Accelerated MR Elastography to Image Brain and Skull Motion

Introduction: : Traumatic brain injury (TBI) is a leading cause of death and disability, affecting an estimated 10 million people each year (Hyder, 2007). Despite its prevalence and severity, the mechanisms involved with injury remain unclear. While most TBI research involves animal or cadaver models for ethical reasons, these are limited in their similarity to the live human brain, emphasizing the need for a technique to quantify force transmission from the skull to the brain in vivo. Magnetic resonance elastography (MRE) is a noninvasive imaging modality that captures small-scale brain tissue displacement from harmonic vibration and can be used to safely observe how the brain deforms in vivo. Previous work introduced a method to image the skull and brain by isolating skull signal from fat within the bone marrow (Yin, 2018). Our group later adapted this technique to acquire high-resolution brain and skull displacement simultaneously (Diano, 2023). However, an accelerated acquisition would allow us to better characterize the brain’s biomechanical response to impact by including multiple frequencies and directions of applied motion (Okamoto, 2023). Here we aim to adapt our fat-water imaging technique by incorporating elastography with distributed generalized encoding to capture high and low sensitivity displacement, eliminate phase wrapping from bulk motion, produce high quality maps of mechanical properties, and accelerate MRE scans to be 3.4x faster than traditional encoding (Kramer, 2024). Data acquired will ultimately provide valuable input parameters for computational models of TBI to define the brain’s response to different injury-inducing forces.

Materials and

Methods: : We incorporated distributed motion encoding into our existing fat-water MRE sequence. Traditional MRE acquires 24 total images which account for motion encoding across the three orthogonal directions, at positive and negative polarities to remove unwanted background phase, and each must be sampled at a minimum of four different phase offsets to capture wave propagation through time. Rather than orthogonal directions, this technique uses encoding vectors distributed across the surface of a sphere and across time (Figure 1), allowing us to sample the theoretical minimum number of images required for MRE – seven (Trzasko, 2015). We tested this sequence in four human subjects (age 23-27, 3M/1F). Protocol parameters included 2 mm isotropic resolution, 40 slices, TR/TE=100/55 ms, 6 spiral shots (R=2), and 2 echo times with a timing gap of 0.5 ms. Experiments were conducted using a Siemens 3T MRI scanner. Vibrations were delivered to the back of the head at 50 Hz using a pneumatic actuator and pillow driver. We then used a joint image reconstruction to resolve fat and water images from the multiple echo times. Reconstructed phase images were used to estimate motion by minimizing the objective function of a general encoding matrix. Resulting brain displacements were put through a nonlinear inversion (McGarry, 2012) to calculate mechanical properties, stiffness and damping ratio, and verify that this technique could capture sufficient harmonic wave propagation. Bulk motion was calculated across the brain and skull to validate that acquired displacements are consistent with those from previous MRE studies.

Results, Conclusions, and Discussions:: The fat-water MRE sequence with distributed generalized encoding was tested and validated in four human subjects, as mentioned previously. Figure 2 shows the resulting fat and water magnitude images from one representative subject, highlighting the defined boundaries of the brain and skull regions of interest. The resulting displacement maps demonstrated that the skull and outer edge of the brain are moving in phase with each other. Wave propagation attenuates as it passes through the skull, demonstrated by higher amplitude of motion observed in the skull compared to the brain. We also saw the highest displacement in the anterior-posterior direction which agrees with the direction of mechanical actuation. This technique was able to capture micron-level harmonic wave propagation throughout the brain tissue, which was sufficient for generating high quality maps of the mechanical properties of the brain. As shown in Figure 3, translational rigid body motion of the skull (10-17 microns) was larger in amplitude than that of the brain (3-9 microns), as expected, exhibiting trends across directions that were consistent with previous studies (Okamoto, 2019).
The proposed multi-echo spiral MRE sequence with distributed encoding was successfully able to capture harmonic, full vector displacement fields in fat and water (skull and brain) simultaneously in approximately 3 minutes, which is a 3.4x reduction compared to traditional MRE. This is a significant improvement in acquisition time compared to the previously described spiral in/out fat-water MRE scan. This technique provides an additional advantage over traditional MRE in that it allows us to capture both high and low sensitivity motion encoding within a single scan to avoid excessive phase wrapping in the skull while still capturing sufficient wave propagation for the nonlinear inversion to reconstruct mechanical property maps. This accelerated acquisition will enable us to expand our imaging session to include a comprehensive analysis of multiple frequencies and directions of actuation to further quantify transmission of force from the skull to the brain. This acquired data will contribute essential parameters to computational models of TBI and ultimately help researchers better understand how different applied impacts can exert injurious strain into the brain tissue.

Acknowledgements and/or References (Optional):: This project was supported by NIH grant U01-NS112120, and Office of Naval Research grant N00014-22-1-2198.