Benefits of Optical Prospective Motion Correction for Single-Shot DTI

INTRODUCTION: Involuntary patient motion causes pixel misregistration and changes the effective diffusion-encoding, which, in turn, results in erroneous estimation of diffusion tensors or higher-order variants (1-3). The most common method of motion correction for DTI is retrospective coregistration of diffusionweighted volumes to a reference volume. However, retrospective correction methods cannot repair incomplete diffusion-weighted volumes, spin history effects, or nonequidistant sampling of the diffusion-encoding direction space. In this study, we propose to use an adaptive optical tracking system (4) to correct for motion artifacts in single-shot EPI DTI and aim to demonstrate the advantages of this prospective approach over retrospective volume-to-volume realignment. THEORY AND METHODS: • Prospective Optical Motion Correction – A small camera was mounted on an 8-channel head coil and acquired images of a selfencoded marker (5) attached to the subject’s forehead (4). The video frames were processed on an external computer. The patient’s pose was determined at a rate of 25Hz and the geometry update was sent back to the sequencer in real-time so that the scan volume followed the subject’s head with minimum delay. • Experiments – A single-shot DTI-EPI sequence was used (TR/TE=10sec/75msec, FOV=24cm, 96x96 matrix, 36 3mm slices @ 1mm gap, b=1000 sec/mm, # diffusion directions = 25 (+3 b=0)). The volunteer was asked to perform mixed (in-plane) shaking and (through-plane) nodding motion throughout the scan, once every ~15 seconds. This scan was repeated with & without prospective motion correction. The data without adaptive motion correction was also reconstructed after retrospective volume-to-volume realignment with SPM5. For reference, two additional datasets were obtained where the subject was asked to stay still and prospective motion correction was turned off and on. Since miniscule subject motion was inevitable even in this case, the dataset with prospective correction was deemed to be the reference. For all datasets, fiber tractography was also performed by planting seed points within the corpus callosum and cortico-spinal tracts. Three quality metrics were used to quantify the quality of reconstructed FA maps and fiber tracts: 1) Correlation coefficient of FA maps between each four datasets and the reference dataset; 2) Normalized High Spatial Frequency Energy (NHSFE) relative to reference dataset to assess the sharpness of the image (6) and 3) the average length of reconstructed cortico-spinal tracts (CST).