Scientist Center for Neuroscience and Regenerative Medicine, HJF
Abstract Text: INTRODUCTION Studying the frequency dispersion of mechanical properties provides a means to probe material microstructure at various length scales. This has the potential to diagnose traumatic brain injury (TBI) which has so far remained invisible in conventional MRI techniques given the large dynamic range of shear modulus compared to MRI relaxation times, diffusivity, etc. Time harmonic MR elastography (MRE) is a method of choice for “palpating” tissue but requires an actuator that can operate over a wide range of frequencies. However, at low frequencies, MRE require large displacements owing to reduced MR sensitivity while at high frequencies operation is often limited by vibrations in the actuator system. In this study, we report the development of a new broadband (0 – several kHz) piezoelectric actuator to perform MRE within a micro-imaging MRI scanner and a Finite Element Modeling (FEM) framework to describe system behavior.
METHODS Our material specimen consists of two layers of agarose gel of 0.1% and 0.12% concentration stacked on top of each other in a glass tube. The actuator is constructed from a piezoelectric stack operable from 0 to 5 kHz (Thorlabs, Sterling, VA) with a stroke length of 100 micron, which is mated to a 3D printed plunger that rests on top of the gel. The piezo transducer is synchronized with the NMR system and is operated at 10 Hz in this study.
The longitudinal motion of the plunger introduces shear waves in the gel which were captured using phase contrast MRI triggered at different phases of the actuation cycle spaced 10 ms apart for one full cycle. The cylindrical geometry of the sample container introduces complex shear waves which can be used to study its mechanical properties. 3D MRE data was acquired on a 7T scanner (Bruker Biospin) using a 25 mm quadrature RF probe with the following parameters: δ/Δ = 1.5/10 ms (Hadamard scheme with b = 0, 250 s/mm2), FOV = 30 x 25 x 25 mm, TR/TE = 200/16 ms, and a 0.5 mm isotropic voxel resolution. The shear modulus of the gel was estimated using algebraic Helmholtz inversion. Diffusion tensor imaging (DTI) data was also acquired with b-values = 0, 1000 s/mm2 and 20 diffusion weighting directions with the actuator turned OFF to compare the sensitivity of DTI and MRE in imaging the layered medium.
Finite element method (FEM) simulations were performed and compared to the observed shear waves and the direct inversion results. Simulated displacement fields were produced in a half-cylinder geometry containing two adjacent linear elastic layers with shear moduli set to 6 Pa (close to the actuator) and 10 Pa (far from the actuator) as estimated from the experimental displacement data. The normal and tangential gel displacements at the walls of the tube were set to zero to model gel adhesion to the rigid boundary. A symmetry boundary condition was prescribed on the remaining flat surface of the half-cylinder. The mesh consisted of 9,216 trilinear hexahedral elements with a maximum side length of 1.87 mm. The model was solved at maximum time steps of 6.25 ms with a dynamic quasi-Newton solver (FEBio).
RESULTS AND DISCUSSION The measured displacement field shows uniform mechanical excitation of the FOV studied. The simulated displacement fields share many characteristics with the measured displacement fields, such as the radial motion of the gel as indicated by horizontal and out-of-plane displacements, and uniaxial shear wave propagation. However, the measured decrease in wave amplitude in the stiffer gel is not as pronounced in the simulation suggesting that this attenuation may be due to presence of an interfacial layer between the two gel layers with a different stiffness or due to differences in viscous dissipation in both the gel layers, which were not modeled. The estimated shear modulus map clearly distinguishes the layers having two different gel concentrations based on their differences in stiffness (6 vs 10 Pa on average). The DTI maps, however, fails to capture this subtle difference—the MD and FA were uniform throughout the material except at the edges where the signal was low as shown in the S0 map.
