An investigation of an adaptive modeling algorithm for magnetic resonance image reconstruction

For estimating the AR (autoregressive) parameters from nonstationary MR (magnetic resonance) data, the authors introduce an algorithm with an adaptive mechanism, which attempts to follow the statistical variation of the data. Medical and material MR images reconstructed using the LSL (least-squares lattice) algorithm were more stable than the original nonadaptive TERA (transient error reconstruction algorithm) ARMA (autoregressive moving average) modeling algorithm. Stability with that algorithm was obtained by using several ad-hoc methods which did not make best use of the available resolution achieved by using modeling. Stability with the LSL algorithm was obtained by its adaptive characteristics to allow it to follow the varying properties