Synopsis

Diffusion-weighted MRI is widely used for diagnosing brain pathology, but is typically not available close to the site of acute brain injury due to size, weight, power, and exclusion-zone constraints. Recently-developed lightweight brain scanners have achieved portability by using inhomogeneous B₀ fields and by correcting for field imperfections during image reconstruction. Building on work done by the petroleum industry with single-sided borehole magnets, we use the built-in spatial encoding field of a portable brain scanner for both diffusion-encoding and spatial-encoding. We also introduce a method for using dummy refocusing pulses to vary the b-value without changing TE_eff and T₂-contrast.

Introduction

Diffusion-weighted MRI (DWI) is an indispensable tool for diagnosing and staging a wide range of brain pathology including cerebral infarcts [1], infections, edema and trauma. Extending this capability closer to the site of acute injury, such as ambulances and remote treatment facilities, could enable faster intervention and improved care. However, conventional MRI scanners are poorly-suited to this application due to size, weight, power, and exclusion-zone constraints. One way to improve portability is to allow the main magnetic field to be inhomogeneous [2, 3], but this raises challenges for performing diffusion-weighted imaging. The petroleum-prospecting industry has previously shown diffusion profiling of rock porosity and fluid content using the “always-on” field gradients surrounding single-sided magnets that are lowered down bore-holes [4].

Building on this work, we show proof-of-concept diffusion-weighted phantom images acquired in a ~100kg brain scanner prototype based on an 80mT permanent magnet Halbach-array [5]. The magnet achieves light-weight and portability by allowing the main magnetic field to be highly-inhomogeneous (ΔB₀/B₀~1%) and by accounting for the field variations during generalized image reconstruction. Because the magnet does not require exclusion zones, cryogens, or industrial power, it is well-suited to acute brain injury diagnosis in settings such as an ambulance. The magnet was designed to have a built-in readout-encoding field with approximately linear variation across the FOV (~7.5mT/m) [5]. Using a modified CPMG pulse sequence, we show that this built-in field can also be used for diffusion-encoding with clinically-relevant b-values. We also show a new method for using dummy refocusing pulses during the diffusion-encoding module to vary the b-value without changing TE_eff or T₂-contrast.

Methods

Figure 1 shows the magnet and gradient coil hardware used for DWI experiments. Previously-shown 2D and 3D T₂-weighted images are shown for reference [6]. Figure 2 shows the modified-CPMG spin-echo sequence that uses the built-in ΔB₀-field for readout and the G_z gradient coil for phase encoding. The readout-field nonlinearity is small enough over the FOV (7cm) to permit simple 2D-FFT image reconstruction.

Frequency-swept RF pulses are used to excite and refocus spins across the wide bandwidth of the inhomogeneous B₀field (~100 KHz) [7]. During the diffusion-encoding module, dummy refocusing pulses are added to change the encoding time, τ, providing different b-values according to Equation 1. For b=0, pulses are played every ΔTE as in a conventional CPMG sequence. To keep T₂-contrast the same across all b-value images, only the echoes during the spatial-encoding module are acquired. Optionally, the switched encoding gradients can provide additional diffusion-weighting according to Equation 2. Figure 3 shows the achievable b-values as a function of TE_eff.

(Eq. 1) $$$b_{built-in}=\frac{1}{12} \gamma^2G_y^2\tau^3$$$

(Eq. 2) $$$b_{switched}= \gamma^2G^2\delta^2(\Delta-\delta/3)$$$

Two petri-dish gel phantoms were prepared. Agarose and nickel-sulfate concentrations were chosen to mimic T₁ & T₂ of white-matter at 80mT. The parameters were then measured on a 60MHz Bruker-Minispec relaxometer. Sucrose was added to one phantom to model the Apparent Diffusion Coefficient (ADC) of infarcted white-matter (ADC~0.8e-3 mm²/s) [8]. The ADC of the second phantom was close to pure-water (~2e-3 mm²/s). ADC maps were acquired using a diffusion-weighted spin-echo EPI sequence on a clinical 3T scanner. Experimental diffusion-weighted images were then acquired on the portable MRI scanner using a range of b-values (0-245 s/mm²) while holding TE_eff constant at 100ms.

Results

Figure 4 shows the petri-dish phantoms along with a short-TE scout image and four diffusion-weighted images. Nonlinearity of the built-in encoding field caused some geometric distortion in the 0%-sucrose phantom. As expected, DWI images show greater signal loss in the 0%-sucrose phantom. Figure 4D shows image intensity versus b-value, providing a good monoexponential fit to Equation 3. Figure 5 compares ADC maps acquired on the portable scanner and 3T-clinical scanner and summarizes best-fit values for ADC, T₁, and T₂ of each gel.

(Eq. 3) $$$s=s_0 exp(-bD)$$$

Discussion

Strong diffusion (ADC) contrast is observed between the two phantoms. The closeness of the measured T₂-values for the two gels indicates that the ADC-maps are dominated by diffusion-contrast rather than T₂-shinethrough. The ADC-maps for both scanners are generally consistent, though the portable scanner does overestimate ADC in the 0%-sucrose phantom.

We show 2D diffusion-weighted phantom images in a highly-inhomogeneous B₀ field as a first step toward using a portable brain scanner for diagnosing brain injury. We also show that dummy refocusing pulses can be used to change the b-value applied by the built-in encoding field without changing TE_eff.

In future work, the switched gradient coils will also be used for diffusion-encoding, enabling higher b-values (500-1000 s/mm_²). Also, by varying the relative amplitudes of the two gradients, diffusion-encoding will be possible along all three spatial directions. The method will then be extended to 3D acquisitions for testing on humans and eventual application to brain injury.

Acknowledgements

References

[1] Wu O, Diffusion MRI (Ch. 31), 2010. [2] Cooley CZ, MRM, 2015. [3] Vaughan T, ISMRM, 2016, p. 498. [4] Casanova F, Single-sided NMR, 2011. [5] Cooley CZ, IEEE Trans Mag, 2017. [6] McDaniel P, ISMRM 2018, p.29. [7] Casabianca L, JMR, 2014. [8] Lavdas I, JMRI, 2013.