Introduction

We propose a protocol and a simple device allowing choreography control by video (ChoCo) for head motion data acquisition in a standardized and reproducible way. In the context of understanding the complex problem of the effects of motion in MRI¹, ChoCo would provide means to collect corrupted labelled MRI datasets. The control of the head choreographies is done by a simple setup: a moving target shown to the subject on a screen combined with customized plastic glasses with a pinhole are constraining the tracking² process by the subject’s head. External synchronization between the stimuli and standard MRI sequences is done by an ad hoc device.
Such a system without known competitors in the literature is compatible with any pulse sequence and clinical MRI scanner, provided that a video stimuli system is available.

Methods

System setup: head choreographies were controlled by a movie displayed on a screen at the end of the MRI bore. A moving dot shows to the subjects the motions to be performed. Each participant was wearing customized plastic glasses with a pinhole to allow dot tracking by the head and avoid eye-tracking motions. The process is analogous to following a moving target with binoculars (Fig. 1). For a proof of concept, two rotational motion types were chosen: abrupt (i.e., short back and forth motion) and ‘’motion and hold’’ (i.e., pause of a few seconds before going back to initial position). Both types of motion were performed in two directions: namely right/left (R/L) and up/down (U/D), which correspond to yaw and pitch rotation directions, respectively (considering the convention illustrated in Fig. 1). Precise timing and synchronization between the pulse sequence and the choreography stimuli was obtained either with the TR output signal of the EPI pulse sequences (fMRI type) or with an external device (Arduino, with optical output) using the external trigger option available on the MRI scanner (see Fig. 2 for setup overview). Six volunteers performed the head choreographies on a 3T MRI scanner (Prisma fit, Siemens).

Technical parameters: multi-volume (>600) multislice (36 slices/volume) EPI SMS pulse sequences (TE = 30 ms, TR = 600 ms, slice acceleration factor = 4) with 3x3x3.3 mm voxels were used for motion estimation. Classical 2D Spin-Echo (TE = 12 ms and TR = 750 ms) and 3D T1 MP-RAGE (TE = 2.36 ms and TR = 1930 ms) were performed to propose samples of motion artifacts for commonly used clinical pulse sequences. Spin-Echo and MP-RAGE sequences were externally triggered with an acquisition window of 800 and 1950 ms, respectively. Voxel dimensions of 0.9x0.9x4.0 mm and 0.9 mm isotropic were used for SE and MP-RAGE, respectively.

Data postprocessing: EPI volumes were realigned using the SPM software³ to estimate rigid-body motion parameters. Average subject angle error with respect to target was measured as well as subject stability (baseline standard deviation).

Results

Subjects are able to reproduce head motion choreographies accurately (Fig. 3) even if subjects may deviate from the baseline (0° angle). The effective angles were computed with respect to the local baseline. The training sessions (5 repetitions of two different motion types) also show that a learning process takes place, usually after about 2 repetitions. Indeed, the angle amplitudes get closer to targets even before the end of the sessions. Moreover, a sliding window standard deviation computed over each subject’s baseline confirms that subjects are able to maintain the latter reasonably stable (with again a slight decrease of the standard deviation near the end of the training session).
Spin-Echo and MP-RAGE acquisitions show that the motion artifacts reproduced across subjects are qualitatively similar, which argues strongly in favour of accurate motion reproduction among all subjects after the training phase. In Fig. 5, a Spin-Echo k-space magnitude (averaged over 27 slices, for the same subject), reconstructed with the ESPIRiT algorithm⁴, exhibits dark bands⁵ at phase columns corresponding to motion onset.

Discussion

Accurate motion reproduction demonstrates the usefulness of our approach, which is simple, universal and can be modified to include other types of motion. Results pertaining to training data encourages the use of this preliminary session before performing the actual representative scan. Given the results we obtained, we are confident that with a reasonable training session (less than 6 minutes), following the ChoCo protocol, subjects will be able to reproduce the head choreographies accurately. The qualitative similarity of motion artifacts in Spin-Echo and T1 3D MP-RAGE, for all subjects, was expected given correct head choreography reproduction and correct timing was further confirmed by visual inspection of classical Spin-Echo k-space magnitude (Fig. 5) with respect to projected movie.

Conclusion

Our work proposes a simple setup and protocol to perform standardized motion-controlled experiments allowing researchers in the MRI community to study the effects of well controlled motion on any MRI pulse sequence without modification or complex external setup.

Acknowledgements

No acknowledgement found.