Source-linked AI summary
A Portable Brain MRI Scanner for Underserved Settings and Point-Of-Care Imaging
Clarissa Z. Cooley, Patrick C. McDaniel, Jason P. Stockmann, Sai Abitha Srinivas, Stephen Cauley, Monika Sliwiak, Charlotte R. Sappo, Christopher F. Vaughn, Bastien Guerin, Matthew S. Rosen, Michael H. Lev, Lawrence L. Wald
TL;DR
MRI access is limited by scanner cost, infrastructure, and the difficulty of transporting critically ill or remote patients. This paper presents a portable 80 mT brain MRI scanner using a lightweight permanent magnet and validates it with in vivo brain imaging across multiple contrasts.
Problem
MRI diagnostics are often unavailable to critically ill, unstable, or low-resource patients because conventional scanners are costly, infrastructure-intensive, and difficult to access.
Method
The authors develop a head-only portable scanner with a compact 80 mT Halbach permanent-magnet array, built-in readout gradient, and field-map-informed iterative reconstruction.
Results
The 122 kg scanner operates from a standard wall outlet without cooling and produced validated in vivo T1, T2, and proton-density-weighted brain images.
Takeaways & Limitations
The design supports point-of-care brain MRI with standard clinical contrasts while reducing siting, power, cooling, and acoustic-noise requirements.
Takeaways & Limitations
The reconstruction does not fully address image distortion from non-orthogonal encoding fields, field-map measurement errors, and temperature-dependent B0 changes.
Abstract
from arXiv · showhide
Access to and availability of MRI scanners is typically limited by their cost, siting and infrastructure requirements. This precludes MRI diagnostics, the reference standard for neurological assessment, in patients who cannot be transported to specialized scanner suites. This includes patients who are critically ill and unstable, and patients located in low-resource settings. The scanner design presented here aims to extend the reach of MRI by substantially reducing these limitations. Our goal is to shift the cost-benefit calculation for MRI toward more frequent and varied use, including improved accessibility worldwide and point of care operation. Here, we describe a portable brain MRI scanner using a compact, lightweight permanent magnet, with a built-in readout field gradient. Our low-field (80 mT) Halbach cylinder design of rare-earth permanent magnets results in a 122 kg magnet with minimal stray-field, requiring neither cryogenics nor external power. The built-in magnetic field gradient reduces reliance on high-power gradient drivers, which not only lowers overall system power and cooling requirements, but also reduces acoustic noise. Imperfections in the encoding fields are mitigated with a generalized iterative image reconstruction technique, that uses prior characterization of the field patterns. Our system was validated using T1, T2 and proton density weighted in vivo brain images with a spatial resolution of 2.2 x 1.3 x 6.8 mm$^3$.
Introduction
The paper addresses the cost, infrastructure, and transport barriers that limit MRI access in point-of-care and low-resource settings. It presents a head-only portable scanner designed to reduce these barriers while retaining useful brain imaging capabilities.
- Motivation: Point-of-care MRI could support time-critical assessment for unstable patients, neonatal imaging, and neurological care in remote or low-resource regions.Examples include trauma, stroke, hematomas, hydrocephalus, intensive-care patients, and pediatric hydrocephalus monitoring.
- Motivation: Conventional MRI requires large, expensive, infrastructure-intensive hardware that is poorly suited to point-of-care operation.High-field magnets, switchable gradients, cryogenics, dedicated power and cooling, and safety infrastructure drive these requirements.
- Research gap: No consensus exists on the best approach for adapting MRI to portable and point-of-care use despite growing research and industrial activity.Recent efforts span superconducting, permanent-magnet, cryogen-free, dedicated-organ, and portable systems.
- Design rationale: The proposed design reduces hardware burden by specializing for brain imaging, using an 80 mT permanent magnet, embedding readout encoding in B0, and shifting constraints toward software reconstruction.The head-only geometry, rare-earth magnet, built-in field variation, and advanced reconstruction depart from the canonical scanner design.
- Scope: The device is intended to provide useful diagnostics where fixed MRI is impractical or impossible, not to replace high-field MRI.High-field MRI retains superior image quality and advanced techniques such as spectroscopy, SWI, and DTI.
- Contribution: The scanner is a 122 kg head-only system that operates from a standard wall outlet without cooling and was validated with T1, T2, and proton-density imaging.Validation used in vivo brain imaging in an RF-shielded room.
Prototype Scanner
The prototype combines a head-only Halbach permanent magnet with built-in readout encoding, switchable phase-encoding gradients, specialized RF pulses, and volumetric RARE sequences. Its compact construction targets portability while preserving multiple brain-imaging contrasts.
- Prototype Scanner: The full scanner system weighs 230 kg and combines a transmit/receive helmet coil, permanent magnet cylinder, gradient coils, RF shield, console, amplifiers, and cart.The patient table detaches from the cart, and the current system can be transported by a single person.
- Permanent Magnet: The NdFeB Halbach cylinder provides an intrinsically self-shielding, head-only magnet designed around the subject’s shoulders.The design trades field homogeneity against compactness and avoids a heavy magnetic yoke.
- Permanent Magnet: The magnet deliberately shapes field variation into a built-in x-direction readout gradient, eliminating the traditional readout-gradient coil and driver.Genetic optimization perturbed the Halbach arrangement using N42 and N52 one-inch NdFeB cubes.
- Permanent Magnet: 641 NdFeB cubes arranged in three layers of 24 rungs form the optimized magnet, with shim magnets used to improve the encoding-field shape.The design includes measured field maps before and after shimming.
- Permanent Magnet: 80 mT over the 20 cm target volume and a 7.6 mT/m built-in readout gradient characterize the 122 kg magnet assembly.The assembly has a 49 cm length, 57 cm outer diameter, and 35 cm inner diameter.
- Gradient Coils: Switchable Gy and Gz gradients provide y- and z-direction phase encoding while the permanent field supplies x-direction readout encoding.The gradient coils are designed for linearity within the 20 cm region of interest.
- Sequences: Frequency-swept chirped RF pulses cover the wide Larmor-frequency range caused by the inhomogeneous low-field magnet.The system uses 3.2 ms excitation pulses and 1.6 ms refocusing pulses with 100 kHz sweeps.
- Sequences: Long multi-echo 3D RARE spin-echo sequences exploit low-field properties and support volumetric proton-density, T1, and T2 imaging.The built-in gradient prevents standard gradient echoes, so acquisition uses spin-echo-based sequences.
Image Reconstruction
Because the compact scanner produces nonlinear encoding fields, reconstruction uses measured field maps in a forward model and solves for the image iteratively. Phantom and in vivo results demonstrate volumetric imaging at millimeter-scale resolution.
- Image Reconstruction: Compact-system nonlinearities can produce aliasing and encoding holes that undermine conventional FFT-based reconstruction.The problem is particularly important toward the periphery of the permanent-magnet gradient.
- Image Reconstruction: The reconstruction model uses measured built-in-gradient and switchable-gradient field maps to model time-domain encoding during 3D RARE acquisition.This generalized approach accounts for the scanner’s known non-linear encoding fields.
- Image Reconstruction: An iterative conjugate-gradient algorithm implemented with GPUs solves for the image from the modeled signal evolution.The reconstruction is based on prior characterization of the encoding fields.
- Validation: Brain images from three healthy subjects reached approximately 2.2 x 1.3 x 6.8 mm resolution, although resolution varied slightly across the field of view.The images were proof-of-principle acquisitions obtained in an RF-shielded room.
- Validation: A T2-weighted 3D grapefruit image acquired in 16:10 min achieved 2.2 x 1.3 x 6.8 mm^3 resolution without visible distortion.The acquisition used a 2500 ms/167 ms sequence with four averages.
Discussion
The scanner produced multiple standard brain MRI contrasts with clinically relevant soft-tissue contrast, while remaining limited by encoding-field distortion, temperature sensitivity, EMI, and workflow constraints. Validation used healthy volunteers and required an RF-shielded room.
- Imaging capability: The scanner generates standard T1, T2, inversion-recovery T2, proton-density, and diffusion-weighted contrasts for point-of-care brain imaging.Its performance is lower than high-field MRI but is reported as sufficient for detecting and characterizing serious intracranial processes.
- Validation: SNR values were 127, 80, 68, 65, and 124 for image acquisitions in rows 1–5, respectively.The generalized reconstruction uses measured field maps, whereas conventional FFT reconstruction shows geometric distortion when nonlinear encoding fields are not modeled.
- Image reconstruction: Encoding-field distortion is most severe for the built-in Gx gradient, especially near the periphery where spatial deformation can approach signal singularities.The Gz and Gy gradient-coil maps show less severe errors and spatial deformation.
- Limitations: Temperature-dependent NdFeB field variation can introduce reconstruction-model errors, although the study assumes no significant temperature change during each approximately 10-minute acquisition.Field probes are proposed to track global or local B0 changes during acquisition.
- Workflow and patient interface: The compact head-only geometry creates a tight fit requiring further attention to patient positioning, airflow, and monitoring, particularly for intubated or highly monitored patients.Healthy volunteers nevertheless found the scanner comfortable during acquisition sessions exceeding 45 minutes, and operation was acoustically quiet.
- Limitations: Human pilot imaging required a traditional Faraday cage because the built-in shielding did not fully eliminate electromagnetic interference.The body outside the shield can conduct EMI into the receiver coil and degrade image quality.
- Future validation: Testing remains focused on healthy subjects, while further work includes pathology validation, improved RF coils, expanded field of view, and diffusion-weighted imaging assessment.Preliminary work suggests diffusion-weighted imaging may be possible with the unconventional scanner architecture.
Outlook
The scanner architecture is presented as a foundation for clinically validating affordable point-of-care MRI devices. Its low-field, self-shielded, low-power design could expand MRI access beyond conventional scanner settings.
- Future applications: The architecture could support affordable point-of-care detection, assessment, and monitoring across diverse medical applications.Potential extensions include minimally modified extremity and neonatal imaging.
- Design outlook: The compact permanent-magnet design uses built-in readout encoding and efficient phase-encoding gradients without cryogenics, enabling mobility and point-of-care operation.The conclusion links modest size, self-shielding, low power consumption, and low acoustic noise to expanded MRI access.
Methods
The scanner combines a genetically optimized, shimmed permanent-magnet design with specialized gradient, RF, acquisition, and reconstruction methods for portable 3D brain MRI. Frequency-swept RF pulses and tailored encoding support imaging across the system’s wide inhomogeneous-field bandwidth.
- Permanent magnet construction: A genetic algorithm selected NdFeB magnet placements and plastic spacers under constraints on mean B0, monotonic encoding, and total field range.The design uses 888 potential locations and contains 342 N42 and 299 N52 NdFeB cubes.
- Permanent magnet construction: Target-field shimming refined the built-in encoding field after initial nonlinearities caused severe image distortion and aliasing.The refinement optimized shim magnets distributed across 48 shim trays.
- Gradient coils: The scanner used switchable gradient coils for y- and z-direction phase encoding, with coil geometry and winding patterns optimized for the imaging region.The coils were designed for a target gradient efficiency of 0.7 mT/m/A and resistance below 2 ohms.
- RF and supporting hardware: A close-fitting spiral helmet coil provided compact transmit-receive operation, while additional hardware included a passive switch, amplifiers, console, and gradient amplifiers.The RF coil’s inner dimensions were 21 cm anterior-posterior by 17 cm medial-lateral.
- Acquisition and imaging sequences: Because the 20 cm region spans an approximately 80 kHz Larmor bandwidth, the sequence used frequency-swept WURST pulses instead of hard pulses for excitation and refocusing.Simulations show hard-pulse profiles spanning only 1-4 kHz, whereas WURST pulses use 100 kHz sweeps and provide wider bandwidth coverage.
- Acquisition and imaging sequences: Three-dimensional RARE sequences supported T2, inversion-recovery T2, T1, proton-density, and diffusion contrasts without slice-selective gradients.The y phase-encoding gradient partitioned the 3D data into approximately 7 mm image partitions, and in vivo images used 2.2 x 1.3 x 6.8 mm^3 resolution.