Improved Reduction of Motion Artifacts in Diffusion Imaging Using Navigator Echoes and Velocity Compensation Chris A. Clark, Gareth J. Barker, and Paul S. Tofts NMR Research Unit, Institute of Neurology, University College London, Queen Square, London WC1N 3BG, United Kingdom Received July 20, 1999; revised September 24, 1999 Navigator echoes provide a means with which to remove motion artifacts from diffusion-weighted images obtained using any mul- tishot imaging technique. However, residual motion artifact is often present in the corrected images rendering the technique unreliable. It is shown that velocity-compensated diffusion sensi- tization when used in tandem with a navigator echo further re- duces the degree of residual motion artifacts present in the cor- rected images and improves the reliability and clinical utility of the technique. This is demonstrated by applying a method for quantification of motion artifact to brain images of healthy volun- teers scanned using both conventional (Stejskal–Tanner) and ve- locity-compensated gradient sensitization. Other factors affecting the efficacy of the navigator echo technique, such as brain pulsatile motion, gradient b factor, and navigator echo signal-to-noise ratio, are also discussed. © 2000 Academic Press Key Words: diffusion; motion; artifacts; navigator echoes; MRI. INTRODUCTION Diffusion imaging is a magnetic resonance technique that provides tissue contrast that is dependent on the motion of water molecules randomly diffusing in the presence of an applied field gradient (1–3). The interaction between the dif- fusing water molecules and the local cellular structure is widely held to be an important mechanism responsible for the phenomenon of directionally dependent (anisotropic) diffu- sion, observed, for example, in the white matter of the human brain (4, 5). Thus, diffusion imaging may be used to investigate in vivo the structural integrity and orientation of not only healthy tissue but also diseased or injured tissue in which some modification of these water diffusion characteristics may be expected (6 –11). In the early years of development, however, diffusion-weighted images (DWIs) obtained with spin-echo sequences suffered from severe motion artifacts rendering them radiologically redundant. Artifacts arising from bulk sub- ject motion during the application of large diffusion sensitizing gradients induce a phase shift in each of the acquired echoes according to  =   G  r dt , [1] where G is the field gradient vector, r is a displacement vector, and  is the gyromagnetic ratio. The subsequent disruption of the phase information in each echo causes the signal intensity to be distributed along the phase-encoding axis subsequent to Fourier transformation (FT). The magnitude reconstructed im- age often has a ghost-like appearance; an example is shown in Fig. 1. The dot product in Eq. [1] indicates that the phase error arises from components of motion along the direction of the applied field gradient G. The principle of the navigator technique, originally de- scribed by Ehman and Felmlee (12) and first applied to the correction of motion artifact in DWIs by Ordidge et al. (13), is to measure and remove this phase shift from each of the acquired echoes. This can be achieved by acquiring a second (navigator) echo, following the initial imaging echo, in which the phase encoding is rewound so that the echo phase change between successive navigator echoes is dependent only on the phase change due to motion between them according to Eq. [1]. Ordidge et al. demonstrated that correction of the image echo phase by reversal of the motion-induced phase error measured by the navigator echo prior to 2DFT can be used to remove motion artifact due to translational rigid body motion. Anderson and Gore (14) subsequently demonstrated that the correction procedure may be derived from and applied to the navigator and image echo projections, respectively, following FT along the read direction, to additionally correct for artifacts induced by rotational rigid body motion. A similar approach was described by de Crespigny et al. (15). The effect of subject rotation is to cause a shift of the echo in k space or equivalently induce a phase gradient or roll across the projection. In sum- mary, six rigid body phase errors are possible, three transla- tional  x ,  y ,  z and three rotational dk x , dk y and dk z , where x , y and z represent the read, phase, and slice-select directions, respectively. The method of Anderson and Gore can be used to correct  x ,  y ,  z , and dk x while the through-plane rotational phase error dk z may be assumed to be negligible compared to the in-plane rotational phase errors. However, the in-plane rotational phase error dk y is not correctable using a single navigator echo. This is because the phase error in this case is restricted to the phase encode direction, whereas the navigator echo is read out along Journal of Magnetic Resonance 142, 358 –363 (2000) doi:10.1006/jmre.1999.1955, available online at http://www.idealibrary.com on 358 1090-7807/00 $35.00 Copyright © 2000 by Academic Press All rights of reproduction in any form reserved.