Correction for Involuntary Motion Related Artifacts in Multi-Shot EPI using k-space Data Swapping

Phase corrections based on readout-asymmetry and navigatorechoes can effectively reduce the motion-related phase errors in multi-shot EPI for diffusion weighted imaging studies. However, occasional involuntary motions might be too severe to be corrected by the navigator echoes. To minimize this effect, kspace data swapping technique is proposed in this study. By replacing the corrupted k-space data segments with non-corrupted data sets from the redundant data sources, this method can effectively suppress occasional motion related image artifacts and dramatically improve image quality. Introduction A very effective approach to reduce motion-related artifacts in multi-shot EPI for diffusion-weighted imaging is based on navigator phase correction [ I ] . In practice, however, occasional involuntary motions might be too severe to be corrected by the navigator echoes. For this type of motion effects, some of the segmented scans might remain corrupted. The aim of this study is to develop a k-space data swapping method to remedy this problem. One or more redundant data sets were then acquired and the corrupted k-space data were substituted with non-corrupted ones from redundant data sources. Methods All scans were acquired on a 1.5 T ACS-NT system (Philips Medical Systems, Best, The Netherlands) equipped with high performance gradients (Powertrak 6000). Diffusion-weighted imaging was accomplished using multi-slice multi-shot echo planar imaging sequences, with cardiac triggering. Phase Correction of directional readout asymmetry was performed using a pair of zero phase-encoded reference scans, corresponding to different scanning directions, acquired before image data collection. The sequence had two 180" RF pulses, with strong diffusion gradients surrounding the 1'' one, and the segmented EPI readout played out after the 2"d one. A navigator echo was acquired without phase encoding after the first 180' RF pulse, in order to correct the motion induced phase errors in the following image echoes [2,3]. Data acquisition was repeated 2-3 times to increase signal to noise ratio in the further diffusion tensor computation, as well as to produce redundant data sets for potential k-space data swapping in post processing. The area of navigator echo profile was taken as the measurement of imaging echo quality according to the central slice theorem of 2D Fourier transform. If a corrupted segmented-scan is identified from the navigator profile, the segmented scan in the k-space is replaced by corresponding segmented scan from another redundant data set. The appropriate k-space data segments can be automatically combined together based on the assumption that all navigator echoes (of different shots) should have similar profiles. This is achieved by maximizing the navigator selection criterion defined as: