Many clinical studies such as whole-body water/fat quantification¹, peripheral vascular angiography², detection of tumor metastases³, and multi-contrast anatomical imaging⁴ demand either whole-body coverage or a field of view (FOV) larger than the uniform z extent of the scanner bore. Continuously Moving Table (CMT) MRI is an imaging technique that enables rapid whole-body MR examination demanded in such studies^5,6. In CMT MRI, data are acquired with concurrent uniform velocity z direction motion of the patient table. However, unlike the alternative of multi-station whole-body scanning, a limitation of CMT MRI is that the pre-scan RF power calibrations are performed at one location of the body and maintained throughout the scan for the full FOV. This results in non-optimal RF transmit power settings in other sections of the body and RF shading artifacts that vary across the length of the FOV. To mitigate these artifacts, a rapid and robust measurement of the whole-body B₁⁺ field (or flip angle α assuming proportionality) is necessary. In this abstract, we present an approach for rapid whole-body measurement of α that combines golden angle radial CMT MRI and dual TR actual flip angle imaging at 3 Tesla⁷.

Acquisition: Whole-body CMT MRI was implemented on a Philips Achieva 3 Tesla scanner (Philips Healthcare, The Netherlands) with a 2-channel transmit and 2-channel receive body coil and table velocity of 20 mm/s. Data were acquired from an adult volunteer with a dual TR golden angle radial imaging sequence with TR₁ = 3.5 ms and TR₂ = 17.5 ms. The other scan parameters were: zFOV = 1800 mm, in-plane FOV = 400 x 400 mm², in-plane resolution = 1.56 x 1.56 mm², TE = 1.08 ms, excited slice thickness = 12 mm, flip angle of 40^o and a total acquisition time of 90 seconds. The RF power calibration and other preparation scans were performed at the level of a landmarked position at the abdomen.

Reconstruction: Complex image volumes of the two TRs were reconstructed offline in Python (Anaconda version 3.4.2, TX) using 64 profiles per reconstructed axial slice. The reconstructed slice thickness was 25.6 mm (64 * 20 mm/s * 21 ms) with a slice separation of 1.56 mm matching the in-plane resolution. GA image reconstruction involved correction for k-space shifts and sampling density prior to gridding, 2D FFT and roll-off compensation.

Flip angle (α) map estimation: Whole-body α was estimated as described in Ref 7, i.e.,

$$\alpha = \arccos\left[\frac{(S_{2}/S_{1})(TR_{1}/TR_{2})-1}{(TR_{1}/TR_{2})-(S_{2}/S_{1})}\right]$$

where S₁ and S₂ were the magnitude intensity values from the two interleaved scans. The flip angle map was estimated only for a masked region of interest extracted by thresholding the TR₁ image.

Figure 1 shows a coronal slice of the whole-body TR₁ image and estimated flip angle map. Example axial slices are shown in Figure 2 (anatomical images: a-c, flip angle maps d-f), illustrating the variation of the B₁⁺ field along the different sections of the body. RF shading artifacts are apparent in the anatomical slices, which correlate spatially with the flip angle maps. While the prescribed flip angle was 40^o, the estimated α in several locations of the body is seen to deviate significantly from the nominal angle. The golden angle radial acquisition scheme allows flexible retrospective combination of an arbitrary number of profiles to reconstruct slices (and hence flip angle maps) without significant undersampling artifacts. This in turn allows required flexibility in the choice of TR₁ and TR₂ . Our future work will focus on understanding the effects of the moving table on the flip angle measurements, establishing the fidelity of the flip angle maps and pursuing other B₁ mapping methods such as dual angle⁸ and Bloch-Siegert⁹ . Rapid mapping the B₁⁺ field in the whole-body is the first step in mitigating the intensity artifacts originating from transmit RF inhomogeneity and anatomical location dependent miscalibration. With the estimated whole-body flip angle maps, strategies for dynamic, table-position-specific RF shimming for continuously moving table MRI can be devised based on traditional multichannel RF shimming methods^10,11.