Comparison of Whole Body Transmit Coil Configurations for RF Shimming at 3T

Introduction Whole body MRI at 3T may be impeded by B1 transmit field inhomogeneities caused by the dielectric shortening of the RF wavelength [1]. RF shimming techniques based on parallel transmission [2,3] can strongly improve the image quality in clinical whole body applications [4]. In this context, it is an important question, how the RF shimming performance depends on the chosen coil topology and, in particular, on the number of transmit channels. In the present work, an 8-channel transmit system is used for B1 mapping and shimming, providing the flexibility of emulating different coil configurations and comparing their RF shimming performance. Methods In vivo experiments (eleven healthy volunteers, all male, age: 27-49, BMI: 18-28) were performed on a 3T MRI system (Philips Healthcare, Best, The Netherlands) equipped with eight transmit channels [5] and an 8-element TX/RX TEM body coil [6]. The AFI (Actual Flip Angle Imaging) technique [7] was used for abdominal B1 mapping (450×270×75 mm3 FOV, 64×38×5 matrix, angle = 60°, TR1 = 20 ms, TR2 = 100 ms, TE = 2.3 ms, transverse orientation, improved spoiling scheme [8], all-but-one coil encoding [9], resulting in 18 seconds scan time per 3D B1 map, cf. Fig.1). For RF shimming, magnitude least square fitting (Fig.2) of the multi-coil maps was performed using a dedicated Java implementation of the local variable exchange method [10]. Tikhonov regularization was used to restrict the optimization to a selected set of coil eigenmodes. To enable or disable a certain coil eigenmode i, the corresponding regularization parameter λi was set to 0 or ∞, respectively. Four different eigenmode sets were compared: a) quadrature coil (λ1=0), b) 2-port birdcage (quad and anti-quad, λ1=λ7=0) [11]), c) 4-channel coil setup proposed in ref. [12] (λ1..4=0) and d) full 8 channels (λ1..8=0). The coefficient of variation cv= RMSD(B1)/ RMS(B1) of the synthesized shimmed maps was taken as homogeneity measure and compared to the quadrature case, which was chosen as a reference. In addition, the required total power (||b||) and the maximum power per coil element (max{bi}) were determined from the fitted drive scales and normalized to the RMS(B1) for the considered coil configurations. Results