Abstract
BACKGROUND: Quantitative perfusion cardiovascular magnetic resonance (QP-CMR) allows the generation of pixel-wise myocardial blood flow (MBF) maps using model-based deconvolution with several models including Tofts, modified-Tofts, and Fermi function models. However, the accuracy of pixel-wise MBF mapping has not been fully investigated in humans. The aim of this study was to evaluate the accuracy of advanced QP-CMR using (15)O-water positron emission tomography (PET) as a reference. METHODS: Thirty-nine patients (29 men, 68±11years) with known or suspected coronary artery disease underwent both CMR including stress and rest QP-CMR and (15)O-water PET at a median interval of 13 days. QP-CMR was performed using dual-sequence technique and a single bolus of gadolinium contrast agent during adenosine triphosphate stress and at rest. MBF maps were generated using three different model-based deconvolution techniques as follows: Tofts, modified-Tofts, and Fermi function models. Agreement of MBF and myocardial perfusion reserve (MPR) between QP-CMR and (15)O-water PET was evaluated using Pearson's correlation, Bland-Altman analysis, and intraclass correlation (ICC). The ability of CMR-derived stress MBF and MPR to detect PET-defined abnormal myocardial perfusion (stress MBF ≤2.3 mL/min/g and MPR ≤2.5) was evaluated by receiver operating characteristic (ROC) analysis. RESULTS: CMR-derived MBF showed a good linear correlation with (15)O-water PET-derived MBF in each of the Tofts, modified-Tofts, and Fermi function models (r = 0.776, 0.752, 0.784, respectively; p<0.001 each) at the patient level. Bland-Altman analysis demonstrated measurement biases for MBF between CMR and (15)O-water PET of 0.31±0.70, 0.05±0.63, and 0.26±0.68 mL/min/g for the Tofts, modified-Tofts, and Fermi function models, respectively. ICCs were 0.734, 0.747, and 0.750, respectively. The area under the ROC curves for stress MBF derived from the Tofts and Fermi function models (0.921 and 0.914, respectively) was significantly higher than that derived from the modified-Tofts model (0.861; p = 0.003 for both). However, there was no significant difference between the Tofts and Fermi function models (p = 0.618). CONCLUSION: Advanced QP-CMR using three different model-based deconvolution techniques demonstrated strong agreement with (15)O-water PET. Of these techniques, the Fermi function and Tofts models were more effective in detecting abnormal myocardial perfusion as determined by (15)O-water PET. Considering our results, the model complexity, and its technical availability, the Fermi function model may possess a practical advantage.