Target Audience

Engineers, physicists and anyone else who likes to tinker with hardware.

Purpose

To give a quick background of what has already been done in the hardware motion tracking field, discuss some exciting new directions to follow and show that augmenting the scanner hardware with new sensors is not necessarily as scary as it looks.

Summary

Augmenting the MR signal acquisition with sensors opens up opportunities for advanced image reconstruction. Measuring motion is of particular importance as it has a significant impact on the interpretation of the NMR spectra, and can be crucial for obtaining high quality images. During this short course we will work through some of the questions you should ask yourself when developing your own motion tracking hardware (Figure 1) and describe the rationale for some existing methods that will hopefully give you some inspiration.

The first question, which you’ve probably already answered by showing interest in this course, is why you can’t solve your motion tracking problem using navigators [1,2] or motion robust pulse sequence designs [3,4]. This can be because these methods typically require a lot of scan time and can be limited in their temporal resolution. As this has nothing to do with hardware development we will skip over these methods, however, it is important to keep them in mind, because a new hardware device should outperform non-hardware methods to have a significant impact.

You don’t always have to create or construct a device, you can repurpose some of the existing scanner infrastructure. Free induction decay (FID) navigators are a fantastic example of hardware repurposing [5], where the multi-channel receive coils that enable parallel imaging are repurposed as motion detectors. The hardware used for SAR monitoring in parallel transmit capable scanners can also be repurposed to measure changes in reflected energy (scattering matrix) associated with cardiac and respiratory motion [6].

If you don’t want to rely on advanced infrastructure or it just isn’t sensitive enough for your problem you can try and augment the hardware with something new. An elegant example of this is transmitting a pilot tone, and detecting how respiration and cardiac motion affect its reception [7] during the conventional MR signal reception. The motion data is perfectly “timestamped” and synchronised with the data acquisition.

If markers are an option, you can construct small passive resonant circuits that couple to the MR scanner’s receive coil [8] and create spikes in the NMR spectra. This can be helpful for highly accelerated navigators. For rigid body motion detection, more than three markers are required. A correspondence problem then needs to be solved in order to uniquely identify each marker in each of a series of spectra. Small capsules filled with acetic acid (off-resonance) can also be used [9], in this case the correspondence problem is solved by taking advantage of the “spatial encoding” of a multi channel receive coil (“repurposing”). You may be noticing a trend, things are generally easier when you digitise signals using the MR scanner.

This approach can be taken one step further with wired markers that can be individually excited and sampled using either the MR hardware or your very own custom built receive/transmit chain that runs in parallel to the imaging experiment. This increases hardware complexity, however, it enables much more freedom in navigator design and very high SNR. In some cases, regular sequences can be made self-navigating [10]. These sensors can also reveal field dynamics at high temporal resolutions (field cameras).
Another advantage of wired probes is the ability to move the signal processing electronics far away from the imaging field of view. Most interventional procedures (iMRI) are compatible with wired probes - although special care should be taken when running longer cables within the scanner. One proposed solution to this is the use of photovoltaics to power an integrated circuit through an -as long as you like- optical cable. The chip is then capable of transmitting information (local NMR frequency) back by modulating the current in a small microelectromechanical (MEMs) mirror that deforms under the forces caused by interactions with the static magnetic field [11].

All the solutions discussed thus far are based on NMR, which is sensitive to both motion and field dynamics. An alternative approach of gaining spatial information from the scanners gradient fields is by detecting their rate of change with pickup coils [12]. In this case the signals are of a much lower frequency and straightforward to process with embedded electronics. This opens the opportunity to build small battery-powered digitisers that can sample sensors near the FOV and transmit the signals using low power 2.4 GHz radios [13]. It does not take a huge leap to include a wide range of commercially available sensors (accelerometers and angular rate sensors), however, the need for a battery and wireless communication limit the minimum size the device/marker can have.

One can envisage smaller passive markers that can be detected by some other means, independent of the scanner’s operation. An interesting new approach is to use a short-wave radio “reflector” that interacts with locally transmitted (~1 MHz) signals [14]. The way the “reflected” signal interacts with a set of gradiometer receive coils is unique for a specific pose of the miniaturised reflector. If you can see the object you're interested in tracking, optical markers make a lot of sense and offer some of the best tracking performance using technology from an already well established field [15]. More recently structured light has been employed to track subject motion without the use of markers [16].For organ tracking multimodal imaging strategies can be employed where a faster imaging modality can be paired with MR to perform motion tracking tasks, such as ultrasound [17].

Acknowledgements

The Karolinska Sjukhuset MR physics group.