Volumetric time-resolved velocity imaging (4D flow) can be used as a single comprehensive MR sequence to quantify blood flow, describe cardiac function, and depict anatomy. However, fat signal can reduce tissue contrast, introduce high-signal-intensity artifacts from motion, and cause errors in velocity quantification^[1,2]. Furthermore, it is difficult to apply a single fat suppression technique for all applications of 4D flow: cardiac, abdomen, etc. Thus, we introduce and investigate two approaches for reducing fat signal in contrast-enhanced 4D flow imaging with minimal time penalty.

To enable the reduction of fat signal in 4D flow, two different approaches are developed:

1. Spectral-spatial RF pulse (SSRF): Fat signal can be avoided (or reduced) with a water-only spectral-spatial RF pulse^[3]. With intravenous contrast administration, fat signal only needs to be reduced with respect to signal from the contrast-enhanced blood pool. Thus, we design a 2.3-msec RF pulse to reduce the excitation flip angle for fat to be less than 2% of that of water. Details of the pulse (Figure 1ab) for 3T include 5 subpulses with spatial time-bandwidth of 8 and a minimum-phase spectral envelope. The pulse is initially designed with a slab thickness of 14 cm and flip angle of 15°, and can be adjusted accordingly for other scan prescriptions.

2. Fat/water separation: In areas with large field inhomogeneities, a more robust approach is fat/water separation. We develop a flow acquisition with one additional velocity-sensitized echo (Figure 1c). This extra echo has the same velocity-encoding gradients as one of the original echoes but with a different echo time (TE) to estimate a velocity-free field map. Two-point Dixon technique^[4] is modified to separate fat/water for all subsequent echoes. To increase robustness, field inhomogeneities are assumed to be slowly-varying in both spatial and cardiac-phase dimensions.

The two techniques were first evaluated in a static phantom study with vials of water, oil, and different concentrations of ferumoxytol. With IRB approval and informed consent, pediatric patients referred for 3T ferumoxytol-enhanced MRI were recruited and scanned on a GE MR750 scanner using a 32-channel cardiac coil. Butterfly navigators were used to measure motion and to suppress respiratory motion artifacts through a soft-gated reconstruction^[5]. With a pseudo-random view-ordering scheme^[6], a compressed-sensing-based parallel imaging reconstruction was implemented to enable high acceleration factors^[5]. Scan parameters included flip angle of 15° and minimum TE/TR. The additional echo for fat/water separation was set with TE of +0.6 msec. Specific parameters are described in figure captions.

In the phantom study, fat signal was reduced to be lower than the ferumoxytol-enhanced vials using the proposed approaches (Figure 2). High concentrations of ferumoxytol increased field variations that impacted the performance of SSRF and caused ferumoxytol-doped solutions to be partially classified as fat.

A reduction of signal from chest wall fat (and of associated motion artifacts) can be appreciated using SSRF (Figure 3). The fat/water separation technique reduced fat signal considerably in the abdominal 4D flow (Figure 4) and cardiac 4D flow scans (Figure 5). Better visualization of vessels near the chest wall fat can be appreciated. Initial flow measurements in the aortic root and main pulmonary artery from the water-only images were in agreement with measurements from the non-fat-suppressed 4D flow scan (Figure 5).

Performance of the fat signal reduction correlates with the amount of additional scan time needed. Since 4D flow already suffers from long scan durations, two approaches with minimal scan penalty (~1 msec/TR) are demonstrated to reduce fat signal sufficiently for improving image quality in ferumoxytol-enhanced scans. These developments can be extended to support 4D flow with gadolinium-based contrast agents.

The SSRF has the potential to considerably reduce image artifacts from moving fat tissue, such as the chest wall. However, the performance of this approach suffers from field inhomogeneities. Results from the fat/water separation technique will be impacted by motion in fatty tissue; thus, motion must be considered when using this approach. Partial classification of contrast-enhanced signal as fat should not impact velocity measurements, because this effect should be consistent among all flow echoes.

Initial results indicate that these approaches do not degrade flow quantification accuracy. Future work will include evaluating whether these approaches will improve velocity measurements. We hypothesize that the SSRF pulse will reduce aliasing of fat signal in vessels of interest and that the fat/water separation technique will improve velocity measurements in vessels surrounded by fatty tissues.

Feasibility of two practical methods for reducing fat signal has been demonstrated for 4D flow. Incorporating these methods will potentially improve tissue contrast and reduce velocity errors associated with fat.