pTX Spoke RF Pulses for Cardiac MRI at 7T: a New Design Robust against Respiration Induced Errors, based on a Virtual Simultaneous Exhale-and-Inhale Calilbration Scan

INTRODUCTION. Cardiac MRI may strongly benefit from high SNR at ultra-high fields (UHF) but is challenged by the short RF wavelength that induces large flip angle variations over the heart. This problem has been successfully addressed in 7T cardiac CINE imaging using parallel transmission (pTX) with 2-spoke pTX RF pulses [1]. Although 2 spokes can provide high contrast uniformity, we sometimes observed significant deviations between measurement and prediction by Bloch simulation as discussed in [1]. Also, at times, the performance changed when using two different 2-spoke trajectories in k-space with similar prediction (Fig1). We suspected motion to be involved (some trajectories may be more prone than others) but subject’s bulk motion in the scanner was excluded. All calibrations and acquisitions are performed under breath-hold, assuming that subjects always return to the same exhale position. If, however, the subject’s respiratory state (ideally exhaled) during cardiac CINE scans is not the same as during transmit B1 (B1+) and ∆B0 calibration scans, degradation in the performance of the 2-spoke RF pulse uniformity can be expected. Here we investigate the impact of breath-hold position on cardiac CINE imaging using 2-spoke pTX pulses and we demonstrate an RF pulse design strategy to increase the robustness of 2-spoke pTX RF pulse to variations in breath-hold position. Our findings may also impact free-breathing pTX acquisitions, including at 3T. METHODS. After obtaining consent, volunteers were imaged at 7T with a 16 channel pTX prototype console (Siemens, Erlangen) using a 16 channel transceiver body coil [2]. The 16 B1+ sensitivity maps were obtained in an axial plane from ECG triggered GRE scans (1 B1+ map per heartbeat) acquired during a single breath-hold using a fast, small flip angle (FA) calibration [1,3]. For each TX map we also acquired within the same heartbeat a sagittal map (16 in total) crossing the diaphragm to monitor breath-hold position during B1+ calibration. A ∆B0 map, required for 2-spoke RF pulse design, was derived from a dual TE GRE acquisition that also included a diaphragm scan. The B1 + and ∆B0 mapping procedure was repeated 3 times while instructing the subject to hold their breath on 3 different breathhold positions (POS): I) full exhale, II) half inhale, III) full inhale (POS verified with diaphragm data). With slice selection along the z-axis, gradient blips applied along the x and y axes were used to symmetrically position the 2 spokes about kx-ky=0 (see Fig. 2a). In polar coordinates, the radius kr varied from 0 to 10m -1 in 1m steps while the angle φ of the ‘spokes axis’ (dashed line in Fig2a) to the kx axis was varied from 0° to 360° (note: not 180°) in 10° steps. RF pulses were designed as described in [1] based on B1 + and ∆B0 maps acquired in POS I, using a magnitude least squares solution [4] for each of the 361 different trajectories. We performed Bloch simulations for each solution, first with the original B1 + and ∆B0 maps (obtained at POS I), then after substituting the latter with those obtained at POS II and then at POS III. For each spoke trajectory characterized by (kr,φ) and each breath-hold position, we calculated the coefficient of variation (CV= std(FA)/mean(FA)) of the simulated FA profile, as well as the total RF energy (En) (sum of energy through all channels and spokes) and the maximum energy per channel (Emax). This procedure was repeated, but with a modified pulse design based on a virtual 2-slice calibration, including in one slice B1+ and B0 maps from POS-I, and in a second slice B1 +