Head motion tracking for functional MRI using an air ultrasound array

Patient motion is known to be a major cause of image artifact during magnetic resonance imaging (MRI). Especially in functional MRI (fMRI), head motion can corrupt signals from brain activity, and deviations of under a millimeter can accumulate nonlinearly, rendering the fMRI technique altogether incapable of brain imaging. It is apparent that the addition of a real time MR-compatible motion tracking system could provide critical information that, if combined with MR-motion correction techniques, could result in more reliable imaging. Here an MR-compatible air ultrasound transducer array was developed, with the goal of quantifying fMRI-relevant motions of under 0.5 mm. The array was designed to track nose position so that, by approximating head movement as a rigid body rotation about the second cervical vertebra (C2), the position of the entire head could be estimated. The array consisted of 5 planar air transducers (400 kHz, 1-cm-diameter) whose axes were oriented to coincide at the tip of the nose, approximately 6 cm in front of the array. One transducer, working as a transmitter, was situated in the array center and aligned approximately parallel with the tip of the nose. The remaining 4 elements were situated left, right, above and below the transmitter, all relative to the head position. After verifying MR-compatibility and optimizing element spacing, preliminary bench top measurements were made using a plastic head phantom affixed to a rotational motor. By analyzing the phase shift in back-scattered acoustic signals, the angle of head shaking or nodding, were estimated; motions that correspond to the most commonly observed movements during fMRI. In five measurements over a range of one degree rotation, predicted angles were found to agree within 85% of the actual head position. Finally, we verified the device on a human volunteer. Overall results demonstrated the ability of the array to identify sub-millimeter (0.2 mm) head motion, thereby allowing estimation- of head rotation angles as small as 0.10. This method, applied as feedback into existing MR-motion correction algorithms, may lead to a practical approach to more robust fMRI imaging.