Introduction

Hyperpolarized ¹²⁹Xe dissolved-phase MRI allows pulmonary gas-transfer to be quantified, permitting lung regions with impaired gas exchange to be identified in pulmonary diseases[1, 2]. Within a ~16s breath hold, the imaging sequence provides gas-phase images showing regional lung ventilation, barrier-uptake images that quantify the amount of xenon dissolved in interstitial lung tissue and blood plasma, and RBC-transfer images that depict the amount of Xe dissolved in the red-blood cells. However, the barrier-uptake and RBC-transfer images can be contaminated with highly-off-resonant (>200 ppm) gas-phase signal from imperfectly selective excitation, hindering accurate quantification. The previous contamination removal method required a second echo where the dissolved Xe signals are no longer present, limiting its usage[3]. We propose an effective method to remove gas-phase contamination using only data collected as part of the more common, single-echo gas-transfer sequence.

Methods

Two phantoms and one healthy subject were imaged according to a protocol approved by our local Institutional Review Board (with FDA IND 123,577). ¹²⁹Xe (83% enriched) was polarized to ~30% via a ¹²⁹Xe hyperpolarizer (Polarean 9820). Phantoms consisted of 300mL Xe with 700mL of N₂, while the subject inhaled 1L Xe. ¹²⁹Xe gas-transfer MRI was acquired using a 3.0T Philips Achieva scanner and custom-built ¹²⁹Xe dual-loop coil[4].
¹²⁹Xe gas-transfer MRI was implemented similarly to the more common 1-point Dixon technique proposed by Wang et al.[2]. Slight modifications were made to illustrate the contamination-removal technique, including a third frequency acquisition measuring the shape and intensity of the contamination, but are optional (Figure 1). Briefly, three interleaved 3D-radial views with identical trajectories but with excitation frequencies corresponding to dissolved-phase (202 ppm, barrier plus RBC), gas-phase, and off-resonance gas contamination (-202 ppm) were acquired. Spectra were obtained for each imaging frequency before and after the imaging. Imaging parameters were: FOV = 325×325×325 mm³, matrix = 56³, TR = 5ms (effectively 15ms between same-frequency pulses), dwell time = 19.1µs, 950 radial projections, and 58 readout points. Spectral parameters were: dwell time = 19.1µs, and 789 points.
Images were reconstructed in MATLAB using open-source reconstruction software[5] utilizing iterative density compensation[6] and a kernel sharpness value (σ) of 0.14 and extent of 9σ. Removal of the gas-phase contamination uses the gas-phase k-space data, s_g(k), as the initial estimate of contamination and requires four steps (Figure 2): 1) first-order phase correction, 2) zeroth-order phase correction, 3) intensity scaling and 4) subtraction of contamination.
First-order phase correction used the spectra acquired at the end of the breath hold to calculate the frequency difference, (Δf), between the gas-phase signal and the dissolved-phase receiver frequency and modulating the data according to s_g(k)e^2πiΔft, where t is the dwell time multiplied by n-1, where n is the number of the k-space point on the FID.
Zeroth-order phase correction accounted for the accumulation of phase before acquisition and was determined by the phase difference (Δφ) of the spectroscopic gas peak and the mean phase of the k₀ points. This fixed phase difference can be accounted for similarly to the first-order phase correction via s_g(k)e^iΔφ.
Scaling of the k-space intensity was determined from spectral and imaging data. Xe signal intensity in the gas-phase (I_gas) imaging data was determined from the intensity of the final k₀ point. The dissolved-phase spectrum was used to determine the intensity of the barrier, RBC, and gas components, the last of which was the gas contamination intensity in the dissolved-phase (I_{dissolved,gas}) imaging data. Thus, the scale factor (β) can be determined by β=I_{dissolved,gas}/I_gas.
Subtraction of the contamination occurred in k-space and modified the gas-phase k-space data according to these equations. Comparisons between pre- and post-removal were completed by generating a lung mask from the ventilation images and noise masks containing all voxels > 7 voxels away from the lung mask. For the phantom, which doesn’t contain dissolved Xe signal, all voxels were taken as noise in the dissolved-phase images.

Results

In phantoms (Figure 3), mean noise and standard deviation decreased by 74% and 83%, respectively. The skew of the noise decrease from 1.65 to 0.79 (expected 0.63 for the Rayleigh distribution of MRI noise[7]) while kurtosis decreased from 6.9 to 3.7 (expected 3.3). Since no dissolved components were present, the remaining gas contamination was 6.4%, a 93.6% efficiency for both phantoms. For the healthy subject (Figure 4), mean noise and standard deviation decreased by 5% and 10%, respectively. The skew of the noise decrease from 2.3 to 2.1 while kurtosis decreased from 13.3 to 11.8. Contamination removal resulted in an 8% SNR increase.

Discussion

Phantom results illustrate >90% effectiveness of removing the contamination using the commonly acquired spectral and imaging data. Human subject results show increased SNR with a smaller reduction in noise, likely due to the blurring of the dissolved-phase image intensity, physiological noise, and the minor levels and disperse nature of the contamination at 3.0T.

Conclusion

Accurate and effective gas-phase contamination removal from dissolved ¹²⁹Xe gas-transfer images is possible without sequence modifications, permitting application to previously-acquired and future data. Contamination removal allows more accurate quantification of gas exchange and more consistent/reliable results which could permit multi-site investigations implementing ¹²⁹Xe gas-transfer imaging.

Acknowledgements

The authors thank the following sources for research funding and support: NIH R01 HL131012, NIH R01 HL143011, NIH R44 HL123299, NIH K99 HL138255, and NIH T32 HL007752.