Serial changes in myocardial perfusion may represent an important marker of disease progression or regression or the effects of therapy for patients with coronary artery disease (CAD). Quantitative methods have not been developed for the assessment of serial changes in perfusion. The objective of this study was to use receiver operator characteristic (ROC) analysis to determine the sensitivity and specificity of direct paired comparisons (DPCs) to detect changes in absolute myocardial perfusion measured with 82Rb PET.
Methods: Repeated dynamic 82Rb PET scans were obtained on 8 dogs at rest and during hyperemia induced with dobutamine (n = 4) or atrial pacing (n = 4). Radiolabeled microspheres were used to verify perfusion changes. Polar maps of absolute 82Rb retention and associated SD were estimated from the dynamic images. Paired comparisons were then performed using a t test on each of the 532 polar map sectors. Rest-rest and stress-stress differences were used to assess specificity and reproducibility, and stress-rest differences were used to assess sensitivity.
Results: 82Rb retention differences of 20% over baseline were detected with 85%-90% sensitivity and specificity, using the optimal DPC probability value and image smoothness. The average 82Rb retention differences correlated well with microspheres (r = 0.74; P = 0.001). Reproducibility of the mean retention values was 4.7% +/- 2.1%. As reproducibility varies, the DPC probability value can be adjusted to maintain specificity. These ROC results are directly applicable to other image modalities that produce measurements with similar SEs (3.7% +/- 0.9%).
Conclusion: The developed method of DPCs is sensitive and specific for the detection of changes in absolute myocardial perfusion measured with 82Rb PET.