Magnetic resonance elastography (MRE) is a phase-contrast based MRI technique that measures displacement due to propagating mechanical waves, from which mechanical material properties such as shear modulus can be recovered. MRE has high potential as a quantitative, physics-based imaging modality for soft-tissue characterization in many clinical applications. This talk gives an overview on the technological aspects of MRE including driver technology, imaging sequences and reconstruction methods.
Overview
MRE basically combines three technical components [1]: 1) actuators for excitation of mechanical waves in the body, 2) MRI hardware and motion sensitive imaging sequences, and 3) postprocessing methods for reconstruction of mechanical parameter maps by wave inversion algorithms. The talk will discus the concepts, challenges, and solutions of these three technological pillars of MRE. In the following a brief overview of concepts in MRE hardware, imaging sequences and reconstruction is given.MRE hardware
A review of the literature demonstrates that MRE can be implemented on any type of MRI scanners at almost all field strengths across a wide range of vibration frequencies, as long as the imaging sequence is synchronized to the tissue vibration [2]. In fact, the timing of an MRE sequence with respect to the induced mechanical motion is a strong requirement for capturing waves at different phases of their propagation through the tissue. In the simplest case, TTL (transistor–transistor logic) or optical trigger pulses are sent from the scanner to the vibration generator to initiate vibration. In more complex implementations, the duration of the trigger pulses can be modulated to convey further information to the vibration generator, for example, to switch the vibration frequency during a multifrequency measurement [3]. Finally, possible interactions between actuators and MRI, leading to imaging artifacts, have to be excluded. For this reason, nonmetallic drivers are mostly used for clinical examinations. The power supply of those actuators is based on either pneumatic vibrations requiring waveguide ducts or electrical signals requiring filters to suppress electromagnetic interference.MRE sequences
MRE sequences are usually equipped with a motion-encoding gradient (MEG) which induces a phase shift to the complex MRI signal proportional to the deflection amplitude of the harmonic motion field. Magnetic field inhomogeneities and susceptibility effects are usually static while the wave propagation is an intrinsically dynamic process with well-defined behaviour in the time domain. The key to isolating wave information is to acquire several images at different phases of the wave oscillation cycle by increasing the delay between the trigger pulse and the start of the MEG. The acquired wave images are Fourier transformed along time in order to extract the complex-valued wave image at driving frequency. Most MRE exams nowadays use a variant of FLASH or echo-planar imaging (EPI). The wave acquisition within a specific slice is repeated several times in order to account for different field components (different MEG directions), time steps (trigger shift) and frequencies.Viscoelastic Parameter Reconstruction Methods
Data processing and parameter recovery in elastography always face an inverse problem. Specifically, the aim is to derive information about the spatial distribution of elastic parameters (shear modulus, shear wave speed, etc.) from measured dynamic quantities, typically the complex-valued time-harmonic displacement field. The inversion methods that have been established in the context of MRE can be broadly subdivided into the categories 1) direct inversions, 2) phase gradient methods (including Local Frequency Estimation - LFE), and 3) variational methods. Most clinical MRE examinations are processed today by methods which fall under 1) or 2). The choice of reconstruction methods in MRE depends on a number of inevitable tradeoffs between the finite support of wave images, noise, suppression of compression waves or stochastic motion artefacts as well as reconstruction time and unknown boundary conditions. A basic assumption that is usually made is local homogeneity – i.e., that the material is homogeneous within a small local processing window, so that the inversion equations greatly simplify but also lead to artefacts especially near boundaries. Combining multiple experimental information such as multifrequency wave data with multiple components stabilizes the inversion and reduces artefacts since the systems equations are over determined [2]. Phase gradient methods are more stable against noise than direct inversion methods but require directional filtering in order to decompose the wave field into planar (unidirectional) waves. The combination of multi-inversion with the phase gradient method has been demonstrated in abdominal applications to provide noise-robust maps of shear wave speed with high detail resolution [3].