Browse Prior Art Database

HYBRID MULTI-RESOLUTION ITERATIVE RECONSTRUCTION

IP.com Disclosure Number: IPCOM000012234D
Publication Date: 2003-Apr-22
Document File: 1 page(s) / 99K

Publishing Venue

The IP.com Prior Art Database

Abstract

ID202142

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 79% of the total text.

HYBRID MULTI-RESOLUTION ITERATIVE RECONSTRUCTION

Two approaches, based on an explicit regridding and a fast convolution respectively, have recently been proposed to rapidly evaluate the matrix-vector multiplication that dominates the computational complexity of iterative recon-struction algorithms for sensitivity-encoded non-Cartesian acquisitions. However, the running times attained with these approaches on commercially available general-purpose hardware are still barely tolerable in practice.

This invention aims at reducing the total operation count and thus the running time of the iterative reconstruc-tion of sensitivity-encoded data with arbitrary distribution in k-space. It proposes to change the resolution in the fre-quency domain during the reconstruction, corresponding to a variation of the field of view in the spatial domain, and suggests to additionally switch between the gridding- and convolution-based approaches.

This invention is based on the observation that a high fidelity of the matrix-vector multiplication is only needed for the last iterations. Beforehand, accuracy may be traded off against complexity, for instance by lowering the resolution or by replacing the regridding with a convolution. Thus, a significant portion of all computations may be saved, particularly because the required number of iterations remains essentially unaffected by these alterations.

Advantages / Improvements:

Considerable acceleration of iterative reconstruction for sensitivity-enc...