Head motion in fMRI and anatomical MRI

The laboratory extensively studies the problem of head motion in functional MRI (fMRI) and anatomical MRI. For example, movement of the head during an fMRI examination introduces "artifact" signals that can potentially degrade data quality and substantially confound the interpretation of brain activity. Motion artifacts are more likely in patient populations, and robust strategies are required that suppress the problem so that medical applications of fMRI can be made practical. Work on this theme has explored several different suppression strategies, including a) careful characterization of head motion during motor tasks in stroke patients, and development of behavioural tasks that produce less head motion during fMRI; b) development of "fMRI simulators", or realistic mock-ups of MRI systems that can be used to optimize experimental design or to train individuals to remain still prior to scanning; and c) use of optical position tracking systems that measure head motion at high temporal and spatial resolution. In particular, we have designed and constructed several such tracking systems for compatibility with the MRI environment, and demonstrated substantial reduction of motion artifact, by adaptively moving the scan plane as determined by the tracking data. This approach provides greatly enhanced image quality and will continue to be a subject of future investigations in the laboratory.

Recent Publications

Maknojia S, Churchill NW, Schweizer TA, Graham SJ. Front Neurosci. 2019 Aug 13;13:825. doi: 10.3389/fnins.2019.00825. eCollection 2019. PMID: 31456656

Golestani AM, Faraji-Dana Z, Kayvanrad M, Setsompop K, Graham SJ, Chen JJ. Brain Connect. 2018 Mar;8(2):82-93. doi: 10.1089/brain.2017.0491. Epub 2018 Jan 22. PMID: 29226689

Faraji-Dana Z, Tam F, Chen JJ, Graham SJ. J Neurosci Methods. 2016 Sep 1;270:46-60. doi: 10.1016/j.jneumeth.2016.06.005. Epub 2016 Jun 8. PMID: 27288867

Screenshot_20221223_091420

Images of the voxel-wise temporal standard deviation of resting-state fMRI time series data for a representative subject. Left: baseline condition where subject attempts to lie as still as possible. Middle: subject nods their head slightly to create motion through the plane of image acquisition. Right: subject undergoes through-plane motion during fMRI where the images are adaptively compensated for motion effects during the data acquisition process. Note that this image has the lowest temporal standard deviation over all voxels (from Faraji-Dana et al., MRI (2016).

Optical Position Tracking Fiducial Marker for High Performance Rigid Body Motion Parameter Estimation

The following information is adapted from Graham Lab Graduate Student, Marina Silic's research on head motion correction in MRI. 

Introduction

 Optical position tracking (OPT) using fiducial markers is an advantageous head motion tracking method because of its high spatiotemporal resolution and precision[1]

• Opportunities remain for improvement via new optical features and multiple camera views[2]  

Methods

1. Fabrication

• Created using two laser printed transparency patterns glued to a clear acrylic block

• Layered patterns necessary for moiré effect

Glued together using liquid optically clear adhesive (LOCA)

Total print area: 4.5 x 4.5 cm

Chessboard square size: 5.5 mm

pic1

2. Marker Feature

pic2

• In-plane DOF (x, y, and roll) measured via perspective n-point problem

• Through-plane translation (z) measured via direct linear transform

• Through-plane rotations (pitch and yaw) measured via moiré pattern phase

3. Benchtop Setup

• Stage information: Newport Corp. 423 & 481-A

   • Translations: 0.1 mm increments across 0-1.5 mm

   • Rotations:

      • 0.5° increments across 0-25°

      • Moiré: 0.1° increments across 0-2°

• Camera information: WAT-204CX Coaxial Transmission camera

   • Focal length: 200 mm; Resolution: 720x480

   • Frame rate: 30fps

   • Measurements averaged over 150 frames

pic3

4. Perspective n-Point Problem (PnP)

pic4

• Standard computer vision 6DOF position measurement

   • Measures pixel change in observed chessboard pattern and known chessboard pattern[3]

5. Direct Linear Transform (DLT)

• Two camera views are used to triangulate points on chessboard to measure depth (z)[4]

pic5

6. Moiré Pattern Operation

pic6

• Moiré pattern is an interference pattern caused by the overlay of two gratings[2],[5]

• Small change in rotation causes drastic change in pattern phase; highly sensitive

• Phase equation: [6]

   : top layer frequency; : layer separation

   : marker rotation; : refraction index

pic7

1. Align marker to a rectilinear grid using the corner symbols

2. Average grayscale values of moiré region are calculated

3. Measured moiré is fit via nonlinear least squares to a sinusoid and compared to a reference sinusoid to obtain phase

4. Marker rotation calculated from phase and phase equation

Results

pic9

Conclusion

• A custom, low-cost (~$5) OPT tool has demonstrated high precision on benchtop tests

• Next steps are to perform MRI validation testing

References

[1] S. Maknojia, N. W. Churchill, T. A. Schweizer, and S. J. Graham, “Resting state fMRI: Going through the motions,” Frontiers in Neuroscience, vol. 13, no. Jul, 2019, doi: 10.3389/fnins.2019.00825.

[2] J. Maclaren et al., “Measurement and Correction of Microscopic Head Motion during Magnetic Resonance Imaging of the Brain,” PLoS ONE, vol. 7, no. 11, Nov. 2012, doi: 10.1371/journal.pone.0048088.

[3] G. Bradski, “The OpenCV library,” Dr. Dobb’s Journal, vol. 25, no. 11, pp. 120–125, Nov. 2000.

[4] R. Hartley and A. Zisserman, Multiple View Geometry in Computer Vision, 2nd ed. Cambridge: Cambridge University Press, 2004. doi: 10.1017/CBO9780511811685.

[5] B. S. R. Armstrong, T. Verron, R. M. Karonde, J. Reynolds, and K. Schmidt, “RGR-6D: low-cost, high-accuracy measurement of 6-DOF pose from a single image,” University of Wisconsin-Milwaukee, 2007.

[6] G. P. Tournier, “Six degrees of freedom estimation using monocular vision and moiré patterns,” Thesis, Massachusetts Institute of Technology, 2006. Accessed: Feb. 22, 2022. [Online]. Available: https://dspace.mit.edu/handle/1721.1/37951 Pitch image credit: D. Tromp, “A guide to quantifying head motion in DTI studies,” The Winnower, May 2016, doi: 10.15200/winn.146228.88496.