|
|
|
“PISCO Software Version 1.0”
by Rodrigo A. Lobos, Chin-Cheng Chan, Justin P. Haldar
March 2023
The estimation of sensitivity maps from k-space calibration data is a common task in many multichannel MRI applications1. In the last decade, subspace-based estimation methods have gained popularity within the MRI community, where ESPIRiT [3] has emerged as the method of choice by many researchers. Even though these methods possess great estimation accuracy and robustness, they can be computationally demanding, and their underlying theoretical principles can be nontrivial to understand. In view of these limitations, we have proposed in [1,2] a novel theoretical framework for subspace-based sensitivity map estimation. This new framework relies on theoretical concepts from the literature on linear predictability and structured low-rank modeling, and we expect it might be more intuitive an easier to understand for some readers. Based on these novel theoretical concepts, we have proposed a nullspace-based algorithm for sensitivity map estimation which is theoretically equivalent to ESPIRiT. In addition, we have also introduced in [1, 2] a set of computational techniques | which we collectively call PISCO (Power iteration over simultaneous patches, Interpolation, ellipSoidal kernels, and FFT-based COnvolution) | that, remarkably, can enable substantial improvements in computation time (up to a 100x-fold improvement in the cases we have tried) when integrated to subspace-based methods as shown in [1, 2].