Sensitivity of quantitative myocardial dynamic contrast-enhanced MRI to saturation pulse efficiency, noise and t(1) measurement error: Comparison of nonlinearity correction methods

Purpose To compare methods designed to minimize or correct signal nonlinearity in quantitative myocardial dynamic contrast-enhanced (DCE) MRI. Methods DCE-MRI studies were simulated and data acquired in eight volunteers. Signal nonlinearity was corrected using either a dual-bolus approach or model-based correction using proton-density weighted imaging (conventional or dual-sequence acquisition) or T1 data (native or bookend). Scanning of healthy and infarcted myocardium at 3 T was simulated, including noise, saturation imperfection and T1 measurement error. Data were analyzed using model-based deconvolution with a one-compartment (mono-exponential) model. Results Substantial variation between methods was demonstrated in volunteers. In simulations the dual-bolus method proved stable for realistic levels of saturation efficiency but demonstrated bias due to residual nonlinearity. Model-based methods performed ideally in the absence of confounding error sources and were generally robust to noise or saturation imperfection, except for native T1 based correction which was highly sensitive to the latter. All methods demonstrated large variation in accuracy above an over-saturation level where baseline signal was nulled. For the dual-sequence approach this caused substantial bias at the saturation efficiencies observed in volunteers. Conclusion The choice of nonlinearity correction method in myocardial DCE-MRI impacts on accuracy and precision of estimated parameters, particularly in the presence of nonideal saturation