), the graphical analysis (method 4), the first-pass extraction method (method 5) and the dose uptake index (DUI; method 6).

Eleven studies in four dogs, 16 studies in eight healthy human volunteers and 14 studies in seven patients were performed using 13N-ammonia and PET. MBF in dogs was varied with dipyridamole and coronary occlusions and was measured independently and simultaneously with microspheres. Volunteers and patients were studied at baseline and after dipyridamole. MBF and DUI were estimated using a time-activity curve (Qi[t]) derived from dynamic images and regions of interest (ROls) and using the six methods. DUI was defined as Qi(t = 2 min) x weight/dose.

MBF estimated by methods 1-5 correlated well with microsphere MBF in dogs. MBF estimates by method 1 correlated well with those by methods 2, 4 and 5 and to a lesser degree with those by method 3 in both dog and human studies. DUI correlated poorly with MBF by microspheres and by methods 1-5 in both dog and human studies. MBF estimates by method 3 showed larger dispersion (SD/mean flow) and higher sensitivity to metabolites correction in arterial blood than those by methods 1, 2, 4 and 5.

MBF can be measured accurately using 13N-ammonia PET and tracer kinetic methods. DUI is a poor indicator of MBF values. The results indicate that preference should be given to the two-parameter model, incorporating geometrical ROI representation (method 2) among the compartment models, and to the graphical analysis (method 4) among the noncompartmental approaches.