Meeting Banner
Abstract #0517

CLEAR: Calibration-Free Parallel Imaging Using Locally Low-Rank Encouraging Reconstruction

SUMMA25Joshua D. Trzasko1, Armando Manduca1

1Mayo Clinic, Rochester, MN, United States

In this work, we present a calibration-free locally low-rank encouraging reconstruction (CLEAR) strategy for accelerated parallel imaging applications. Whereas existing calibrationless parallel MRI methods operate entirely in k-space, using globally-constrained reconstruction models, our proposed strategy imposes constraints locally in the image domain. As we demonstrate, this approach offers substantial computational advantage, is very amenable to parallelized implementation, and naturally incorporates with Compressive Sensing-type sparsity constraints.

Keywords

amplification anal analogously appear approaches bilinear block blocks boundary calibration called challenging channel clear coil coils collision combined comprises computational computationally compute conditions conjunction constraints constructed convex cosine count decomposition define degrade despite discarded disjoint domain efficient efficiently element employ encouraging envelope evaluated evaluation example executed explicitly exploits expressed extracts fashion favorable fine floating footprint free fully generality generalization generalize hardware highlight implementations include inherently innovative integrates investigation iteration joint larger latter like limit locally loss magnitude many matrices matrix memory model models modern noise norm note novel nuclear objective observe onto operations operator optimization orthonormal parallel parallelization penalty performance post problem profiles projected projection promoted properties proposed proximal quality rank recent reconstruction reconstructions regularization retrospectively reviewing root sampling scalability sensitivity settings similarly since singular slice smooth soft space sparsity splitting squares standalone strategies strategy suppose system target tendency terms thresholding transform translation truly typically unavoidable update utility variable whereas wise