PURPOSE

To develop and evaluate a phase-contrast MRI (PC-MRI) technique with hybrid one- and two-sided flow-encoding and velocity spectrum de-aliasing using low spatial resolution reference four-point PC-MRI for accelerated blood flow and velocity measurement.

METHODS

In conventional PC-MRI, the temporal sampling period between two cardiac phases is 4*TR*views-per-segment (VPS). In the recent proposed HOTSPA technique(1), each cardiac phase is shortened by 50% to 2*TR*VPS. This is achieved by applying two-sided flow encoding (FE) in the z-direction and alternating two orthogonal in-plane FE directions (i.e. X+Y and X-Y directions) in two consecutive cardiac phases. Specifically, the two-sided FE_z data includes a temporal modulated velocity waveform $$$\phi_{0}(n)\pm\phi_{v,z}(n)=\phi_{0}(n)+e^{i*\pi*n}*\phi_{v,z}(n)$$$, where $$$\phi_{0}(n)$$$ is the the flow-compensated (FC) background phase, $$$\phi_{v,z}(n)$$$ is the z-directional velocity signal phase and n is the cardiac phase number. The hybrid one-sided (x-direction) and two-sided FE (y-direction) are used to generate a $$$\phi_{0}(n)+\phi_{v,x}(n)\pm\phi_{v,y}(n)=\phi_{0}(n)+\phi_{v,x}(n)+e^{i*\pi*n}*\phi_{v,y}(n)$$$ waveform in the two in-plane directions. $$$\phi_{v,x/y}(n)$$$ is the x/y-directional velocity signal phase. If one applies a temporal Fourier transform of the temporal modulated velocity waveforms, the spectra of each encoding component will overlap with half of the frequency support shift in Fourier velocity domain, as shown in Fig. 1d. When the bandwidth is sufficient large for both spectra, a temporal filter can easily separate the overlapping signals. However, when the bandwidth is limited and the two spectra are severely overlapped, we propose to use the waveforms from an additional low spatial resolution reference four-point acquisition to estimate the ratio between F{V_x(f)} and F{V_y(f)} (temporal Fourier transform of x- and y-directional velocity waveforms), as shown in Fig. 1c-d. The hypothesis is based on the assumption that the phase signal is mainly carried by center k-space(2). Thus, we can differentiate the aliased Fourier velocity spectra by using the ratio of the two overlapping spectral components calculated from the low-resolution reference acquisition (Fig. 1d). Six volunteers were scanned at the common carotid arteries (CCAs) using two sequences: 1) 2D reference four-point PC-MRI with three FE directions, i.e. 2D FC/3FE; 2) 2D HOTDEAL strategy. The sequences parameters were: TE/TR=4.0/6.3ms, Flip Angle=20°, VENC=120cm/s, FOV=256*176cm², Acquired matrix size=256*176 for fully sampled acquisition(Fig. 1b) and 24 lines were kept for center k-space acquisitions(Fig. 1a), slice thickness=7mm, VPS=1 and 2 for FC/3FE, 2 and 4 for HOTDEAL. The net acceleration rate of HOTDEAL was 1.6 because of the extra acquisition of reference four-point PC-MRI.

RESULTS

Fig. 2a-b & Fig. 3a-b show examples of through-plane mean velocity and peak velocity measurements (average and maximum within the vessel lumen, respectively) of the CCA comparing two sets of different measurements: 1) the 1-VPS 2D FC/3FE(blue) with 25.28 ms temporal sampling period and the 2-VPS 2D HOTDEAL(red) with 25.28 ms temporal sampling period; 2) the 2-VPS 2D FC/3FE PC-MRI(blue) with 50.56 ms temporal sampling period and the 4-VPS 2D HOTDEAL(red) with 50.56 ms temporal sampling period. The velocity measurements (both through-plane mean velocity and peak velocity) were similar between the 2-VPS 2D HOTDEAL and the 1-VPS 2D FC/3FE, as well as the 4-VPS 2D HOTDEAL and the 2-VPS 2D FC/3FE, although HOTDEAL acquisition speed was 1.6 times faster. However, low temporal resolution FC/3FE (2-VPS) and HOTDEAL (4-VPS) underestimated the peak velocity due to its long temporal footprint and temporal sampling period. Across the six subjects, using the 2D 1-VPS FC/3FE as the reference, the bias of 2D HOTDEAL was -0.08 mL (-1.54% relative bias error) with 95% Confidential Interval (CI) [-0.6, 0.4] mL for total volumetric flow and -2.32 cm/s (-2.54% relative bias error) with 95% CI [-9.2, 4.6] cm/s for peak velocity (Bland-Altman plots of Fig. 2c-d). The Bland-Altman(Fig. 3c) comparisons between 2-VPS FC/3FE and 4-VPS HOTDEAL showed good agreements of total volumetric flow measurements. The bias of total volumetric flow was -0.05 mL (-0.97% relative bias error) with 95% CI [-0.3, 0.2] mL.

DISCUSSION

The HOTDEAL enables an approximately 50% shorter temporal sampling period for each cardiac phase. It allows two spectra share the same bandwidth, which improves the acquisition speed by reducing the temporal sampling rate. Furthermore, the HOTDEAL technique can be combined with other fast imaging techniques, such as parallel imaging, compressed sensing, and HOTSPA (1) to further increase the acquisition speed.

CONCLUSION

The proposed HOTDEAL technique accelerates PC-MRI using low spatial resolution reference four-point PC-MRI for overlapping spectra de-aliasing in Fourier velocity spectrum domain and it maintains the measurement accuracy for total volumetric flow and peak velocity quantification.

Acknowledgements

NIH R01HL127153

References

1. Wang D., et.al. MRM 2016 Aug 9. doi: 10.1002/mrm.26366. [Epub ahead of print] 2. Markl M., et.al. MRI 2001; 19: 669-676