Magnetic resonance elastography: from invention to standard of care

Abstract

In 1995, a vivid image of diffracting waves in red and blue was published on the cover of the journal SCIENCE. An article in that issue described a new imaging technology called magnetic resonance elastography (MRE) (Muthupillai in Science 269:1854–1857, 1995). In 2004, quantitative images of liver stiffness in vivo, obtained with MRE, were demonstrated for the first time at the annual meeting of the International Society for Magnetic Resonance in Medicine. Only five years later, the technology had become widely available as an FDA-cleared diagnostic tool for patient care. MRE has emerged as a reliable non-invasive diagnostic method for detecting and staging liver fibrosis. Deployed on more than 2000 MRI systems worldwide, MRE has received a Category I CPT code from the American Medical Association, based on clinical availability and efficacy. For many patients, MRE now provides a safe, more comfortable, and much less expensive alternative to liver biopsy for diagnosing liver fibrosis. Although progress in radiology is notable for a history of very rapid translation of technology innovations to patient care, the path is rarely linear. This article reflects on the story of MRE, the advances and the setbacks, and the lessons that were learned in the process.

Graphical Abstract

Introduction

The advent of first-generation MRI systems in the early 1980’s launched a remarkable period of advancement. The field was wide open and drew brilliant and curious minds from many disciplines. Within a decade, MRI methods had been successfully developed to reveal the physical states of water in tissue, to measure and image blood flow, perfusion, water diffusion, tissue microstructure, cardiac motion, tissue biochemistry, tissue composition, and dozens of other properties.

Yet even after a decade of rapid progress, there were still many unexplored opportunities. For instance, physicians had known for centuries that many disease processes such as cancer, inflammation, and fibrosis can cause marked changes in the hardness of tissue, as sensed by touch. However, standard imaging technologies such as CT, MRI, and ultrasonography were not capable of quantitatively assessing tissue stiffness.

In informal terms, the property called “stiffness” relates the amount of deformation (strain) that results when a measured force (stress) is applied to a material. In engineering terms, “stiffness” is called elastic modulus and is the ratio of stress divided by strain. Early attempts to measure this property in living tissue used radiography, CT, MRI, or ultrasonography (US) to measure tissue displacements resulting from an applied force at the skin surface. However, this approach can only be used to evaluate tissues that are within a few cm of the location of the applied force. Another fundamental limitation is that while the distribution of strain values (displacement) can be measured with imaging, the distribution of applied stresses within the tissue is unknown, so it is generally not possible to create images that quantitatively depict tissue stiffness (elastic modulus) with this approach [1].

In the early 1990’s, our research team at Mayo Clinic focused on a different approach. It was known that when mechanical shear waves at a given frequency propagate through a material, then the wavelength at each location in the object will be determined by the local stiffness of the material. We reasoned that if shear waves could be generated in the body and somehow imaged, then the observed wavelength at each location in a cross-sectional image could be converted to a quantitative measurement of tissue stiffness.

The key challenge was to find a way to visualize propagating mechanical shear waves in vivo. Even at low frequencies, propagating shear waves would only cause a few microns of cyclic motion in tissue. This is far below the spatial resolution of whole-body MRI systems. Our team devised a solution based on phase contrast MRI with novel motion-sensitizing gradients that are modulated at the same frequency as the applied vibration. It took several years of work to implement the novel pulse sequences and to develop devices suitable for generating shear waves. The device had to be safe for use in the MRI magnet, while not distorting the magnetic field or generating radiofrequency noise. A way had to be found to precisely synchronize complex motion-encoding gradient waveforms with vibrations applied by the external device. MRI-compatible phantoms simulating the mechanical properties of tissue had to be developed.

The complexity of the experiment created many points of failure that had to be systematically identified and addressed. Persistence was necessary, but after unsuccessful attempts spanning many months, the feasibility of the technique was finally confirmed by imaging propagating shear waves in a gel phantom. The technique proved to be capable of resolving shear waves with amplitudes as small a few hundred nanometers (comparable to the wavelength of light) in tissue-simulating phantoms. In parallel, our team developed novel mathematical algorithms for processing wave images to create quantitative maps of tissue stiffness. The first publications on this new technology, in SCIENCE and NATURE Medicine, described the theoretical basis and results in phantoms and organ specimens, along with validation by mechanical testing (Fig. 1) [2, 3]. The first in vivo application of MR elastography was reported in 1996 [4].

Fig. 1.

Figures from the first publication in 1995 describing dynamic MR elastography. A MR image mapping the propagation pattern of shear waves at 250 Hz in a tissue-simulating gel phantom with stiffer and softer inclusions on the left and right respectively. The wavelength is longer in the stiff inclusion and shorter in the soft inclusion. B Quantitative stiffness map (elastogram) computed from the wave images. [From ref 2, with permission]

Diagnosing liver fibrosis

We speculated at an early stage that MRE might be a useful technique for detecting liver fibrosis. It was long recognized by surgeons that while healthy liver tissue is very soft, advanced fibrosis makes the liver very hard to the touch. Intrigued by the potential, our team initially attempted to image shear waves in animal liver tissue specimens. This proved very difficult because the shear waves in the specimens were heavily attenuated. Our early attempts to generate and image shear waves in the liver in vivo also failed. In these tests, shear waves generated at the body surface (by vibratory motion parallel to the skin) proved to be heavily absorbed by subcutaneous tissues, with very little mechanical energy reaching deeper structures. Due to these difficulties, our team set aside the goal of assessing liver fibrosis and refocused on exploring other applications such as evaluating skeletal muscle with MRE. Generating observable shear waves in muscle proved to be very easy. An unexpected result of that work was that we found that a better way to generate shear waves in deeper tissues was to apply vibrational motion perpendicular to the skin surface. This generates longitudinal waves that are far less attenuated than shear waves. The longitudinal waves propagate to deeper structures and generate shear waves at tissue interfaces due to a physical mechanism known as mode conversion.

With that important lesson, our team refocused on the liver. Experiments showed that longitudinal waves applied to the body wall created shear waves in the liver that could be readily imaged with MRE. Furthermore, the waves were much less attenuated in the living liver than in the early experiments with ex vivo liver specimens. This was a second important lesson. We have learned that the mechanical properties of tissue in vivo are often much different than the properties of the same tissue, ex vivo.

Our team worked to develop an efficient solution for applying longitudinal vibrations at the body surface. An “active driver” device, located outside the MRI scanner room, generates an air pressure waveform at the desired frequency. The pressure waveform is transmitted through tubing to a simple non-metallic drum-like “passive driver” that transmits the mechanical energy into the body within the scanner [5]. The first successful application of MRE for assessing liver stiffness in vivo was presented by our team at the 2004 annual meeting of the International Society for Magnetic Resonance in Medicine (ISMRM) [6]. Preliminary results showing the ability to discriminate between healthy and fibrotic liver in vivo were presented by our team in 2005 at the annual meetings of the ISMRM and the Radiological Society of North America [7]. Published articles in 2006 provided further evidence that liver stiffness as measured with MRE was a highly promising biomarker for detecting and staging liver fibrosis [8–10].

Recognizing the potential of this application, our team worked intensively in 2005 and 2006 to advance the implementation. We developed a new version of the active driver device that integrated all components into one unit. This made it much easier to install the technology on MRI systems for clinical evaluation. Through extensive testing, our team found that the ideal location for the passive driver is on the anterior right chest wall, directly over the liver. The passive driver causes the diaphragm to vibrate, generating shear waves which propagate largely transversely from the capsular margins the liver. Therefore, it was possible to visualize and measure the wavelength adequately in individual 2D axial images. This eliminated the need to image the wavefield in three dimensions, which would complicate and lengthen the acquisition. A more advanced MRE processing algorithm based on direct inversion of the wave equation was developed and integrated into the MRI system. By 2006, the technology was installed on multiple MRI scanners on the Mayo Rochester campus for full scale clinical testing. Systems were also installed on clinical scanners at Mayo facilities in Arizona and Florida. An additional six MRE systems were constructed and shared at no cost with external collaborators in the US and Europe. These collaborations rapidly accelerated clinical testing of the technology.

Dissemination of MRE technology

By 2007, hepatologists at Mayo Clinic had concluded that MRE was a reliable alternative to liver biopsy for diagnosing liver fibrosis and were routinely using it in their clinical practice (Fig. 2) [11]. Published evidence of efficacy in this role began to accumulate [12]. Patients from distant locations were contacting Mayo Clinic to inquire about having this diagnostic test. Radiologists and hepatologists from around the world inquired about gaining access to MRE in their practices. To address this groundswell, the technology transfer office of Mayo Clinic approached the major MRI original equipment manufacturers (OEM’s) about integrating MRE technology into their systems so that it could be more widely available to patients. While recognizing the value of MRE, the companies were reluctant to take on the costs of implementing the technology on their systems and, in particular, developing and manufacturing regulatory-compliant versions of the specialized MRE driver system.

Fig. 2.

MR elastography in 54-year-old male with a clinical diagnosis of non-alcoholic steatohepatitis (NASH). The stiffness of the liver is heterogeneously increased from the normal value of less than 2.5 kPa. The mean liver stiffness of 5.7 kPa indicates the presence of stage 4 fibrosis

Motivated by the strong demand for the technology from Mayo Clinic physicians, the benefits to patients, and the steadily increasing external requests for access to the technology, Mayo Clinic founded a company (Resoundant Inc.) in mid-2007 to serve as a low-cost source of regulatory-compliant MRE driver systems for global use by the MRI OEM’s. Resoundant would provide the OEM companies with licensing and technical assistance to easily implement MRE technology into their systems. A third generation MRE driver system was developed, compliant with ISO13485 and other quality and regulatory standards for medical devices (Fig. 3). A next-generation version of MRE processing algorithm, suitable for wide deployment on MRI systems, was also developed and extensively validated.

Fig. 3.

Third-generation driver system for MR elastography. The active driver (left) is located outside the magnet room and is controlled by the MRI system. The active driver contains a powerful digitally controlled actuator that can generate air pressure waveforms, as directed by the MRI system. During imaging, the pressure waveform is transmitted through flexible tubing to a drum-like “passive driver” that vibrates the chest wall to generate shear waves in the liver

In early 2008, one of the major MRI OEM’s made the landmark decision to make MRE available as an option on their systems. The plan was announced at the annual meeting of the ISMRM in May, 2008. Our team worked closely with the company in the following months to integrate MRE acquisition techniques, hardware, and processing algorithms into the system, adhering to procedures necessary for regulatory clearance and good manufacturing practices. Test systems with the third-generation driver and inversion software were deployed with collaborators at multiple institutions in the US, Europe, and Asia, providing important feedback and eventually leading to further published evidence of efficacy. The MRI OEM submitted an application to the FDA for MRE in November 2008 and successfully received 510(k) clearance in July, 2009. Similar regulatory clearances were obtained in Europe, Asia, and South America. With growing recognition and demand for the technology, other major MRI OEM’s followed suit, collaborating with our team to integrate MRE into their systems. Currently, MRE is available as product from five major OEM’s, who supply over 95% of the MRI systems used globally.

Evolution as a standard of practice

MRI is intrinsically a quantitative imaging modality, capable of providing measurements such as relaxation times, water diffusion metrics, and many other biomarkers useful for characterizing tissue. Unfortunately, proprietary technical differences in the way that these quantities are imaged and measured on different MRI systems have limited comparability across platforms and wider use in clinical practice. Recognizing this issue, we made a strategic decision to promote standardization in the implementation of MRE by the OEM’s in a way that would support cross-platform reproducibility. This approach has been successful. All regulatory approved commercially available versions of MRE use similar acquisition techniques and shear wave driver systems, and identical processing technology. All versions provide quantitative results in a standardized format, including image scaling and display color tables [13].

The Quantitative Imaging Biomarkers Alliance (QIBA) was established by the RSNA in 2007 to bring researchers, clinicians, and industry together to advance the use of quantitative imaging in patient care and clinical trials. A key objective of these efforts is to develop and publish guideline documents, called “Profiles”, for each biomarker that assure good performance through evidence-based standardization. These profiles are being developed by over 20 individual committees within QIBA, each focused on a specific quantitative biomarker generated using CT, MRI, PET/SPECT, or US. A biomarker committee for liver MRE was formed in 2015. Leveraging the intrinsic standardization of MRE and the availability of an extensive evidence base, the committee quickly developed a Consensus Profile that was approved by QIBA and published in 2018. In early 2022, the MRE Profile was advanced to “Stage 3: Technically Confirmed” by QIBA. MRE is the first MRI-based quantitative technique to reach this stage [14].

A standard four slice MRE examination of the liver can be acquired as quickly as a single breath-hold and therefore can be included in a standard abdominal MRI protocol with minimal effect on total exam time. In appropriate settings, MRE can also be used as a focused, standalone exam. Adding a proton density fat fraction (PDFF) acquisition adds only one extra breath-hold to such an exam. This abbreviated “Hepatogram” protocol provides a quantitative screen for hepatic steatosis, fibrosis, and iron overload in a very short examination.

In the United States, the standard way that medical services by healthcare professionals are identified and reported is through “CPT codes”. These codes are assigned by the American Medical Association based on criteria that include confirmation that the service has received FDA clearance, has been shown to be widely available and frequently provided in clinical care, and that the efficacy of the service has been documented in the medical literature. In MRI practice, CPT codes have generally only been assigned for imaging of specific regions of the body (such as examinations of specific regions of the body (such as MRI code #74,183: MRI abdomen with and without contrast). Except for MR spectroscopy of the brain and MR angiography, CPT codes have generally not been assigned for specific MRI techniques. For instance, an MRI exam of the liver is coded and reimbursed the same, whether or not diffusion-weighted imaging is included in the protocol.

However, in 2018 the American College of Radiology (ACR) recognized MRE as a distinct modality that can be used either in combination with MR imaging or as a standalone Test. The ACR submitted an application to the AMA to approve a full Category I CPT code for MRE. After review, the AMA approved the application and assigned CPT code #76,391 to MRE in 2019. Subsequently, the Centers for Medicare and Medicaid Services (CMS) assigned a value to the service that equates to a government reimbursement of approximately $240. This relatively low cost reflects the modest scanner time required for the procedure and positions MRE for broad clinical use, particularly in the management of fatty liver disease.

The efficacy of MRE for assessing liver fibrosis has been established in dozens of peer-reviewed published studies to date and has been evaluated in several meta-analyses [15–23]. Consensus has emerged that among available non-invasive tests (laboratory or imaging-based), MRE has the highest diagnostic performance for detecting and staging liver fibrosis. The role of MRE in clinical care of patients with chronic liver disease is being increasingly reflected in clinical guidelines developed by professional organizations [24, 25]. Further, MRE is becoming widely adopted as a key quantitative biomarker in clinical drug trials of prospective therapies for chronic liver disease.

Recent publications have indicated that measurements of baseline liver stiffness by MRE in patients with chronic liver disease are highly predictive of the development of cirrhosis within five years [26]. Baseline measurements of liver stiffness in patients with biopsy-proven cirrhosis are also predictive of future decompensation and mortality [27]. In addition, liver stiffness by MRE is an independent predictor of cardiovascular disease in patients with NAFLD [28].

Future prospects

By all measures, the story of MRE is far from over. The next generation of liver MRE technology, known as 3D vector MRE, provides access to new biomarkers in addition to tissue stiffness (Fig. 4) [29, 30]. Many other applications are showing promise for addressing important clinical needs. Researchers around the world have worked to develop and explore clinical applications for abdominal, pelvic, pulmonary, cardiac, musculoskeletal, and neurologic diseases [31–36]. It seems likely that next widely adopted application will be for brain imaging [31, 35]. MRE is now routinely used at the Mayo Clinic for preoperative assessment of the mechanical consistency and adhesion of skull base tumors to aid surgical planning. MRE is also emerging as a unique way to study neurodegenerative diseases. For instance, studies are showing that MRE has substantial promise for discriminating normal pressure hydrocephalus, a surgically treatable condition, from other causes of dementia [37].

Fig. 4.

MR elastography can provide access to additional biomarkers for characterizing tissue. The biomarker informally called “stiffness” is calculated from the magnitude of the complex shear modulus, which consists of the storage modulus (measuring elastic properties) and the loss modulus (measuring viscous properties). Measurements of the loss modulus at low frequency are promising for characterizing inflammation. MRE data can be processed to obtain other biomarkers which include measurements of attenuation, volumetric strain, and local octahedral shear strain, each of which has shown promise for specific diagnostic applications

Lessons learned

The journey recounted here highlights a number of strategies that may be helpful to innovators in radiology as they strive to translate their work to clinical practice.

1. Seek validation in clinical practice

Many novel imaging techniques that are published in our journals never become widely adopted because their clinical value has not been established. In the case of MRE, the strategy of deploying the newly developed technology at multiple locations for clinical testing under IRB oversight provided rapid validation and showed that it could be implemented at scale. Meaningful participation of referring hepatologists as co-investigators in the trials provided important practical perspectives. After gaining experience with the results, these clinicians provided motivation and advocacy for transitioning the technique from research to clinical practice at Mayo Clinic.

2. Share your innovation with other investigators

Proactively sharing MRE technology with talented external collaborators was extremely helpful in moving the field forward. Beginning in 2007, our research team provided prototype MRE hardware and advanced MRE acquisition and processing software for exploring liver, abdominal, cardiac, muscle, and brain elastography applications at no cost to more than 100 research collaborators in five continents. This has fostered development of global expertise in the technology, contributed to an expanding evidence base, and has resulted in nearly 200 independent publications by collaborators.

3. Engage available institutional technology transfer expertise

Engaging the institutional technology transfer organization at Mayo Clinic provided a critical boost that was needed in order to make MRE available to patients everywhere through the MRI OEM’s. Most academic institutions have technology transfer offices that can provide important assistance in disseminating new technology. The modest investment by Mayo Clinic to create a source of specialized hardware that was needed by the MRI manufacturers to implement MRE was a key factor that allowed the extraordinarily rapid transition of this technology from laboratory to FDA-cleared product in just five years.

4. Proactively manage conflicts of interest

Conflicts of interest are common and unavoidable in productive academic life. Conflict-of-interest Review Boards of academic institutions recognize that the appropriate goal is not to eradicate conflicts of interest but rather to identify and manage them appropriately [38]. Conflict of interest management plans for research activity are designed to ensure the safety of human research subjects, to protect the scientific integrity of the research, and to protect the interests of participating colleagues, students, and the institution. Management can include measures such as disclosure, as well as requiring participation and corroboration of research by non-conflicted co-investigators. During the planning for dissemination of MRE technology by the Mayo Clinic, our team recognized that this would create financial interests that would need to be managed effectively in order to allow ongoing research to take place. Accordingly, our team sought and obtained a comprehensive management plan from the institutional Conflict of Interest Review Board before a decision was made by Mayo Clinic to work with industry to disseminate the advance.

5. Promote evidence-based standardization

Over the last decade there has been increasing awareness of the benefits for patients, providers, and industry from standardization of quantitative imaging technology. In the case of MRE, the strategy of implementing the technique in a consistent manner across different MRI platforms had the benefit of streamlining the process of regulatory approval in the US and in other countries, promoting the use of the biomarker in multicenter clinical trials, and accumulating consistent multicenter evidence for clinical use and insurance coverage.

6. Tap the power of Convergence Science

Multidisciplinary team science was an essential strategy in the journey from invention to standard of care. From the very beginning, the project required participation of individuals with expertise in physics, mathematics, and engineering, in addition to the disciplines of biomedical science and patient care. This approach has been called “Convergence Science” [39]. In addition to the participation of colleagues from many different disciplines, our experience showed that students can make truly significant contributions during their learning experience. In fact, graduate and post-doctoral students are named inventors in the majority of the nearly 30 documented inventions related to MRE technology assigned to the Mayo Clinic.