Introduction

Multinuclear MRI scan protocols have been developed to probe the structural and functional properties of the human lung^1-3. ¹⁹F-MRI of inhaled perfluoropropane (PFP) can report on lung ventilation properties without the need for gas tracer hyperpolarisation⁴. Our studies to date have employed a ¹⁹F transmit/receive single-channel coil, employing fully sampled and accelerated gradient echo scan protocols with compressed sensing (CS)^{5, 6}. Combined compressed sensing-parallel imaging (CS-PI)⁷ with multi-channel coils offers additional scan acceleration that can enable higher spatial and/or temporal resolution. We performed a systematic evaluation of the impact of CS-PI scan acceleration on image quality using single- and multi-channel ¹H RF coils and an aqueous ¹H test object that replicated the spin density and relaxation properties of inhaled PFP observed by ¹⁹F-MRI at 3.0T (T₁= 12.4 ms, T₂*= 1.8 ms⁸), enabling assessment of lung ¹⁹F-MRI CS-PI image quality prior to purchase/construction of ¹⁹F-MRI receive array hardware. We assessed the performance of gradient echo scan protocols with a scan duration of no more than 15s (ie. achievable in a breathhold for patients with respiratory disease⁹) in their ability to accurately identify lung ventilation defects of three sizes, with CS/CS-PI acceleration factors of up to 4x, using three ¹H RF receiver arrays (single-channel body coil, 6-channel torso and 16-channel torso coils).

Methods

Test object construction
Two containers of H₂O-doped D₂O solutions (98.75 mM), 2kg and 1.5kg, were constructed containing GdCl₃ (6.55 mM) and MnCl₂ (1.72 mM) to mimic the relaxation properties (T₁= 12.4 ms, T₂*= 1.8 ms)⁸ and spin density (197.5 mM)¹⁰ of a 79%/21% inhaled PFP/oxygen gas mixture observed in ¹⁹F-MRI. Concentrations of Gd³⁺ and Mn²⁺ were derived by simulating the T₁ and T₂* for an array of contrast agent concentration combinations using Equation 1&2¹¹, where CA is concentration of contrast agent, T_1o and T_2o* denote relaxation rate without contrast agent and r₁ and r₂* are molar relaxivities of each contrast agent. By finding the minima of resultant rates against our desired relaxation rates (Equation 3), required concentrations of Gd³⁺ and Mn²⁺ were determined. Two 2L bottles containing acrylic rods with diameters of 15.0±0.3 mm, 10.0±0.3 mm, and 5.0±0.3 mm were employed to generate signal voids of known sizes within the coronal imaging plane (Figure 1). Deviation of measured rod diameter from ground truth was used to compare scan protocols.
(1)$$ \frac{1}{T_1} = \frac{1}{{T_1}_o}+r_1\left[CA_{{Gd}^{3+}}\right]+r_1\left[CA_{{Mn}^{2+}}\right]$$ (2)$$\frac{1}{T_2^*} = \frac{1}{{T_2}_o^*}+r_2^*\left[CA_{{Gd}^{3+}}\right]+r_2^*\left[CA_{{Mn}^{2+}}\right]$$(3)$$min\left\{\parallel \overrightarrow{T_1}\left(\left[Gd^{3+}\right],\left[Mn^{2+}\right] \right)-12.4\parallel+\parallel \overrightarrow{T_2^*}\left(\left[Gd^{3+}\right],\left[Mn^{2+}\right] \right)-1.8\parallel\right\}$$
MRI data acquisition
Coronal 3D ¹H spoiled gradient echo sequences were acquired with field of view (FOV)= 400x330x250 mm³, TR= 7.5 ms, TE= 1.7 ms, flip angle= 50°, and bandwidth= 500 Hz/pixel on a Philips Achieva 3T scanner. Scans were repeated with varying spatial resolution (range: 4x4x4 mm³‑6.25x10x10 mm³). Acquisition time was held approximately constant by varying the NSA used in retrospective CS/CS-PI image reconstruction. Table 1A shows matrix size, resolution, and scan times for an NSA=1 scan at each resolution. Table 1B shows the NSA employed at each resolution and retrospective CS/CS-PI acceleration factor to achieve a scan duration of 15s. This yielded 28 unique scan protocols per RF receiver array (single-channel body coil, 6-channel torso coil, and 16-channel torso coil (Philips)). CS/CS-PI reconstruction was as previously described, extended to CS-PI using ESPIRiT^6,7. Scans were zero-filled in MATLAB by a factor of 3 in the in-plane directions.
MRI data analysis
Reconstructed images were segmented using an in-house threshold based semi-automated approach used in our patient ¹⁹F-MRI lung ventilation imaging studies⁹. We measured the diameter of signal voids from a central slice of the binary segmented image in L-R and H-F directions. This was repeated at n=8 known positions for each diameter (Figure 2). Diameter deviation from the ground truth was calculated for each rod size and plotted against acceleration for all resolutions/RF receiver arrays.

Results

Gd³⁺ and Mn²⁺ concentrations of 6.55 mM and 1.72 mM achieve ¹H T₁ and T₂* of 12.4 ms and 1.8 ms respectively (Figure 3). Average absolute diameter deviation across acceleration was largest in 5 mm rod diameters for all scans. We expected this as 5 mm was chosen to test scan protocol capability limitation. Correlation coefficients for 5 mm diameter deviations against acceleration factor were: negative (range: -0.9 to -0.6) for single-channel CS reconstructions; mixed (range: -0.1 to 0.9) for 6-channel CS-PI reconstructions and positive (range: 0.3 to 0.9) for 16-channel CS-PI reconstructions (Figure 4). This suggests improved image fidelity for increased acceleration using CS-PI and decreased image fidelity using CS.

Discussion and conclusion

Aqueous ¹H test objects have been constructed with independent control of spin density, T₁ and T₂*, enabling replication of inhaled PFP properties as observed by ¹⁹F-MRI. We have tested ¹H-MRI scan protocols using conventional MRI hardware for sequence optimisation that can be applied to patient lung ¹⁹F-MRI studies. Higher acquisition resolutions achieved the lowest absolute error of simulated ventilation defect diameter. A negative impact of acceleration on all measured diameter was observed for CS-accelerated scans. However, no systematic impact of increasing acceleration was observed for multi-channel CS-PI reconstructions. Our data illustrates the value of CS-PI for improvement of ¹⁹F-MRI lung scan spatial resolution without impact on scan duration. We plan to integrate optimised CS-PI accelerated scan protocols into studies of patients with respiratory disease.

Acknowledgements

This project is funded by the UK Engineering and Physical Sciences Research Council’s Centre for Doctoral Training in Molecular Sciences for Medicine. The authors acknowledge assistance from Matthew Clemence (Philips Clinical Science), and the radiography team at the Newcastle Magnetic Resonance Centre, for help with data acquisition.

References

1. Shukla, Y., et al., Hyperpolarized 129Xe Magnetic Resonance Imaging. Academic Radiology, 2012. 19(8): p. 941-951.

2. Van Beek, E.J.R., et al., Functional MRI of the lung using hyperpolarized 3-helium gas. Journal of Magnetic Resonance Imaging, 2004. 20(4): p. 540-554.

3. Schreiber, W.G., et al., Dynamic (19)F-MRI of pulmonary ventilation using sulfur hexafluoride (SF(6)) gas. Magn Reson Med, 2001. 45(4): p. 605-13.

4. Couch, M.J., et al., (19) F MRI of the Lungs Using Inert Fluorinated Gases: Challenges and New Developments. J Magn Reson Imaging, 2019. 49(2): p. 343-354.

5. Lustig, M., D. Donoho, and J.M. Pauly, Sparse MRI: The application of compressed sensing for rapid MR imaging. Magn Reson Med, 2007. 58(6): p. 1182-95.

6. Neal, M.A., et al., Optimized and accelerated 19 F‐MRI of inhaled perfluoropropane to assess regional pulmonary ventilation. Magnetic Resonance in Medicine, 2019. 82(4): p. 1301-1311.

7. Uecker, M., et al., ESPIRiT--an eigenvalue approach to autocalibrating parallel MRI: where SENSE meets GRAPPA. Magn Reson Med, 2014. 71(3): p. 990-1001.

8. Couch, M.J., et al., Pulmonary ultrashort echo time 19F MR imaging with inhaled fluorinated gas mixtures in healthy volunteers: feasibility. Radiology, 2013. 269(3): p. 903-9.

9. Pippard, B.J., et al., Reproducibility of 19 F‐MR ventilation imaging in healthy volunteers. Magnetic Resonance in Medicine, 2021. 85(6): p. 3343-3352.

10. Denison, D.M., M.D. Morgan, and A.B. Millar, Estimation of regional gas and tissue volumes of the lung in supine man using computed tomography. Thorax, 1986. 41(8): p. 620-628.

11. Nofiele, J.T. and H.-L.M. Cheng, Ultrashort Echo Time for Improved Positive-Contrast Manganese-Enhanced MRI of Cancer. PLoS ONE, 2013. 8(3): p. e58617.