Purpose
To compare the precision and inter-observer agreement of ventricular volume, function and mass quantification by three-dimensional time-resolved (4D) flow MRI relative to cine steady state free precession (SSFP).
Materials and Methods
With research board approval, informed consent, and HIPAA compliance, 22 consecutive patients with congenital heart disease (CHD) (10 males, 6.4±4.8 years) referred for 3T ferumoxytol-enhanced cardiac MRI were prospectively recruited. Complete ventricular coverage with standard 2D short-axis cine SSFP and whole chest coverage with axial 4D flow were obtained. Two blinded radiologists independently segmented images for left ventricular (LV) and right ventricular (RV) myocardium at end systole (ES) and end diastole (ED). Statistical analysis included linear regression, ANOVA, Bland-Altman (BA) analysis, and intra-class correlation (ICC).
Results
Significant positive correlations were found between 4D flow and SSFP for ventricular volumes (r = 0.808–0.972, p<0.001), ejection fraction (EF) (r = 0.900–928, p<0.001), and mass (r = 0.884–0.934, p<0.001). BA relative limits of agreement for both ventricles were between −52% to 34% for volumes, −29% to 27% for EF, and −41% to 48% for mass, with wider limits of agreement for the RV compared to the LV. There was no significant difference between techniques with respect to mean square difference of ED-ES mass for either LV (F=2.05, p=0.159) or RV (F=0.625, p=0.434). Inter-observer agreement was moderate to good with both 4D flow (ICC 0.523–0.993) and SSFP (ICC 0.619–0.982), with overlapping confidence intervals.
Conclusion
Quantification of ventricular volume, function and mass can be accomplished with 4D flow MRI with precision and inter-observer agreement comparable to that of cine SSFP.