Adaptive reconstruction of phased array MR imagery.

An adaptive implementation of the spatial matched filter and its application to the reconstruction of phased array MR imagery is described. Locally relevant array correlation statistics for the NMR signal and noise processes are derived directly from the set of complex individual coil images, in the form of sample correlation matrices. Eigen-analysis yields an optimal filter vector for the estimated signal and noise array correlation statistics. The technique enables near-optimal reconstruction of multicoil MR imagery without a-priori knowledge of the individual coil field maps or noise correlation structure. Experimental results indicate SNR performance approaching that of the optimal matched filter. Compared to the sum-of-squares technique, the RMS noise level in dark image regions is reduced by as much as the square root of N, where N is the number of coils in the array. The technique is also effective in suppressing localized motion and flow artifacts. Copyright 2000 Wiley-Liss, Inc.