Direct reconstruction of B1 maps from undersampled acquisitions

Introduction: Parallel Transmission (PTx) offers more control over B1+ than single channel systems by independently controlling elements of transmit arrays, each with spatially distinct sensitivity patterns. The majority of PTx methods require accurate knowledge of these sensitivities, the measurement of which is often time consuming for the whole array. In this work, we exploit the observation that B1+ maps are smoothly varying in space and are therefore likely to be sparse in several transform domains, a property previously used for receive fields [1]. The B1+ mapping source images contain anatomy and so are less sparse than the B1+ maps themselves. This suggests that the latter might be effectively obtained directly by a Compressed Sensing (CS) type reconstruction [2]. We have tested this idea for the AFI mapping sequence [3]. It is shown that accurate reconstructions of simulated and measured in-vivo B1+ fields are possible using only 40% of the full dataset.