Optimization of high resolution contrast-enhanced magnetic resonance angiography (CE-MRA) is a non-trivial problem due to the linked interactions of acquired voxel size, sequence duration, bolus timing, contrast recirculation, and desirability to maintain R1 enhancement throughout the acquisition of k-space without inducing R2*-related reduction of signal. Some investigators indicate superior resolution and image contrast may be achievable with slower gadolinium injection rates than those typically applied. We seek to further evaluate this hypothesis.

We applied a computer-based model implemented in MATLAB (Mathworks, Natick, MA) to measure both resolution and image contrast through analysis of generated modulation transfer functions (MTF). [1] Contrast injection volume and rate, cardiac output, magnetic field strength, and elliptical-centric image acquisition parameters were varied over a wide gamut. We modeled the effects of gadobenate dimeglumine, although the model allows for simulation of other gadolinium formulations. Recirculation was modeled, with parameters chosen to approximate levels observed in healthy volunteers (unpublished data), with baseline signal-to-noise ratio (SNR) ranges derived from clinical renal and carotid examinations. Ultimately, a MTF was generated for each discrete set of parameters. Resolution at which 50% absolute modulation transfer is achieved (MTF50%), maximum SNR achieved at the optimal (i.e., minimum MTF50%) injection rate (SNRmax), and area under the modulation transfer curve (AUC) were recorded for each run, with parameter combinations resulting in maximum SNRs of less than 10 excluded from analysis as non-diagnostic. [2]

Each set of 0.1-5.0 mL/sec injection rate data acquired at a particular combination of parameters was analyzed, with the injection rates resulting in minimal resolution loss (minimum MTF50%) and maximal image contrast (maximum AUC) recorded. Modulation transfer curves were exported directly, AUC and MTF50% data were visualized as a scatter plot, and 3-D surface renderings were created from these data.

Modulation transfer functions vary significantly with injection rate and imaging parameters (Figure 1). Generally, the height of the initial plateau (left side of graph) represents the maximum contrast concentration during the bolus, and varies surprisingly little with injection rate. The subsequent downslope and nadir (right side of graph) represent recirculatory effects. Acquisition at 1.5T (vs. 3.0T) does not significantly alter optimal injection rates (not depicted), although it does result in lower observed SNRs. Maximum area under the MTF curve typically occurs at slightly higher injection rates than does minimum MTF50% (Figures 2, 3, and 4). Optimal resolution for typical breath-hold renal CE-MRA is achieved with injection rates on the order of 0.5-1.0 mL/s, dependent on contrast injection volume, whereas for carotid these rates are on the order of 0.2-0.3 mL/s (Figure 3). Maximum SNR varies with both contrast injection volume and cardiac output, with slightly greater variance across the gamut for carotid imaging parameters (Figure 5).

Optimization of spatial resolution and vessel contrast in CE-MRA is non-trivial. In particular, increasing gadolinium concentration does not always lead to increased signal intensity. Recent work indicates that the whole blood gadolinium concentration must remain at a level that produces high R1 relaxation without inducing R2*-related reduction of signal. [3, 4]

Most prior investigation of the relationship between gadolinium injection rate in CE-MRA and resulting resolution and image contrast has been hindered by use of subjective measures of image quality. Multiple extant studies in which the authors attempted to define imaging parameters to optimize CE-MRA image quality also have the limitation of no systematic variation in injection rate, or the choice of several discrete injection rates.

Using a computer-based model allowed us to explore a wide gamut of injection rates, contrast volumes, and cardiac outputs, and we created objective metrics based on the modulation transfer function. We found that cardiac output, contrast injection volume, and image acquisition times all affect the optimum injection rate for which maximum area under the MTF curve and resolution is achieved. This work suggests optimal image resolution is achieved at intuitively slow injection rates (0.2-1.0 mL/s, dependent on imaging parameters). Higher injection rates rely purely on recirculation to maintain R1 enhancement during peripheral k-space acquisition, and this modeling suggests that this leads to significant resolution loss.

This work suggests that slower injections better optimize resolution by virtue of the longer injection duration, allowing for greater uniformity of R1 enhancement throughout k-space acquisition. This increased resolution comes with only a minimal loss of SNR due in part to the non-linearity of R1 relaxivity and the R2* degradation seen with high gadolinium concentrations, as can be seen graphically in our modulation transfer function curves (Figure 1).