Mario Zeller1,
Alexander Mller1, Marcel Gutberlet2, Andreas J.
Bartsch3, 4, Daniel Stb1, Dietbert Hahn1,
Herbert Kstler1
1Institute
of Radiology, University of Wrzburg, Wrzburg, Germany; 2Institute
for Interventional and Diagnostic Radiology, Hannover Medical School,
Hannover, Germany; 3Department of Neuroradiology, University of
Heidelberg, Heidelberg, Germany; 4FMRIB Centre, Oxford University,
Oxford, United Kingdom
In Cartesian imaging, optimal SNR can be achieved by filtering the k-space proportional to the signal (SNR matched filter). This however leads to Gibbs artifact amplification. In contrast, Gibbs artifacts are reduced by filters that apodize the k-space periphery, leading to non-optimal SNR. K-space density weighting allows combining both approaches. The application of an SNR matched filter ensures optimal SNR, while a non-Cartesian k-space sampling allows achieving a prospectively defined point spread function. In this work, k-space density weighting was applied to echo planar imaging. The results indicate significant SNR advantages of density weighting over Cartesian imaging with retrospective filtering.