Comparison of myocardial perfusion and permeability estimates from dynamic contrast-enhanced MRI with three quantitative analysis methods
Wang Jing1, Yudong Zhang2, Yang Fan3, HaiBin Shi2, and Xiaoyan Liu1

1Center for Medical Device Evaluation, CFDA, Beijing, China, People's Republic of, 2Department of Radiology, the First Affiliated Hospital with Nanjing Medical University Nanjing, Nanjing, China, People's Republic of, 3GE Healthcare, MR Research China, Beijing, China, People's Republic of


This study compares three methods for quantitatively analyzing dynamic contrast-enhanced MRI datasets for the extraction of myocardial perfusion, permeability, and other hemodynamic parameters. The two-compartment exchange model and the adiabatic approximation to the tissue homogeneity models are compared with reference to model-free deconvolution. The goal was to determine if these three models could reliably estimate hemodynamic parameters and be able to distinguish between normal patients and those with hypertrophic cardiomyopathy. Results demonstrate that all three methods yield consistent estimates for perfusion, blood volume, and mean transit times, and that these parameters were significantly different for the hypertrophic heart.


First-pass dynamic contrast-enhanced (DCE) MRI with kinetic analysis models was know as a reliable tool to quantify myocardial F1, but not yet reached agreement for myocardial permeability. Recently, an adiabatic approximation to the tissue homogeneity model (AATH) and two-compartment exchange model (2CXM), have been recognized as promising ways for F and PS quantification in brain, breast and prostate2. This study was to explore the feasibility of cardiac function evaluation by first-pass DCE-MRI with 2CXM and AATH in healthy subjects and hypertrophic cardiomyopathy (HCM) patients.


All examinations were performed on a 1.5 T GE MRI (14 HCM patients, 12 healthy subjects, no age difference). For MF analysis, h(t) was determined by deconvolving AIF to tissue enhancement signal, then calculated F, Vp and MTT maps. For 2CXM and AATH, the plots of F, Vp, MTT, E, and PS were performed using Levenberg-Marquardt nonlinear least squares algorithm. Because ventricular septum is the most representative place for characterizing an asymmetrical hypertrophy in HCM patients, regional F, Vp, MTT, E and PS were calculated using manual segmentation in middle short-axis slices. Bland-Altman plot was adopted to verify the concordance of 2CXM and AATH with MF (the average of F, Vp and MTT as X-axis, the difference as Y-axis). Mean difference, 95% limits of agreement, and the Pearson correlation coefficients between both methods were reported. With regards to the validation of model outputs, the quantitative cardiac parameters (F, Vp, MTT, E and PS) was compered between normal and HCM groups by independent-samples t test.


Fig. 1 shows representative myocardial DCE results estimated by MF, 2CXM and AATH. Fig. 2 shows the ROI-based fitting results of all three methods. Bland-Altman plot (Fig.3) shows that the mean bias of F between the three methods is 0.20 mL/g/min (-0.25 to 0.64 mL/g/min, 95% confidence interval [CI]). There are one sample in 2CXM below, and one in AATH above the 95% limits of agreement, demonstrating high concordance of both methods with MF in F quantitation. The correlation plots show that the slop of Y (F2CXM) versus X (FMF) is 1.42 (intercept = -0.19 mL/g/min, ρ = 0.88), and the slop of Y (FAATH) versus X (FMF) is 1.23 (intercept = 0.12 mL/g/min, ρ = 0.82). The mean bias of Vp between all three methods is 0.02 mL/g (-0.11 to 0.08 mL/ g of 95% CI). There are two samples in 2CXM and one in AATH below the 95% limits of agreement. The slop of Y (Vp2CXM) versus X (VpMF) is 1.28 (intercept of -0.04 mL/g, ρ = 0.70), and the slop of Y (VpAATH) versus X (VpMF) is 1.15 (intercept = -0.01 mL/g, ρ = 0.44). The mean bias of MTT of all three methods is 2.8 s. (-8.7 to 14.4 s of 95% CI). There are two samples in 2CXM below the 95% limits of agreement and non for AATH. The slop of Y (MTT2CXM) versus X (MTTMF) is 1.28 (intercept = -5.96s, ρ = 0.60), and the slop of Y (MTTAATH) versus X (MTTMF) is 0.59 (intercept = 1.83 s, ρ = 0.39). For myocardial permeability comparison between AATH and 2CXM, mean bias of E is 35.1% (-10.8% to 80.7% of 95% CI), and mean bias of PS is 0.23 mL/g/min (-0.75 to 1.20 mL/g/min of 95% CI). However, there is no statistically significant correlation in E or PS between both methods (p > 0.05). Fig. 4 shows the representative parameters acquired by AATH in a HCM patient. The hypertrophic myocardium shows evident delay-enhancement in ventricular septum and inferior wall (arrow) from the short-axis delay-enhanced T1-weighted image. For permeability maps derived from AATH, it shows inhomogeneous decrease in regional F, combining with inhomogeneous increase in E and Tc, extremely in these regions with replaced myocardial fibrosis (arrows), indicating an impaired blood supply and structural abnormality of involved myocardium.


This study demonstrated that myocardial perfusion can be well obtained from first-pass DCE-MRI with currently proposed approaches. All the three models show high accordance in determining myocardial F (coefficient correlation r > 0.8). With regards to quantitation of myocardial permeability, although myocardial E and/or PS obtained by 2CXM and AATH reflected significant difference between HCM and normal group, there is yet great inconsistence of the results between the proposed two methods, indicating the limitation of current imaging techniques in evaluation of myocardial permeability in physiopathologic conditions.


No acknowledgement found.


[1] Coelho-Filho OR,et al. MR myocardial perfusion imaging. Radiology 2013. [2] Korporaal JG, et al. Tracer kinetic model selection for dynamic contrast-enhanced computed tomography imaging of prostate cancer. Invest Radiol 2012.


Fig.1: Typical myocardial perfusion and permeability maps from a HCM patient. Top row: myocardial F obtained by MF(a), 2CXM(b) and AATH(c); Middle: myocardial MTT obtained by MF(d), 2CXM(e) and AATH(f); Bottom: myocardial PS obtained by 2CXM(g) and myocardial E obtained by 2CXM(h) and AATH(i).

Fig.2: ROI-based analysis of myocardial perfusion and permeability estimated by MF (a), 2CXM (b), and AATH (c), respectively. The black dot-lines are raw data and red dot-lines are fitted curves.

Fig.3: Left: Bland-Altman plots of myocardial F (a), Vp (c) and MTT (e). Y-axis is the mean bias and X-axis is the average; Right: Passing & Bablok regression analysis of myocardial F (b), Vp (d) and MTT (f) (black, 2CXM vs. MF; red, AATH vs. MF).

Fig.4: The representative AATH results from a HCM patient. The DE image (a) shows extremely thickened LV wall, and patch-like DE in the ventricular septum and inferior wall(arrow). The fibrotic myocardium is characteristic with markedly decreased F(b), increased E(c) and Tc(d), indicating an impaired blood supply and structural abnormality.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)