In this talk, motion correction methods for head MRI are presented that rely on signals from external sensors as opposed to ones from tissue magnetization. Head motion is primarily rigid, that is, the head's pose can be captured by determining three rotational and an additional three translational degrees of freedom, which simplifies the motion correction task to a degree that enables the use of markers as a proxy for the head pose. This shifts the head pose encoding from pulse sequence design to marker tracking. This shift enables independent MR image encoding and pose tracking, whose consequences are manifold and the content of this talk. Finally, once the head pose has been faithfully estimated, the sequence geometry must be adapted using the updated head pose information, which requires the transmission of the geometry information to the real-time sequence controller.We will cover the elements of a prospective motion correction system and present current solutions and remaining challenges:

1. Marker and Receiver Subsystem

2. Signal Encoding Principles

3. Signal Processing Chain

4. Transmission of the Geometry Updates to the Sequence Controller

Optical markers and cameras are the most commonly employed motion correction systems in head MRI, and several different implementations thereof have been devised. This technology has matured since the early publications An early system published in [1] is based on a set of four reflective markers whose individual positions were recorded with a camera standing outside of the scanner bore. The markers are attached to a bite bar in order to ensure a rigid configuration among themselves as well as the skull of the subject.A second approach [2] relies on a single marker whose relative position to an initial reference is encoded in moiré patterns, i.e. the motion-dependent interference pattern between two gratings. The camera system employed in [2] was now placed inside the scanner bore, which proved to be more robust and practical compared to the camera systems outside the bore.Another approach that was presented in [3,4] employs a checkerboard marker that is recorded with an in-bore camera placed directly on the head coil. An important feature of that type of marker is a digital code printed on checkerboard pattern, which provides tracking stability in cases when the line-of-sight from the camera to the marker is partially interrupted.This talk will cover the individual strengths and weaknesses of the different optical marker designs and their technology.NMR-based markers will be discussed as an important alternative to optical systems. They consist of a set of small NMR active droplets (diameter ~ 0.5 - 2 mm) whose position is encoded with dedicated gradient waveforms, which modulate the phase of the droplet's signal as a function of position. The marker's signal is received by a small RF coil that is tightly wrapped around the droplet and connected to tuned RF receive hardware.NMR Markers for motion correction were developed as early as 1998 by Derbyshire et al. [5], but gained traction only with the first in-vivo data presented by Ooi and Krueger [6,7]. A step forward in NMR marker design was done by Barmet et al. [8] whose markers, or "field probes", were applied to motion correction by Haeberlin et al. in 2013. The sensitivity of their field probes was high enough to enable two things that will be covered in this talk: First, a novel position encoding method that enabled concurrent field probe localisation and image encoding ("gradient tones"), a feature that had not been possible with the previous approaches. Second, they allow to additionally correct for imperfections in the image encoding field by field monitoring that is performed concurrently to the image acquisition.We will give an overview on how NMR marker systems are designed and applied in practice.

In this part of the talk, we will discuss how the different markers actually encode the head pose. For the optical methods, we will include the functionalities of the moiré-patterned marker as well as the checkerboard marker. For the NMR-markers, we will discuss the sequence module using static gradients as well as methods employing dynamic gradient waveforms such as gradient tones.

An important aspect of all motion correction system is the speed at which the position updates can be computed. We will discuss signal processing aspects for both the optical methods, which contain 2D pattern recognition algorithms, as well as NMR-based methods, which process 1D NMR signals.

An overview on how the geometry updates are fed back to the MR system will be presented. The discussing includes the communication protocol between the motion correction system and the MR scanner, and the incorporation of the updated geometry information into the running MR sequence.