Presentation: Thursday @ 2pm

1. # 5036 Presentation: Thursday @ 2pm Real-Time Motion Correction for High-Resolution Imaging of the Larynx: Implementation and Initial Results Juan M. Santos Dwight G. Nishimura Joëlle K. Barral Electrical Engineering Stanford University

2. In a Nutshell We propose a real-time algorithm to combat the main types of motion that corrupt high-resolution larynx imaging. Our algorithm combines navigator-based motion correction with a reacquisition strategy. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

3. MOTIVATION

4. The Larynx Anterior commissure Vocal cords Thyroid cartilage Thyroid cartilage Cricoid cartilage Axial Sagittal http://www.antiquescientifica.com -- Drawing courtesy of Julie C. DiCarlo Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

5. Laryngeal Motion Real-time acquisition: 13 frames per second Notice swallowing at time t = 18 s! Healthy volunteer Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

6. Laryngeal Motion : Outliers (Sporadic motion) : Bulk motion (Drift) High-frequencies: Respiration, 14 cycles per min Motion detected by Cartesian navigators Cancer patient Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

7. Laryngeal Motion Types How to mitigate their effects • Intermittent, sporadic motion: • Swallowing, coughing, jolting  Alternative ordering schemes • Continuous motion: • Flow (carotid arteries)  Phase encodes L/R • Bulk motion (drift) •  Physical restraints; Coaching; Navigators • Respiration  Diminishing Variance Algorithm (DVA) If a continuous drift happens, DVA never converges. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

8. Diminishing Variance Algorithm (DVA) Sachs, MRM 34: 412-422, 1995 -- Sachs, IEEE-TMI19: 73-79, 2000 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

9. METHODS

10. Proposed Approach We propose to first correct the data based on the shift information. We then reacquire encodes whose projections could not be properly corrected. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

11. Implementation RTHawk 1.5 T Santos, IEEE-EMBS 2: 1048-1051, 2004 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

12. Pulse Sequence Fast Large Angle Spin Echo = FLASE • Spin echo: immune against flow & off-resonances • 3D: high-resolution • T1-weighted contrast Ma, MRM 35:903-910, 1996 -- Song, MRM 41:947-953, 1999 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

13. Encodes Ordering Examples with 32 phase encodes and 16 slice encodes kz ky Elliptical (concentric) Sequential Square spiral Pseudo-random Wilman, MRM 38: 793-802, 1997 -- Bernstein, MRM 50: 802-812, 2003 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

14. Reconstruction Pipeline The user stops the scan when satisfactory image quality is obtained. Barral, ISMRM Motion Workshop 2010, p. 18 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

15. GUI X Y Z S S Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

16. Experimental Parameters FOV 12 cm - Matrix size 256x128x32 - TR/TE = 80/10 ms Sequentialencodes order Three-coil larynx dedicated array First pass (full acquisition: 4096 encodes): 5 min 28 s Each additional pass (64 encodes reacquired): 5 s Phantom (orange) scans: coronal acquisitions In vivo (larynx) scans: axial acquisitions Barral, ISMRM 2009, p. 1318 -- Coil picture courtesy of Marta G. Zanchi Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

17. PHANTOM EXPERIMENTS

18. Phantom Experiment 1: No Motion • An orange was scanned. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

19. Phantom Experiment 1: No Motion • As expected, image and corrected image are identical One pass = Full acquisition Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

20. Phantom Experiment 2: DVA • Non-rigid motion was simulated by switching from the coronal acquisition to an axial acquisition towards the middle of the scan, for several seconds. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

21. Phantom Experiment 2: DVA • As expected, motion correction fails Pass # 1 = Full acquisition: 4096 encodes acquired • Motion detection successful • Shift information meaningless Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

22. Phantom Experiment 2: DVA • When corrupted encodes are reacquired, a motion-free image is obtained. Pass # 1 Pass # 6 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

23. Phantom Experiment 3: Motion Correction • Towards the middle of the scan, the table was manually translated. It was brought back to its original position several seconds later. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

24. Phantom Experiment 3: Motion Correction Pass # 1 = Full acquisition: 4096 encodes acquired • As expected, motion correction works Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

25. Phantom Experiment 3: Motion Correction • Blurry: the final position of the table did not perfectly match the original position. Pass # 1 Pass # 4 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

26. Phantom Experiment 4: Combined Algorithm • Non-rigid motion was simulated by switching to an axial acquisition towards the middle of the scan, for several seconds. The table was then manually translated. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

27. Phantom Experiment 4: Combined Algorithm • Motion correction successfully accounts for the translation Pass # 1 = Full acquisition: 4096 encodes acquired Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

28. Phantom Experiment 4: Combined Algorithm • Reacquisition needed to correct for non-rigid motion Pass # 1 Pass # 6 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

29. IN VIVO EXPERIMENTS

30. In Vivo Experiment 1: Without Instructions • A healthy volunteer was scanned. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

31. In Vivo Experiment 1: Without Instructions One pass = Full acquisition Slice 20/32 X Y Slice 26/32 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

32. In Vivo Experiment 1: Without Instructions Sagittal reformat Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

33. In Vivo Experiment 2: With Instructions • A healthy volunteer was scanned. He was asked to swallow at will and to accentuate motion when the center of k-space was being acquired. For this experiment, 192 encodes were reacquired each additional pass. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

34. In Vivo Experiment 2: With Instructions Pass # 1 = Full acquisition: 4096 encodes acquired • Swallowing properly detected • Only bulk motion corrected by motion-correction X Y Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

35. In Vivo Experiment 2: With Instructions • When corrupted encodes are reacquired, motion correction is needed to account for bulk shift (drift) that happened between passes. Pass # 1 Pass # 3 Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

36. WRAP-UP

37. Conclusion & Future Work • Our real-time algorithm corrects for rigid-body motion and reacquires encodes that could not be corrected. • Additional scans are needed to validate the robustness of the method in vivo. • Future work will improve the flexibility of the algorithm and improve the user interface. Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.

38. Thank you! Contact: jbarral@stanford.edu On larynx imaging, see also posters # 2410 and 2416! Real-Time Motion Correction for Larynx Imaging -- J.K. Barral et al.