Optimized Coil Design for Parallel Imaging

INTRODUCTION: Parallel MRI utilizes multiple RF receiver coils as encoding engines, whereby the spatial sensitivity profiles of these coils is used in unison with phase encoding to obtain higher accelerations with reduced folding artifacts. The current state-of-the-art in electronics and manufacturing allows construction of coil arrays with a greater and greater number of similarly shaped coil elements, resulting in a considerable increase in data volume., and leading to overload of computational resources and impractical image reconstruction times. This increasing number of coil elements has not been proportional to the increase in the speed of imaging, or to a decisive scalable advantage in image quality. This is due in part to the physics of electromagnetic fields, which limit the RF coil sensitivity profiles to smooth functions, as well as to the coil design techniques that are currently being adopted. Spatial sensitivity profiles and B1 field penetration of individual coil elements are the major determinants of the effectiveness of a coil array, and are a direct function of coil size, shape and location with respect to the desired Region of Interest (ROI);, However, no systematic investigation of the effect of varying individual coil element size and shape has been conducted to date. Massively parallel image reconstructions provide an SNR advantage close to the coil surface, but no substantial advantage deeper within the imaged structure. Hardy et al. showed that smaller elements on the anterior side of the torso, and larger elements on the posterior side produced better g-factor maps at the heart, as compared to similarly sized elements on both sides [1]. These results indicate that, for parallel imaging, optimal coil arrays should have smaller coil elements in the vicinity of the ROI, which deliver high SNR and large intensity variations in the ROI, and also contribute most of the spatial encoding, and larger coils for elements further away, in order to insure that these elements receive a significant signal contribution, as opposed to noise from the ROI, and primarily contribute to SNR enhancement, but less to spatial encoding. In this work, we show that it is possible to follow a simple intuitive approach, to reduce the size of the array without loss of image quality or acceleration speed by linearly combining subsets of small coils into larger coil elements, where these elements have differing sizes.