Preventing Signal Dropouts in DWI Using Continous Prospective Motion Correction

Introduction: Diffusion weighted imaging (DWI) has become indispensable in clinical routine, especially due to its sensitivity to early stages of brain ischemia. Patient motion during measurements is a major source of artifacts. A distinction must be made between motions during the acquisition of a single diffusion encoded image (intrascan) and movements in between these images (interscan) as well as between in-plane and through-plane motion [1]. In-plane motion in between scans can be corrected retrospectively by realigning the images, correcting the phase shift and applying a b-matrix rotation [2, 3]. For through-plane motion these methods fail since the measured object is inconsistent. It has been shown that prospective motion correction can correct for both types of interscan motion using navigators [4] or an external tracking device [5]. However, intrascan motion remains problematic. Even for single-shot EPI intrascan motion can lead to severe signal dropouts. Movements during diffusion encoding and between refocusing pulses can result in severe artifacts making it impossible to apply retrospective correction methods. This work presents a new method where prospective motion correction is continuously applied during the diffusion gradients to correct for six degrees of freedom intraand interscan motion. Methods: All experiments were performed on a 3 T Magnetom Trio (Siemens Healthcare, Germany) using an optical tracking system (ARTtrack3, Advanced Real-Time Tracking GmbH, Germany) for prospective motion correction. The twice-refocused SE-EPI sequence with diffusion weighting (Fig.1) was modified in two steps. First, prospective motion correction was enabled once per slice excitation. In a second step, the diffusion gradients were split into small 2 ms segments. During the execution of each segment, the subsequent one is prepared, including new position information if available. This allows for the continuous correction of intrascan motion. In the following human experiments the head motion of a healthy volunteer was tracked using a mouthpiece with reflective markers. In the measurement presented 11 slices (matrix 110×110, voxel size 2×2×2mm, TR = 6000ms, TE =94ms) were acquired with different b-values (b0, b1 =500, b2 =1000s/mm2) and three weighting directions each (trace weighting). For illustrative purposes, a simulation of the corrected gradients during head rotation was performed and is shown in Fig.1. Results: Fig. 2 shows 10 slices with one diffusion weighting (b1 =500s/mm2). In the first line an experiment with no deliberate motion was performed for