Reconstruction as a source of artifact in nongated single-shot diffusion-weighted EPI

A controversy has existed over the requirement to cardiac gate diffusion-weighted MRI acquisitions of the brain. Conventional wisdom suggests gating to be a necessary requirement to allow acquisition of accurate data, but recent applications find gating not necessary. The signal-to-noise and acquisition duration of these two approaches can be quite different; thus, this difference in methodology is important. This is particularly relevant when performing quantitative work such as diffusion tensor imaging. Here, the convention to gate is explained as being due to the historical use of low spatial resolution and more recently to the use of different reconstruction approaches. It is demonstrated that the Margosian reconstruction approach only yields high quality results when used in a gated fashion. Zero padding of the acquisition matrix provides an alternative reconstruction method that is not found to accentuate the artifacts that are due to pulsatile motion in the diffusion-weighted acquisition and thus do not require a gated acquisition. The relative merits of each reconstruction approach are discussed, including estimates of the relative signal-to-noise ratio and resolution benefits. It is concluded that both gated methods and nongated methods can each provide high quality results with appropriate reconstruction methods. (c) 2005 Elsevier Inc. All rights reserved.