Introduction

When presenting a new motion correction (MOCO) method results are regularly validated through volunteer experiments in which a predetermined set of movements are repeated. A major drawback of this is that the prevalence of artefacts is strongly dependent on the type of pulse sequence, in what order k-space is filled, and how this is correlated to when and how the volunteer moved. Reproducing results between different subjects can therefore be challenging. This makes comparisons for different subjects, different setups or even different implementations of the same setup difficult.

Furthermore, when translating a method to real patients it may be impractical to make these image comparisons. We therefore identified a need for accurately repeating artificial or even real patient motion patterns in volunteer experiments to gain better insight into the efficacy of MOCO methods.

To this end, we developed a motion visualisation tool that can be used to playback a motion parameter log file to a volunteer. The playback is synchronised with the pulse sequence execution by storing a frame number that links each update to an encoding step of the pulse sequence. The volunteer's head pose is constantly recorded using a tracking device. The “playback” pose from the desired motion pattern can then be displayed alongside the volunteer’s current pose, allowing them to adjust their head so that they reproduce the motion stored in the log file.

Methods

The modifications to the scanner setup are shown in (green) Figure 1. A laptop, running the visualisation tool, was connected to:

• the realtime computer of the scanner (to synchronise the intended motion pattern with the sequence playout),

• a motion tracker (to obtain realtime feedback of the volunteers head position) and

• an MRI compatible monitor (for displaying the information to the volunteer).

The 6 degrees of freedom in motion needed to be visualised on a 2D screen in a way that is intuitive to the volunteer. To achieve this we took advantage of the strong correlation between orientation and position as subjects typically rotate their heads about the base of their skull. We therefore use only the three degrees of freedom in orientation to represent the subject’s head pose. Changes in orientation are shown to the subject using a symbol on the display: Rotating in a nodding “yes” motion moves the symbol up and down. A “no” movement moves the symbol left and right, and a tilting movement rotates the symbol about an axis normal to the plane of the screen. The display then moves a “target” symbol according to the log and the volunteer is tasked with keeping their symbol aligned with the log symbol.

The visualisation tool was written in Julia programming language¹ and made use of the Makie² package for real time data visualisation. The display included scatter plots of the current motion on the left of the screen. The centre showed the “target” and “current” markers and the right section displayed a 3D model of a head transformed by the current pose. An example data snippet is animated in Figure 2.

A Wireless Radiofrequency-triggered Acquisition Device (WRAD)³ was used to capture the volunteer motion for the visualisation feedback on the display and if required for prospective motion correction (PMC). The motion packets from the WRAD include an RF pulse counter. This counter was used to synchronise the logfile “target” screen position to the “current” position. For asynchronous PMC methods a synchronisation signal from the scanner will be required for tagging the data. The volunteers were shown a “target” that is approximately 250 ms in the future, to account for their reaction time.

Experiment 1) A log file was recorded during an acquisition in which the volunteer performed strong circular head motion. Three different volunteers were then tasked with tracing this log file.

Experiment 2) The motion recorded during a coronal T2FLAIR PROPELLER acquisition of a paediatric patient was played back to a volunteer. The volunteer then repeated the motion twice, once with and once without PMC.

Results

Experiment 1) The motion plots displayed in Figure 3 show a strong correlation between different volunteers.

Experiment 2) The volunteer was able to reproduce the motion profile and gain insight into how the patient peered out of the scanner in order to see their guardian (Figure 4).

Discussion and Conclusion

Realtime visual feedback enables reproducible volunteer motion experiments. This effectively allows a researcher to be “put in the shoes” of a real patient to gain a better understanding of how they move. Retrospectively acquiring “motion corrupted” images with patient motion logs can also be helpful in justifying the use of MOCO in a clinical setting, where comparisons can be difficult to obtain.
Other applications could include:

• Applying the same motion pattern to different contrasts.
• Improving retrospective correction techniques.

Acknowledgements

This research was supported by Barncancerfonden under grant MT2020-0006