[Brainmap]: Stephen Robert Frost PhD-Markerless optical head tracking for real-time correction of MRI motion artifacts

Wednesday, December 13, 2017 - 12:00 to 13:00
Building 149 Rm 2204

 

Abstract:

Motion is widely recognised as a major problem in MR neuroimaging, with consequences including impaired clinical diagnosis, wasted time and money for repeat scanning, and biased research results. Motion during 3D-encoded scans, such as T1-weighted MPRAGE and T2-weighted SPACE, creates complicated artifacts due inconsistencies in the acquired k-space data. These artifacts are difficult to correct post-hoc, and some estimate of the head motion is usually necessary. One way to estimate motion is with fast volume navigator (vNav) images inserted every ~3 s, and using the tracking information to adapt the sequence in real-time to maintain a consistent view of the head has been shown to substantially reduce artifacts. Motion can also be estimated with camera and marker systems, and this allows sequence-independent, high-frequency real-time correction, however, the requirement to attach a marker to patients or subjects (e.g. on the forehead or with a bite-bar) can be prohibitive, and can also result in less accurate estimates of head motion.

In this talk I will present our latest work on markerless real-time motion correction for MPRAGE and T2-SPACE. We used the “Tracoline” optical camera system (TracInnovations, Copenhagen, Denmark), which reconstructs 3D “point clouds” of the subject’s face and registers them to a reference point cloud to estimate head movement. This is appealing because: i) it can provide high-frequency motion estimates (every ~30 ms), ii) it is expected to have low impact on the MRI workflow. Our results suggest that the ability to update the FOV rapidly can have benefits over slower updates (e.g. vNav correction every ~3 s). We also show that the morphometry measures from the corrected images acquired during substantial motion are consistent with those derived from artifact-free images acquired without intentional motion.

 

About the Speaker:

My background is in MRI physics and pulse sequence development for application to diffusion imaging, and more recently, real-time motion correction. During my PhD research with Peter Jezzard and Karla Miller at the FMRIB Centre, University of Oxford, we worked on the readout-segmented EPI (or RESOLVE) sequence for high-resolution diffusion imaging with reduced EPI distortion and blurring artifacts. In collaboration with David Porter (Siemens), we demonstrated partial Fourier and simultaneous multi-slice (SMS) accelerations, which reduce the otherwise prohibitive scan times to clinically feasible durations and enable multi-direction diffusion studies. We also worked on a cardiac synchronised multi-slab implementation, which like SMS, improves the SNR efficiency by reducing the repetition time (TR).

In 2013 I started post-doctoral research on real-time motion correction with Peter Jezzard at the FMRIB Centre. In collaboration with Andre van der Kouwe and Dylan Tisdall we deployed the volume navigator (vNav) technique in some specialised research sequences, including vessel-encoded dynamic ASL angiography. In work on vessel wall imaging in the neck, we developed a navigator approach to identify motion-corrupted data, which were then estimated with parallel imaging. Since 2016 I have been working at the MGH Martinos Center with Andre van der Kouwe and Bruce Fischl, primarily on motion correction with markerless tracking and vNavs.