Amplified Magnetic Resonance Imaging (aMRI)
Samantha J Holdsworth1, Wendy W Ni1, Greg Zaharchuk1, Michael E Moseley1, and Mahdi S Rahimi1

1Lucas Center for Imaging, Department of Radiology, Stanford University, Palo Alto, CA, United States

Synopsis

This work introduces a new visualization method called amplified Magnetic Resonance Imaging (aMRI), which uses Eulerian Video Magnification to amplify subtle spatial variations in cardiac-gated brain MRI scans and magnify brain motion. This approach reveals deformations of brain structures and displacements of arteries due to cardiac pulsatility, especially in the brainstem, cerebellum, and spinal cord. aMRI has the potential for widespread neuro- and non-neuro clinical application, because it can amplify and characterize barely perceptible motion, and allows visualization of biomechanical responses of tissues using the heartbeat as an endogenous mechanical driver.

Introduction

Intrinsic motion of brain parenchyma, spinal cord and arteries can be altered by pathology or natural factors as a result of changing tissue properties, vascular compliance, and pressure and flow dynamics. Conventional MRI has limited ability to image these changes, partly because of its lack of sensitivity to small but macroscopic motion. Here, we introduce a new motion detection and visualization method called amplified Magnetic Resonance Imaging (aMRI). aMRI uses the Eulerian Video Magnification technique (EVM) method [1], originally developed for use in conventional videography, to magnify cardiac-induced brain motion recorded by retrospectively cardiac-gated MRI. The heart is assumed to act as an endogenous mechanical driver, producing subtle changes that can be amplified to reveal useful biomechanical information.

Methods

Image acquisition: With IRB approval, we scanned 4 healthy volunteers (2M/2F, 29-64 years old) using 3T GE MR750 Discovery MRI system (GE Healthcare, Milwaukee, WI) and an 8-channel head coil. The aMRI method is implemented with the cardiac-gated balanced steady-state free precession (bSSFP) sequence [2] (Fig. 1a) to produce cine MRI images. For all scans, a 2D mid-sagittal slice of the brain was obtained (matrix size = 2242, flip angle = 45°, TR/TE = 3.6/1.3 msec; FOV = 24 mm2; slice-thickness = 4 mm, acceleration factor [R]= 2, peripheral cardiac gating, views/segment [Nseg] = 1, retrospective re-binning of data to 150 cardiac phases, scan time = 3:30 min). Phase-contrast cine MR images were acquired as a gold standard [3] (matrix size = 2242, 42 slices, flip angle = 15°, TR/TE = 5.5/12 ms; FOV = 24 cm2; slthck = 4 mm, Venc = 5 cm/sec in A/P and S/I directions, R = 2, Nseg = 10, scan time = 5 min). To test the hypothesis that aMRI reflects tissue compliance, we manipulated intracranial pressure in three volunteers by performing a 20sec physiologic challenge called the Valsalva maneuver [4]. Motion amplification: Fig. 1b describe the EVM method: it takes cine MRI data as ‘video input’, performs spatial decomposition, then temporal filtering and frequency-selective amplification of the spatial components to synthesize a motion-amplified cine data set. For this study, we used: an ideal narrowband filter with a pass band of 0.5-1.5 Hz, corresponding to a cardiac cycle of 30-90 beats per minute; a user-defined amplification parameter, α = 200; and spatial frequency cutoff, λc = 200. Normalized variance maps (calculated over the cardiac cycle) and difference maps (amplified frames minus corresponding unamplified/reference frames) were produced.

Results

Figure 2 and its associated movie Video 1 illustrate a dataset where output images displayed amplified subtle motion of brain structures arising from cardiac pulsatility. While movement of the brain was barely perceptible in the input data, it was significantly amplified in the output images, especially near the brainstem, cerebellum and spinal cord. This motion is highlighted in a spatiotemporal x-t and y-t slice-time profile through the brain (Fig. 2a-c). Difference maps and variance maps (Fig. 2d-e) also show amplified motion. Figure 3 clearly demonstrates that amplified displacement of aMRI produces better visualization of brain motion than the subtle, unamplified displacement revealed by phase-contrast MRI. For the Valsalva experiment, participants showed considerably reduced brain tissue displacement during the Valsalva maneuver compared with that during free breathing (Figure 4). These findings support the expectation that increased intracranial pressure induced by the Valsalva maneuver reduces brain tissue motion [5]. It is important to note that there is no simple relationship converting values of α and λc to a physical amplification factor, as noted in the original EVM paper [1]. In our implementation, our heuristic focused on achieving adequate motion amplification without the introduction of confounding image artifacts and noise, by independently adjusting α and λc.

Conclusion

Here we introduce a new MR-based method called aMRI, in which near-imperceptible brain motion can be enhanced through spatio-temporal processing of dynamic cardiac-gated MRI acquisitions. Using aMRI, one may observe important biomechanical aspects of the brain and potentially other organs in a non-invasive manner that may shed light on a large number of clinical applications in medical imaging. This approach may enable the characterization of brain tissue integrity in a wide range of neurological diseases in which changes in the biomechanical properties of the brain can lead to dramatic changes in pressure and flow dynamics, and hence tissue motion.

Acknowledgements

Helpful discussions – Sam Cheshier, Thomas Christen, David Feinberg, Gary Glover, Gerald Grant, Jürgen Hennig, Bronwen Holdsworth, Michael Iv, Rob Lober, Julian Maclaren, Manoj Saranathan, Stefan Skare, Allan White, Kristen Yeom. Grant support and other assistance: NIH 2RO1 NS047607-06, Stanford Interdisciplinary Graduate Fellowship, Center of Advanced MR Technology at Stanford (P41 EB015891), and the Lucas Foundation.

References

[1] Wu H, Rubinstein M, Shih E, Guttag J, Durand F, Freeman WT. Eulerian video magnification for revealing subtle changes in the world. ACM Transactions on Graphics (Proc. SIGGRAPH) 2012; 31:4.

[2] Oppelt A, Graumann R, Barfuss H, Fischer H, Hartl W, Schajor W. FISP: a new fast MRI sequence. Electromedica (Engl Ed) 1986; 54,15–18.

[3] Nayler GL, Firmin DN, Longmore DB. Blood flow imaging by cine magnetic resonance. J. Comput. Assist. Tomog. 1986; 10, 715–722.

[4] Prabhakar H, Bithal PK, Suri A, Rath GP, Dash HH. Intracranial Pressure Changes During Valsalva Manoeuvre in Patients Undergoing a Neuroendoscopic Procedure. Minim. Invasive Neurosurg. 2007; 50(2), 98-101.

[5] Bhadelia RA, Madan N, Zhao Y, Wagshul ME, Heilman C, Butler JP, Patz S. Physiology-based MR imaging assessment of CSF flow at the foramen magnum with a Valsalva maneuver. Am J Neuroradiol. 2013; 34(9),1857-1862.

Figures

Figure 1: aMRI acquisition and reconstruction. a) A bSSFP acquisition is used to generate 2D cine images. b) The EVM algorithm [1] is applied: each cardiac phase is spatially decomposed (Laplacian pyramid approach), band-pass filtered, amplified by a factor related to α, added to the original image components, and then summed.

Figure 2: Brain motion induced by cardiac pulsatility is too subtle for direct visualization, but is clearly seen using aMRI. a,b,c) x-t and y-t profiles from reference (unamplified input) and aMRI-processed images. d) aMRI difference images around brainstem. e) variance maps.

Video 1: Animated movie of subject (64y, M) in Fig. 2 shows aMRI-amplified motion around brainstem, which was barely perceptible on the input bSSFP cine MRI images. Low-frequency shading across the brain may be considered as an artifact, possibly arising from CSF variation over the cardiac cycle.

Figure 3: Brain pulsatile motion and ‘nodding’ motion (arrow) is subtle on phase-contrast MRI, but can be clearly seen using the aMRI technique. a) 10 frames from a cardiac-gated phase-contrast sequence (A/P, Venc=5 cm/s). b) 10 frames from aMRI. Both sets of difference images use frame 1 as reference.

Figure 4: Reduced motion with the Valsalva maneuver is evident in the midbrain and spinal cord after motion amplification, revealed by (a) variance maps overlaid on a grayscale cine MRI frame, and (b) difference maps of regions of interest outlined in (a).



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0346