MRI with TRELLIS: a novel approach to motion correction Julian R. Maclaren, Philip J. Bones ⁎ , R.P. Millane, Richard Watts Computational Imaging Group, Department of Electrical and Computer Engineering, University of Canterbury, Christchurch, New Zealand Received 3 May 2007; revised 18 July 2007; accepted 8 August 2007 Abstract A motion-correcting pulse sequence and reconstruction algorithm, termed TRELLIS, is presented. k-Space is filled using orthogonal overlapping strips and the directions for phase- and frequency-encoding are alternated such that the frequency-encode direction always runs lengthwise along each strip. The overlap between strips is used both for signal averaging and to produce a system of equations that, when solved, quantifies the rotational and translational motion of the object. Results obtained from simulations with computer-generated phantoms, a purpose-built moving phantom, and in human subjects show the method is effective. TRELLIS offers some advantages over existing techniques in that k-space is sampled uniformly and all acquired data are used for both motion detection and image reconstruction. © 2008 Elsevier Inc. All rights reserved. Keywords: MRI; FSE; Motion correction; Artifacts 1. Introduction One disadvantage of MRI in comparison with other scanning modalities is the relatively long data acquisition time required. Thus, image quality is often degraded by motion artifacts, including image blurring and ghosting [1]. A number of techniques are employed to help ameliorate these problems: one is to prevent the motion occurring using sedation or physical restraints. Sedation involves risk [2] and also adds complication to the scan. Physically restraining patients is only partially effective. Navigator echo techniques have been used extensively and are becoming increasingly sophisticated [3–5] but still require the acquisition of extra data to determine motion. It is also possible to track subject motion using ultrasound [6], digital imaging [7], head- mounted tracking devices [8], or techniques such as using spectroscopy to identify the angle of rotation of a crystal containing deuterium [9]. An advantage of these techniques is that motion can be compensated for in real time by adjusting scanner gradients. However, the added complexity of these approaches is a significant disadvantage and none has yet achieved widespread clinical acceptance. An alternative approach is to collect the data in such a way that the motion can be detected and corrected in post- processing. The advantage of this method is that no additional equipment is needed. One such technique that has achieved clinical success is PROPELLER MRI [10]. While PROPELLER has proved effective in reducing motion artifacts in both simulations [11] and clinical trials [12], it is slower than standard imaging sequences since the center of k-space is sampled multiple times [10]. Recent work combining EPI [13] and k-space under-sampling [14] with PROPELLER has produced promising results and may lead to faster implementations of PROPELLER in the future. We have developed an alternative post-processing techni- que to correct for patient motion which we call ‘TRELLIS’— an acronym for Translation and Rotation Estimation using Linear Least-squares and Interleaved Strips and so named because its sampling pattern looks like a trellis. This technique has some similarities to PROPELLER: it does not require extra hardware; it samples k-space more than once in order to obtain information about patient motion; and the final reconstruction requires some form of gridding. However, the data acquisition and image correction methods differ substantially and TRELLIS may offer significant advantages because the whole of k-space is uniformly sampled instead of concentrating sampling in the center of k-space as in PROPELLER. Available online at www.sciencedirect.com Magnetic Resonance Imaging 26 (2008) 474 – 483 ⁎ Corresponding author. Tel.: +64 3 364 2987x7275. E-mail address: phil.bones@canterbury.ac.nz (P.J. Bones). 0730-725X/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.mri.2007.08.013