Motion during MR image encoding produces inconsistencies in the acquired k-space data, which results in well-known motion artifacts. In clinical settings patient motion during MRI examinations renders a significant fraction of scans non-diagnostic. Recently, due to the increased availability of ultra-high filed imagers of 7T and above capable of sub-millimeter resolution in vivo, it became apparent that even in normal subjects involuntarily motion of about a millimeter may limit significantly the achievable image quality. Motion correction based on additional motion tracking hardware has demonstrated its ability of achieving excellent image quality both in clinical and research settings.
Motion artifacts in MRI can be suppressed effectively if the underlying motion in known with a sufficient precision and accuracy [1-3]. To enable prospective or retrospective correction of motion object position and orientation need to be measured continuously. Oftentimes rigid body model is assumed in such measurements, requiring six degrees-of-freedom (6 DOF) – three translations and three rotations – to be defined.
Optical tracking is an established technology capable of providing motion information in 6DOF with a minimal latency (typically well below 100ms). A range of different motion tracking approaches can in principle be applied in MRI settings to provide position information.Laser tracking is an industry standard for position accuracy [4].
Laser trackers achieve micrometer-scale position accuracy under typical conditions. For rotations, 3 or more landmarks need to be placed on a tracking marker. The laser tracker operates as a single unit, so no inter-unit calibration is required. However, high costs, eye-safety and magnetic field compatibility of the beam deflection mechanics are prevent widespread of this technology in MRI applications.
Stereoscopic motion tracking mimics the visual perception of the 3D world of the living beings. It involves two or more cameras delivering independent views of the scene and is a special case of a more generic technique termed convergent photogrammetry. The use of on-axis lighting and retro-reflecting spheres enables fast automated detection of landmarks and assignment of correspondence between images. 3D positions of every retro-reflective sphere are determined independently by determining the intersection of rays from multiple viewpoints. To track orientation, the positions of three or more landmarks on a rigid body are combined and fitted to a corresponding model.
3D tracking by convergent photogrammetry depends on two calibration sets: relative positions and orientations between the cameras and relative positions between the markers. Instability of these parameters or inadequate separation along either set affects tracking accuracy. 3D tracking with a single camera is also feasible if the marker 3D geometry is precisely known and provides sufficient information based on the detectable landmarks. A landmark is located in a camera image is characterized by its X and Y coordinates. The pose of a rigid marker comprising several landmarks has 6 DOF, therefore the marker needs at least 3 landmarks. The accuracy of the position determination depends on the imaging geometry, with low accuracy arising when large rotations produce only a small apparent shift of the landmarks in the image. The sensitivity of the method is improved when marker is located close to the camera and the camera uses wide angle optics. Planar markers [5] like a checker board [6], or more advanced targets are often used for tracking or camera calibration. Extended 3D-structured markers containing multiple planar checker board segments [7] have been successfully employed in MRI settings to address a sub-optimal marker orientation and limited optical FOV.
MRI environment poses numerous challenges to the optical tracking technology. In addition to the magnetic field safety, eddy currents and vibrations due to the gradient switching and RF interferences, also mechanical and optical constraints of the magnet bore and used coils render most of the off-shelf technologies unusable. For stereoscopic systems it is difficult to achieve a sufficient separation both between the cameras and within the target, furthermore multiple optical paths need to be maintained constantly. An interesting adaptation of a multi-camera technology to the tight constraints of the MR scanner has been presented recently, where three cameras observe three separate retro-reflective markers mounted on a rigid frame, where each camera can only view a single marker [8]. Planar markers like checker boards are known to have sub-optimal orientations where tracking accuracy drops down significantly especially for rotations. It is therefore desirable to reduce sizes of both cameras and markers and reduce the number of optical paths and minimize requirements on the cross-sections of such optical paths in order to fit into the tight spatial constraints within the MR device itself and RF coils in particular.
As the optical markers have to be mounted externally (e.g. on the subject’s face) and therefore remotely to the MRI target organ (e.g. brain) an additional requirement on the tracking technology artises. Due to a substantial distance between the tracking marker and the target organ the errors of rotation determinations manifest themselves as an apparent decrease of the position accuracy. Therefore the rotation accuracy is particularly important for the tracking technology of choice. Actually, similar concerns are applicable to the marker mounting method.
One possible tracking technology able to comply with the above requirements is based on the Moiré Phase Tracking (MPT) principle, which uses a single camera to track a single marker also previously known as a Retro-Grate Reflector (RGR) [9]. The MPT principle consists in an independent measurement of rotations using the 3D structured marker capable of encoding these rotations into the interference patterns. Due to a small marker dimensions and independent measurements of rotations MPT compares very favorably to stereoscopic tracking if considered within the geometric constraints of an MRI scanner [10]. The efficacy of the in-bore MPT system for prospective motion correction in MRI has been demonstrated recently for a range of applications and field strengths [11]. In particular, one of the highest resolution T1-weighted in vivo brain data set acquired with prospective motion correction has been recently made available [12].
Despite of the major advances in the field of real-time motion tracking and correction significant challenges remain. Handling of non-rigid motion and inadequate marker fixation is one of the yet unsolved problems. One possibility to avoid the problems of marker fixation is to eliminate the tracking marker completely. Approaches based on surface tracking with a structured light illumination [13,14] have recently been presented and demonstrated effective for motion correction in positron emission tomography (PET) [15]. It remains to be seen if the accuracy and precision of the motion tracking data originating from surface tracking may become sufficient for motion correction in MRI.
More information about the field of motion correction using external devices can be acquired from recent reviews [16-19].
[1] Nehrke, K. and P. Börnert, Prospective correction of affine motion for arbitrary MR sequences on a clinical scanner. Magn Reson Med, 2005. 54(5): 1130-8.
[2] Zaitsev, M., C. Dold, G. Sakas, J. Hennig, and O. Speck, Magnetic resonance imaging of freely moving objects: prospective real-time motion correction using an external optical motion tracking system. Neuroimage, 2006. 31(3): 1038-50.
[3] Maclaren, J., O. Speck, D. Stucht, P. Schulze, J. Hennig, M. Zaitsev. Navigator accuracy requirements for prospective motion correction. Magn Reson Med. 2010 Jan;63(1):162-70.
[4] Gassner G, Ruland R. Laser Tracking Calibration Testing the Angle Measurement System, SLAC-PUB-13476, December 2008.
[5] Qin, L., P.v. Gelderen, F. Jin, Y. Tao, and J.H. Duyn. Prospective head movement correction for high resolution MRI using an in-bore optical tracking system. ISMRM 2007, Berlin; 1828.
[6] M. Aksoy, M. Straka, S. Skare, R. Newbould, S. Holdsworth, J. Santos, and R. Bammer, In-vivo Applications of Optical Real-time Motion Correction Using a Monovision System, ISMRM 18, p. 4599, 2009.
[7] Forman C, Aksoy M, Hornegger J, Bammer R. Self-encoded marker for optical prospective head motion correction in MRI. Med Image Anal. 2011 Oct;15(5):708-19.
[8] Schulz J, Siegert T, Reimer E, Labadie C, Maclaren J, Herbst M, Zaitsev M, Turner R. An embedded optical tracking system for motion-corrected magnetic resonance imaging at 7T. MAGMA. 2012 Dec;25(6):443-53.
[9] Armstrong B, Verron T, Heppe L, Reynolds J, Schmidt K. RGR-3D: Simple, cheap detection of 6-DOF pose for tele-operation, and robot programming and calibration. In Proc. 2002 Int. Conf. on Robotics and Automation, pp 2938–2943. IEEE: Washington, 2002.
[10] Armstrong B, Andrews-Shigaki B, Barrows RT, Kusik TP, Ernst T, Speck O. Performance of stereo vision and retro-grate reflector motion tracking systems in the space constraints of an MR scanner, Proc. ISMRM 17, p 4641, 2009.
[11] Maclaren J, Armstrong BS, Barrows RT, Danishad KA, Ernst T, Foster CL, Gumus K, Herbst M, Kadashevich IY, Kusik TP, Li Q, Lovell-Smith C, Prieto T, Schulze P, Speck O, Stucht D, Zaitsev M. Measurement and correction of microscopic head motion during magnetic resonance imaging of the brain. PLoS One. 2012;7(11):e48088.
[12] Lüsebrink F, Sciarra A, Mattern H, Yakupov R, Speck O. T1-weighted in vivo human whole brain MRI dataset with an ultrahigh isotropic resolution of 250 μm. Sci Data 2017; 4:170032.
[13] Olesen OV, Paulsen RR, Højgaar L, Roed B, Larsen R. Motion tracking in narrow spaces: a structured light approach. Med Image Comput Comput Assist Interv. 2010;13(Pt 3):253-60.
[14] Olesen OV, Paulsen RR, Højgaard L, Roed B, Larsen R. Motion tracking for medical imaging: a nonvisible structured light tracking approach. IEEE Trans Med Imaging 2012. 31(1):79-87. doi: 10.1109/TMI.2011.2165157.
[15] Olesen OV, Sullivan JM, Mulnix T, Paulsen RR, Højgaard L, Roed B, Carson RE, Morris ED, Larsen R. List-mode PET motion correction using markerless head tracking: proof-of-concept with scans of human subject. IEEE Trans Med Imaging 2013. 32(2):200-9. doi: 10.1109/TMI.2012.2219693.
[16] Maclaren J, Herbst M, Speck O, Zaitsev M. Prospective motion correction in brain imaging: a review. Magn Reson Med. 2013; 69(3):621-36. doi: 10.1002/mrm.24314.
[17] Zaitsev M, Maclaren J, Herbst M. Motion artifacts in MRI: A complex problem with many partial solutions. J Magn Reson Imaging 2015; 42(4):887-901. doi: 10.1002/jmri.24850.
[18] Godenschweger F, Kägebein U, Stucht D, Yarach U, Sciarra A, Yakupov R, Lüsebrink F, Schulze P, Speck O. Motion correction in MRI of the brain. Phys Med Biol. 2016; 61(5):R32-56. doi: 10.1088/0031-9155/61/5/R32.
[19] Zaitsev M, Akin B, LeVan P, Knowles BR. Prospective motion correction in functional MRI. Neuroimage 2017; 154:33-42. doi: 10.1016/j.neuroimage.2016.11.014.