Emerging Techniques and Future Directions

INTRODUCTION

This article focuses on emerging technologies that have the potential to improve the speed, efficiency, and availability of MRI in emergency settings, specifically, fast MRI and portable MRI. Although these technologies have been topics of research interest for some time, the past several years have seen increasing efforts toward clinical evaluation and adoption of fast MRI and portable MRI in the emergency department and other acute care settings. These techniques are increasingly adopted in clinical research studies, and we anticipate will play an increasing role in emergency radiology over the coming decade.

An increasing number of hospitals are moving toward 24/7 availability of on-site MRI, which is currently required by The Joint Commission for Comprehensive Stroke Center certification.¹ While stroke may be the most compelling indication needing improved MRI speed and availability for neuroimaging in the emergency setting (see R. Gilberto González’s article, “Diffusion MRI of Large Vessel Occlusion Ischemic Stroke for Treatment Selection,” in this issue), improved MRI could also add value in the diagnosis, prognosis, and management of a wide variety of acute intracranial pathology (see Damien Galanaud and Rajiv Gupta’s article, “MR Imaging for Acute Central Nervous System Pathologies and Presentations in Emergency Department,” in this issue). Furthermore, its lack of ionizing radiation makes MRI a safer choice for pediatric and pregnant patients (see Abigail Stanley and colleagues’ article, “Magnetic Resonance Imaging of Acute Abdominal Pain in the Pregnant Patient”; and Maria Gabriela Figueiro Longo and colleagues’ article, “Pediatric Emergency Magnetic Resonance Imaging,” in this issue).

The use of MRI in the emergency department has traditionally been limited by availability, long scan times, and sensitivity to motion.² Lack of timely scanner availability or delays induced by the difficulty transporting a patient to the MRI suite can prolong inpatient hospital stays and negatively impact hospital revenue and patient/provider satisfaction.^3,4 Long MRI scan times require longer booking slots, limiting patient throughput and potentially delaying diagnosis and management. Long scan times are also associated with higher sensitivity to motion, a particular challenge in the emergency department.

For all these reasons, techniques to improve the efficiency of clinical brain MR examinations have gained increasing attention. Past efforts to develop ultrafast brain MRI examinations suffered from artifacts and poor image quality, limiting robust clinical use across multiple contrasts. Recent advances in MR technology have led to the development of accelerated acquisition techniques that have the comparable diagnostic quality to conventional sequences.^5–9 Decreased acquisition time has the further benefit of reducing image degradation due to patient motion and has shown value for optimizing workflow and throughput.¹⁰

Complimentary to the benefits of increasing the speed of the conventional examination, portable MRI used at the point-of-care is emerging as a revolutionary technology with the potential to bring rich diagnostic information directly to the patient’s bedside. This paradigm shift requires dramatic changes in the hardware design, clinical workflow, and potential applications of MRI in the emergency setting.

This article will review ultrafast MR technologies and portable MRI systems (including the first FDA approved and commercially available portable MRI scanner), and the major clinical use cases in the emergency department setting.

ULTRAFAST MR TECHNOLOGY

From some historical perspective, fast imaging has not always been motivated by the desire to rapidly evaluate critically ill patients in the emergency setting. Echoplanar imaging (EPI) was first described by Sir Peter Mansfield in 1977, far before the technology became commercially available and successful. Ian Pykett and Richard Rzedzian, both of whom trained under Mansfield, founded Advanced NMR in the 1980s, a company to develop EPI technology for cardiac imaging. Advanced NMR installed its first EPI dedicated gradient system at the Massachusetts General Hospital in 1990.¹¹ Early applications of EPI focused on cardiac motion in an attempt to freeze the motion of the beating human heart. However, the technology truly exploded with the advent of dynamic time-series imaging of brain hemodynamics by Bruce Rosen and Arno Villringer at the MGH-NMR Center, and became a ubiquitous research tool after the discovery of the blood oxygen level dependent (BOLD) effect and the application of EPI to enable functional MRI of human brain activations.^12–14 The development functional MRI, diffusion MRI and dynamic susceptibility contrast (DSC) perfusion MRI spurred the wide clinical and commercial adoption of EPI brain imaging. However, it has only been in the past few years that the hardware, pulse sequence development, and acquisition strategies have emerged to allow ultrafast multi-contrast brain imaging entirely based on EPI.^15,16

However, EPI is just one of the many fast imaging strategies that have been used to facilitate fast brain imaging.^{5–9,17–19} These techniques typically fall into 3 broad categories: (1) faster image data acquisition, (2) acquisition of less imaging data (ie, under-sampling), or (3) a combination of both (Fig. 1). The fundamental principles for achieving short scan time have been known for some time, but successful implementation requires gradient improvements, adequate coil sensitivity, sequence development, and reconstruction algorithms. Advances in these diverse technologies for faster clinical scanning have propelled them to the forefront of clinical imaging in recent years. An overview of the major approaches for fast MRI follows later in discussion.

Fig. 1.

Ultrafast MR acceleration techniques. CAIPI, controlled aliasing in parallel imaging; ms-EPI, multishot echo-planar imaging; ss-EP, single-shot echo-planar imaging.

Ultrafast Spoiled Gradient Echo

Gradient echo (GRE) imaging readouts use 2 gradients that are opposite in direction in rapid succession. The first dephasing gradient accelerates the dephasing and squelching of the free induction decay (ie, artificially shortens T2*).^20,21 The second rephasing gradient, which is opposite in polarity but same in strength, reverses the phase scramble to generate a GRE. In the course of the unscrambling, a single line of k-space data is collected. To acquire k-space lines as fast as possible, ultrafast GRE uses a very small flip angle, as low as 5° to 10°, very short repetition time (TR), and short echo time (TE). In comparison, conventional spin-echo sequences use flip angles close to 90° and 180° for the excitation and refocusing pulses, respectively, very short repetition time (TR) and short echo time (TE). Unfortunately, short TR and small flip angles can result in decreased tissue contrast. To preserve T1 contrast, a magnetization preparation pulse, commonly a 180° inversion pulse, is often applied at the beginning of acquisition, that is, Magnetization Prepared RApid Gradient Echo (MP-RAGE) as initially described by Mugler in 1990.²² Optimized k-space filling can be performed following variable trajectories (linear, centric, square-spiral or elliptical spiral). As ultrafast GRE depends on the T2* decay, it is susceptible to field inhomogeneity.

Echo-Planar Imaging

EPI has developed dramatically since it was first described by Mansfield in 1977 and is now used clinically to achieve rapid acquisition of images in 20 to 100 msec per slice.²³ Unlike conventional spin-echo sequences whereby one line of k-space is acquired with each TR, EPI acquires multiple lines of k-space after a single RF pulse. To achieve this, the frequency-encoding gradient oscillates from a positive to a negative amplitude, continuously producing the dephasing and rephasing phenomena described under the GRE image. Thus, rather than a single GRE forming (as in GRE), a continuous GRE train is formed, each providing a line of k-space data. The echo train length, also known as the EPI factor, is the number of k-space lines encoded in a single shot and is an important determinate of acquisition speed. Other prepared contrasts such as a spin-echo (to achieve T1- or T2-weighted imaging), inversion recovery, and diffusion sensitizing gradients can also be applied immediately prior to the EPI readout.

There are 2 main types of EPI—single-shot and multi-shot. In single-shot EPI (ss-EPI), all the needed k-space lines for a 2D image are acquired after a single RF excitation, and the k-space lines are filled by an echo train with multiple gradient reversals. A constant phase-encoding gradient was used in the original EPI technique resulting in zigzag filling of k-space. This has been nearly completely replaced by the newer “Blipped” technique, whereby each echo is phase-encoded by its own “blip” which advances the k_y line location at the end of each readout, resulting in recti-linear sampling of k-space.^24,25

In multi-shot EPI (ms-EPI), the k-space data are acquired in multiple segments (“shots”), with each shot acquiring a fraction of the total k-space.²⁵ The shots are repeated until the full set of k-space data is collected. As fewer k-space lines are collected per shot, there is less time to build up phase errors.²⁶ This results in reduced image distortion from regions of the head whereby the tissue adversely affects the field homogeneity and causes a susceptibility artifact. Because of the image distortion and as higher resolution scans require more k-space coverage, ms-EPI has been the preferred way to acquire high-resolution EPI scans (at a cost of increased scan time and increased sensitivity to motion). Figs. 2 demonstrate example images of msEPI sequences compared with conventional reference sequences.

Fig. 2.

(A)ms-EPI MR sequences (T1-weighted, T2/T2*-weighted, FLAIR, DWI/ADC) compared to a conventional reference clinical protocol based on turbo-spine echo (TSE) brain MR sequences in a patient with old cortical infarct in the right precentral gyrus with associated hemosiderin staining. (B) The red magnification box demonstrates similar conspicuity of hemosiderin staining in the right precentral gyrus (red arrowheads). The blue magnification box demonstrates a similar appearance of FLAIR hyperintensity in the right precentral gyrus, suggestive of the old cortical infarct (blue arrowheads). No DWI hyperintensity was seen on the MSEPI or conventional images.

Parallel Imaging

One way to speed up the acquisition is to acquire, for example, only every other or every third k-space line (k-space undersampling). Undersampling k-space reduces the image FOV in the under-sampled direction. This causes aliasing artifact unless directly addressed in the image reconstruction through a parallel imaging reconstruction such as SENSitivity Encoding (SENSE) or GeneRalized Autocalibrating Partial Parallel Acquisition (GRAPPA) which utilize multiple receiver coils with known placements and sensitivities to determine where the MR signal arises and “undo” the aliasing. The positional information provided by the multichannel phased-array coils supplements the incomplete k-space data and prevents aliasing.^27,28 The parallel imaging acceleration factor, R, is the ratio of k-space data required for a full y sampled image to that acquired in the under-sampled image. It is also the factor by which the image acquisition is sped up compared with the conventional approach. An acceleration factor of 1.5 to 4 is commonly used.¹⁹

There are 2 general reconstruction classes of parallel imaging reconstructions. The first being image domain reconstruction (reconstruct then correct), which unfold images using a coil-sensitivity map from the aliased image.¹⁹ The second method is k-space domain reconstruction (correct then reconstruct), whereby the under-sampled k-space information is auto-calibrated before Fourier transformation. In both cases, the image suffers an SNR reduction due to 2 factors; (1) the reduced data from having acquired fewer k-space lines, and (2) a noise amplification in the image reconstruction due to imperfect spatial information supplied by the array coil geometry. The first scales simply as the square root of the acceleration factor R, and the second factor is termed the “geometry factor” or g-factor and is typically between 1 and 3, although it can become very large if too much acceleration is attempted.

Simultaneous Multislice Imaging

While conventional parallel imaging speeds up the acquisition by omitting in-plane phase encoding steps, some acquisitions can also be sped up by encoding multiple slices simultaneously. The principle underlying simultaneous multislice (SMS), also known as a multiband imaging, is the simultaneous excitation and readout of multiple slice planes to reduce TR and acquisition time and/or extend slice coverage in a given TR.²⁹ SMS is most widely used for EPI sequences as simultaneous slice acquisition directly translates to shorter TR in most cases, although SMS versions of turbo spin-echo and GRE sequences have also recently been introduced and are commercially available.

The acceleration achieved with SMS is similar to standard parallel imaging, but there are a few subtle differences arising from operating on the slice selection process rather than the phase encoding. For example, the RF power demand and thus SAR concern is higher in SMS as it uses a high peak-power composite RF pulse for multi-slice excitation.²⁹ But SMS incurs less SNR penalty than conventional parallel imaging methods as k-space is not under-sampled (so there is no √R SNR penalty). If the g-factor of the acquisition is near unity, there is little to no penalty in SNR; a substantial advantage of SMS over conventional parallel imaging. This is because SMS acceleration does not reduce the echo train length, number of phase-encoding steps, or number of k-space samples. SMS can introduce artifacts, such as the spread of signal across simultaneously acquired slices, termed inter-slice leakage, which can amplify the effects of motion.^30,31 While methods to reduce inter-slice leakage such as slice-GRAPPA have been developed, inter-slice leakage artifacts can still occur and affect diagnostic quality.³² Fig. 3 demonstrates an example of SMS accelerated DWI and ADC image pairs.

Fig. 3.

Brain MRI in a patient who underwent right frontal craniotomy for the subtotal resection of right frontal anaplastic astrocytoma. Magnetization-prepared rapid gradient-echo (MPRAGE) image demonstrate ill-defined and heterogenous areas of enhancement in the right frontal lobe, basal ganglia, right anterior thalamus, and corpus callosum, which is concerning for tumor progression. TSE FLAIR image demonstrates hyperintensity throughout the right frontal lobe, subcortical structures, corpus callosum, and the left frontal lobe with associated worsening mass effect, right ventricular effacement, and leftward midline shift. Simultaneous multislice (SMS) accelerated DWI/ADC images demonstrate restricted diffusion with ADC hyperintensity in the right basal ganglia corresponding to the region of ill-defined enhancement, furthering concern for tumor progression. There are surrounding areas of DWI and ADC hyperintensity in the right frontal lobe that reflects T2 shine through. There is no evidence of acute infarction.

Wave-Controlled Aliasing in Parallel Imaging

Volumetric imaging has become increasingly important in neuroimaging for surgical planning and assessment of various pathologies such as brain tumors.³³ Acquisition times, however, are often long and greatly increase the total examination time. Parallel imaging is often applied to reduce the acquisition time of volumetric sequences. In conventional 3D parallel imaging, the image data can be under-sampled in the two phase-encoded directions (commonly in the y and z planes) to reduce acquisition time. While better than putting all the acceleration into a single direction, large total acceleration factors can still result in a high SNR penalty, often prohibitive above a total acceleration factor of approximately 3 to 4.³⁴

A limitation of parallel imaging is that the modern coil array is relatively poor in discerning elements in untangling aliasing in the z-direction, thus, in the SMS method, the excitation of the multiple slices must be spaced widely apart. CAIPI assists the role of the coil sensitivity in the z-direction by using the phase of the RF pulse to shift the position of adjacent slices in a controlled fashion making the aliasing easier to undo.³⁵ CAIPI spreads voxel aliasing more uniformly and allows for more efficient disentanglement of adjacent slices.

Wave-CAIPI builds on CAIPI by using sinusoidal G_y and G_z gradients applied during readout of each k-space line.³⁶ The characteristic corkscrew trajectory leads to the optimal synergy between aliasing and coil sensitivity profiles, allowing for highly accelerated volumetric imaging with low artifact and negligible g-factor penalty.³⁶ Wave-CAIPI has been used to accelerate 3D sequences such as MPRAGE (Fig. 4), SPACE-FLAIR (see Fig. 4), and SWI (Fig. 5).

Fig. 4.

WAVE-CAIPI SPACE-FLAIR (top row) and WAVE-CAIPI MPRAGE (bottom row) images demonstrate scattered foci of T2/FLAIR hyperintensity with corresponding T1 hypointensity in the periventricular, subcortical, and juxtacortical white matter throughout the supratentorial brain (arrow). This is consistent with a reported history of multiple sclerosis. There was no parenchymal enhancement to suggest active demyelination.

Fig. 5.

Motion artifact obscured visualization of supra- and infra-tentorial microhemorrhages on the standard SWI sequence due to the long acquisition time (4 min 33 sec). The presence of microhemorrhages is better seen on the WAVE-CAIPI SWI sequence, which is less affected by motion due to the shorter acquisition time.

Compressed Sensing

Compressed sensing (CS) is founded on the basis of reconstructing an image from incomplete k-space sampling by exploiting the existence of a sparse representation of the image (ie, a representation that will be represented with only a few variables), which is mathematically related to the ability to compress an image such as the well-known jpeg compression scheme.^37–39 Indeed, the inspiration for CS partly came from attempts to solve medical image compression. One of the solutions to medical image compression was lossy compression, whereby certain data elements are discarded.⁴⁰ Several research groups then applied this approach in reverse for MR imaging—if medical image data can be compressed by discarding certain information, the initial image data acquisition can be limited to this compressed (under sampled) data-set which can be acquired much faster. Compression is simply a more efficient way of describing the image data; a general property of “naturalistic” images, whether photographs of scenes or medical images, is that they can be stored in this compressed representation and then reverted back with relatively little image degradation. The “compressed” data contain enough information to reconstruct images of similar quality as those that would arise from a fully sampled set.^37–39

Application of CS in MR imaging is possible because of redundant image data and transform sparsity, meaning the object or image we are trying to recover has a sparse representation in a constrained transform domain.³⁹ Sparse representation means that essential information is contained in a few high-valued coefficients and most of the other coefficients can be discarded with no or negligible loss of information.⁴¹ The final MR image can thus be recovered from under-sampled data using iterative knowledge-based reconstruction to fill in the empty k-space lines. The reconstruction method is nonlinear and enforces 2 conditions: (1) the image data exhibit transform sparsity and (2) consistency of the reconstruction with the acquired data.⁴² The first condition allows aliasing artifacts to be separated from the actual image signal and the second condition ensures that the actual image data are not replaced with arbitrary data. These 2 conditions balance the sparsity of the image and reconstruction of an image that is consistent with the collected image data. With CS, acquisition time can be reduced by half while maintaining high diagnostic quality. Because CS relies on image sparsity and not hardware, it is also not limited by the Nyquist limit like other techniques such as parallel imaging.³⁹ Compressed sensing has been applied to various 2D and 3D brain MR sequences such as T2-weighted imaging, FLAIR, 3D T1-echo-spoiled gradient echo (SPGR), and TOF MRA (Fig. 6).^18,43,44

Fig. 6.

Example of CS-TOF MRA (A) and standard TOF MRA (B) MIP images in a healthy volunteer. Visualization of the proximal major intracranial vessels is similar between CS-TOF MRA and the standard TOF MRA images despite a 5-fold reduction in acquisition time for the CS sequence.

CLINICAL USE OF FAST MAGNETIC RESONANCE IMAGING

Pediatric Patients

MRI is preferred over CT as the imaging modality of choice for pediatric patients due to MRI’s lack of ionizing radiation and excellent anatomic assessment of the brain. The same ionizing radiation dose from CT results in a 10-fold higher neoplastic potential for a child than an adult.⁴⁵ MRI is also superior to CT for evaluating the posterior fossa and ischemic stroke.⁴⁶ The long acquisition time, together with motion artifact associated with imaging young patients, can limit the diagnostic quality of MRI and may require sedation in some settings. Techniques to accelerate MRI acquisition have been used to overcome these limitations and studies have demonstrated clinical benefit for imaging pediatric patients in the emergency setting.⁴⁷

Pediatric patients with ventricular shunt placement for hydrocephalus receive routine and emergency imaging to evaluate for shunt dysfunction. Single-shot fast spin-echo and ultrafast GRE T2-weighted imaging have been shown to be sufficient in most cases for catheter visualization and evaluation of shunt dysfunction.^48,49 Application of ultrafast brain MR imaging can also help avoid sedation and reduce motion artifact.⁵⁰

Combining different acceleration techniques such as parallel imaging and EPI, a recent group from Korea demonstrated that a 1-min ultrafast pediatric brain protocol consisting of T1-, T2-, and T2*-weighted imaging, FLAIR, and DWI sequences showed sufficient image quality for diagnostic use for a broad range of indications.¹⁷ Accelerated MRI sequences have also been used for pediatric traumatic brain injury, including the use of single-shot fast spin-echo T1-and T2-weighted sequences, and accelerated EPI T1-weighted, FLAIR, and DWI sequences.⁵¹ Kralik and colleagues reported similar sensitivity and specificity for the detection of trauma using accelerated EPI-DWI and EPI-T2*-weighted images when compared with their conventional counterparts.⁵² Limitations of ultrafast brain MRI in pediatric patients include reduced sensitivity for certain pathology such as skull fractures and very small hemorrhages.^52–55

Stroke Evaluation

Neuroimaging plays a critical role in the initial evaluation of acute ischemic stroke. MRI is superior to CT for the assessment of parenchymal integrity, tissue viability, and ischemic core.^56,57 DWI / FLAIR mismatch on MRI is suggestive of ischemic stroke less than 4.5 hours, which is a crucial patient selection criteria for IV-tPA, and is extremely helpful when the symptom of onset is unknown.^58,59 DWI also allows for the superior assessment of ischemic core volume than other techniques such as CT perfusion, and facilitates selection of patients likely to benefit from mechanical thrombectomy and late-window thrombolysis.

To reduce scan time and reduce motion degradation, fast brain MR protocols for acute stroke evaluation have been developed using ss-EPI and parallel imaging techniques.⁶⁰ A previously proposed protocol for comprehensive stroke evaluation included accelerated versions of DWI, FLAIR, GRE, DSC perfusion, and MRA.⁶⁰ The acquisition time of this protocol was 6-min long and demonstrated diagnostic image quality in greater than 90% of the cases. When the acceleration of ss-EPI and parallel imaging is pushed too far, however, image distortion and dropout artifacts can occur due to local field inhomogeneity and geometric distortion. Furthermore, SNR and sensitivity of detecting small areas of ischemia or hemorrhage are reduced.

Multi-shot echoplanar imaging (ms-EPI) is a high efficiency interleaved EPI imaging technique that uses multiple excitations, resulting in significantly reduced geometric distortion and higher SNR than ss-EPI.^61,62 Prototype ms-EPI brain MR protocol (T1-, T2-, T2*-weighted, FLAIR, and DWI) are being evaluated at our institution (Fig. 7).⁶³ The prototype fast brain ms-EPI-based protocol totals approximately 2 minutes. Preliminary results demonstrated that while noise and artifact were increased on ms-EPI sequences compared with conventional sequences, the detection of image findings, including ischemic stroke evaluation, was not compromised.⁶³ Further efforts using machine-learning algorithms to improve image quality and to decrease image noise and artifact are currently ongoing.

Fig. 7.

ms-EPI MR sequences compared with standard brain MR sequences in a patient presenting with aphasia for 2 weeks. In both the ms-EPI and standard brain MR images, there was obvious T2/FLAIR hyperintensity in the left temporal lobe with subtle associated high DWI signal, suggestive of late subacute infarction. The overall diagnostic quality of the images was similar between the ms-EPI and the standard images. The acquisition times for the sequences are listed below each image in seconds.

Acute Nontraumatic Neurologic Presentation

Approximately 15% of emergency department visits are due to acute neurologic symptoms.⁶⁴ EPI and parallel imaging have been used in the emergency setting beyond ischemic stroke evaluation in these scenarios. Kazmierczak and colleagues and Prakkamakul and colleagues deployed ultrafast brain MRI protocols in the emergency and neurologic intensive care unit settings, respectively.^65,66 These studies included a broad range of acute neurologic symptoms such as vertigo, paresthesia, impaired vision, aphasia, motor deficits, memory impairment, tremor, and others. They found that the ultrafast brain MRI protocols were superior to CT for detecting acute intracranial pathologies. While the ultrafast protocol was noninferior to the standard brain MRI protocol and did not impact clinical care, several pathologies were missed or were less conspicuous on the ultrafast MR images. Depending on the symptoms and suspected pathology of the patients, the protocols can be individualized by adding or replacing certain sequences (eg, a 3D SPACE FLAIR offers higher sensitivity for demyelinating lesions than conventional 2D FLAIR if this is an acute concern; susceptibility-weighted images offer greater sensitivity for small foci of intracranial hemorrhage, and so forth). Recently, an adaptable ultra-fast brain MRI protocol “Neuro-Mix” was proposed, whereby image quality can be dynamically adjusted by substituting different fast MRI sequences depending on the clinical indication and the patients’ ability to tolerate MRI scanning.⁶⁷

Operational Impact of Fast Magnetic Resonance Imaging

Reducing scan time is beneficial for improving patient care and access. For certain populations, including patients who are acutely ill, in pain, have claustrophobia, or altered in mental status, remaining still for prolonged periods of time for conventional brain MRI can be very difficult. Implementation of accelerated MR sequences can increase access for patients that would otherwise not tolerate the long scan times of standard MR protocols. Further advantages include a reduced number of repeat scans from motion degradation and improved patient satisfaction and comfort, all the while maintaining adequate diagnostic imaging quality.

The utilization of diagnostic imaging, including brain MRI, in the emergency setting has also been increasing.^68,69 This is partly due to increased MRI scanner availability, increased demand by clinicians and patients, and increased surveillance of patients with diseases such as cancer.^70–72 Strategies to meet the increasing imaging demand include acquiring more scanners, hiring of additional personnel, and acquisition of new space, but these are all associated with significant financial costs. Accelerated MRI protocols can improve scanner productivity without the additional financial burden and mitigates downstream costs.⁷³ Implementation of accelerated brain MRI sequences at our institution was shown to reduce acquisition time by approximately 40% in the outpatient setting, suggesting that substantial improvement in operational efficiency can potentially be achieved in the emergency department as well.

POINT-OF-CARE AND PORTABLE MAGNETIC RESONANCE IMAGING

Expanding the Role of Magnetic Resonance Imaging in the Emergency Department with Small Foot-Print and Portable Scanners

The size, footprint, and siting needs of conventional high-field MRI scanners have traditionally prohibited their installation in specialized areas such as the emergency department (ED) or Intensive Care Unit (ICU). Recent technological work has focused on altering the system design to reduce the siting constraints and even produce truly portable MRI scanners that can be wheeled to the patient bedside. This has been primarily enabled by the design of scanners focusing on a particular body part, such as the brain, and by lowering the field strength of the magnet and compensating for the reduced image quality with computational filtering approaches (denoising) and utilization of state-of-the-art sequences and receive-coil technology. Other technical components that must be addressed to reduce system cost and siting needs include the magnet’s magnetic footprint, power consumption, cooling of electrical components, cryogen use and venting, acoustic noise, and reduction of artifacts from electro-magnetic Interference (EMI) sources. The latter are typically eliminated through the use of a specialized shielded room, which must be eliminated in a truly portable scanner. When discussing portable and point-of-care MRI, it is important to keep in mind that the goal is not replacing conventional high-field scanners, but rather to supplement them by providing systems that expand the use of MRI in a new setting.

Rethinking System-Level Approaches for Portable Magnetic Resonance Imaging

To reimagine MRI as a portable device for use in the ED, a new system-level approach is needed. The simplest approach is to continue the industrial effort into shortening the bore of conventional super-conducting magnet-based systems such as the current standard (the 1.5 T scanner). However, even after decades of effort, this system remains a multi-ton, nontransportable system with high power and cooling needs and a relatively large magnetic field footprint. Relaxing some of the features of this workhorse conventional scanner is required to make it truly easy to site in an ED (or portable). Reducing the focus from whole-body to brain-only yields the largest potential gains for increasing the range of siting locations. Other obvious departures include lowering the static magnetic field strength (at the expense of sensitivity) and/or the magnet’s homogeneity, which alters the imaging pulse sequences that can be used.^74–76 The result is cheaper, smaller superconducting magnets or permanent magnets where the magnetic energy density can be stored without the use of cryogens. Other departures include nonswitched readout gradients built into the static magnet design (saving power, cooling, and reducing acoustic noise),⁷⁷ encoding by the rotation of a built-in gradient (further reducing encoding electronics),^78–83 and shrinking the imaging field of view (FOV) even further to a subset of the organ and perhaps not fully encoding all spatial dimensions.

Three Strategies for Portable Magnetic Resonance Imaging in the Emergency Department

We arbitrarily divide ED MRI use cases into 3 levels based on their degree of deviation from the standard 1.5 T scanner suite. The closest level uses modest deviations and attempts to improve siting (and perhaps cost) to facilitate siting within a tight ED space. This “easy-to-site suite” scanner could use a standard superconducting solenoid magnet architecture, perhaps at reduced field strength, but with modifications to decrease its cost, size, and stray-field footprint (Fig. 8). For example, the system might use a short-bore, conduction-cooled superconducting magnet to eliminate cryogens and the quench pipe. The footprint, size, and cost can be further reduced compared with conventional whole-body scanners if the magnet is sized for brain-only imaging and operated at mid-field (between 0.5 T and 1.0 T). This intermediate field strength is attractive because it can provide sensitivity and imaging contrasts similar to conventional 1.5 T scanners. This direction has been recently reviewed and put into historical perspective.⁸⁴ An additional technology that is on the near horizon is the addition of active Electro-Magnetic Interference (EMI) mitigation to eliminate the standard RF shielded room. High-field (3T) superconducting head-only scanners can also be considered similarly “easy-to-site.” These include the Siemens Allegra 3T clinical scanner⁸⁵ introduced in the early 2000s but no longer produced, and the more recent GE high-performance 3T head scanner using a conduction-cooled magnet with no cryogen vent pipe.⁸⁶ While the magnet and gradients of these high-field head-only systems are more compact, the focus of these 2 systems was on performance rather than siting alone.

Fig. 8.

Commercially available point-of-care and portable MRI systems. (A) Dedicated head-only MRI scanner for point-of-care imaging from Synaptive. (B) Siemens Healthineers moves into new clinical fields with its smallest and most lightweight 80 cm bore whole-body MRI. (C) Small-footprint, lightweight, high-performance 3T MRI scanner for advanced brain imaging with reduced installation costs. (D): Hyperfine Swoop^® Portable MR Imaging SystemTM. (From [A] Panther et al (reference 88), Proc. ISMRM 2019 p. 3679, owner ISMRM; [B] Siemens Healthineers SHAPE 21 Imaging Press Conference, November 18, 2020. Available at: https://www.siemens-healthineers.com/press/releases/magnetom-free-max.html. [C] Foo TKF et al, Magn Reson Med. 2018 Nov;80(5):2232–2245. PMID: 29536587. [D] Proc. ISMRM 2020 abstract 555, owner ISMRM.)

Reduction of the magnet’s B₀ field yields further siting benefits. A 0.5 T or 1.0 T superconducting magnet retains many aspects of conventional suites including high-power electrical hookups (for conventional gradients), maintenance-prone cryogenic equipment, water cooling, and a safety exclusion zone, but the potential siting benefits have motivated several commercial MRI manufacturers to initiate the development of this type of device (see Fig. 8). Two “mid-field” approaches (0.55 T⁸⁷ or 0.5 T^88–90) have been introduced leveraging superconducting systems with cryogen-free refrigeration systems. Both use modern, high-performance gradient systems in standard architecture. Other efforts are underway with an even smaller head-focused 1.5 T high-temperature superconducting magnet.⁹¹

The second level would be a truly portable scanner that could be pushed from location to location within the ED. This device must operate using a standard electrical outlet or perhaps battery power. The mobile brain scanner would likely operate at low field (50 mT to 200 mT), need unconventional EMI mitigation, and must operate with substantially reduced electrical power compared with conventional systems and without water cooling or cryogenics. This class of POC devices is being actively pursued by several companies and academic groups. Notably, the first fully portable clinical MRI product scanner is now FDA approved for clinical care, the 64 mT Hyperfine scanner shown in Fig 8D.^92,93 Fig. 9 shows some brain images taken with this system. Figs. 10 and 11 show another approach under development, based on an 80 mT “Halbach-bulb” rare earth magnet configuration with a magnet weight less than 100 kg.^77,94,95

Fig. 9.

Clinical examples of portable 0.063T MR images (axial T1, T2 and FLAIR) compared to non-contrast CT images respectively from left to right. Top row: Subacute right MCA ischemic infarct (a-d). Middle row: Bilateral subacute subdural hematomas (e-h). Bottom row: Suspected right splenial hemorrhagic mass (i-l) in a patient with metastatic Ewing’s sarcoma.

Fig. 10.

The “Halbach Bulb,” another portable MRI approach under development using an 80 mT rare earth magnet configuration with a magnet weight less than 100 kg. (A) A volunteer with head positioned inside the scanner (orange cylinder). The subject’s shoulders remain outside the scanner allowing for a lightweight small bore design that fits the head only. (B) Another view of the scanner. (C) Representative T2 (top row) and T1 (bottom row) weighted images in a volunteer, obtained using this device. (From [A] Figure 1 of Cooley, C.Z., McDaniel, P.C., Stockmann, J.P. et al. A portable scanner for magnetic resonance imaging of the brain. Nat Biomed Eng 5, 229–239 (2021); [B] Original artwork; [C], Original artwork.)

Fig. 11.

Illustration of the permanent low field magnet design used in the “Halbach Bulb” scanner. (A) Computer rendering of the rare earth metal array which creates a permanent magnetic field of 80 mT. (B) Photograph of the actual permanent magnet array used in the scanner. (C) Diagram of the scanner components, including the gradient coils, which are placed outside the permanent magnet array, and RF coil, which is placed inside the permanent magnet array. (D) Rendering of the scanner mounted on a portable frame that can be wheeled to the bedside for point-of-care imaging. (From: Original artwork.)

Finally, the third class is a more speculative device that extends MRI to a near “hand-held” level, likely with greatly reduced imaging capabilities, but inexpensive and small enough to be considered an MR detector or monitoring device more than a diagnostic imaging device. Such a lightweight device could reach into the bed and monitors the brain, perhaps using 1D imaging or just the MR signal itself. This rethinks the role of MRI as a tomographic imager and, as such, is the most distant from conventional MRI scanner architectures. Nonetheless, examples of this more speculative device are starting to emerge in the literature, such as the 7 kg device shown in Fig. 12.⁸² Even if detailed anatomy is not visualized, the MR data can be collected and monitored for changes that might accompany, for example, important intracranial pathology. This type of MRI is thus a patient monitoring device in the way a pulse-oximeter or ECG device is used at the bedside in the ED or ICU. It is new territory for MRI and pushes the technology into a more radical configuration. As a monitoring device, the device must “reach” into the bed, operate adjacent to the patient (vs placing the anatomy inside the magnet), and be light and cheap enough for sustained operation as a monitoring device. An ED or ICU might benefit from such an MRI device to continuously image the brain watching for intracranial hemorrhage or changes in cerebral mass-effect through monitoring a ventricular/CSF left-right hemisphere asymmetry. An intracranial MR monitor could provide an early warning sign of impending herniation, particularly in patients whereby the clinical examination is difficult (eg, sedated patients). Single-sided spectrometers have also been introduced for assessing breast tissue^96,97 and muscle hydration.^98,99 Single-sided full imaging systems are less common, but have been demonstrated¹⁰⁰ and applied to burn depth.¹⁰¹

Fig. 12.

(A) Concept drawing of the “MR Cap,” a 7 kg device that can be positioned above the subject’s head like a helmet. (B) Concept drawing illustrating the sensitive ROI of the scanner, superimposed on a sagittal T1-weighted image high-resolution image of the brain. (C) Images of a phantom were obtained using this scanner. Six images are shown on the right, at different depths of the phantom in the YZ (transverse) plane. (From: [A, B] from Figure 1 and C from Figure 9 of McDaniel, P.C., et al., The MR Cap: A single-sided MRI system designed for potential point-of-care limited field-of-view brain imaging. Magn Reson Med, 2019. 82(5): p. 1946–1960. Owner: Wiley; [C], Original artwork.)

The above 3 visions for point-of-care portable MRI are not exhaustive, but we hope will stimulate emergency radiologists and emergency department clinicians to begin thinking of MRI as a tool that may soon be accessible at the bedside, and to understand and imagine how this emerging technology could benefit acute care in the near future.

SUMMARY

Challenges to MRI in the emergency setting include limited availability, limited throughput, siting constraints, costs, and motion artifacts. Fast brain MRI and portable MRI are emerging as promising technologies that can improve the efficiency of MR imaging in the emergency setting, and improve the availability of MRI in patient care settings whereby cost, siting, or scan-time considerations were previously prohibitive. MR acceleration techniques result in a substantial reduction in total acquisition times, while providing accurate diagnosis for the detection of major acute pathologies such as stroke, hemorrhage, mass effect, and hydrocephalus. Point-of-care and portable MRI systems can further expand the use of MRI by increasing the number of MRI departments able to afford and site and MRI scanner, by bringing the scanner to the bedside, and one day perhaps providing real-time monitoring of acute intracranial pathology. Though it remains to be seen, the ultimate goal of this technology development and clinical translation is to provide more timely, accurate, and available diagnostic information for acutely ill patients in emergency and acute care settings.

KEY POINTS.

• Fast MRI methods decrease scan times by reducing the amount of time required to acquire the k-space data, acquiring less k-space data (under sampling), or a combination of the two.

• Fast brain MRI protocols are particularly useful in motion-prone patients such as pediatric patients or adults with altered mental status.

• Point-of-care and portable MRI methods offer the possibility to reduce costs, relax siting constraints, and extend the diagnostic power of MRI to the bedside for patient diagnosis and monitoring in acute care settings.

CLINICS CARE POINTS.

• Use of fast MRI methods can be helpful in the evaluation of motion prone patients such as pediatric patients, critically ill patients and those with altered mental status.

• Decreasing the length of MR acquisitions can have beneficial impact on operational workflow and reduce the time required to obtain critical diagnostic information for acutely ill patients.

• Point-of-care and portable MRI technology is emerging as a promising tool to bring the diagnostic power of MRI to the bedside with reduced costs and relaxed siting constraints.