3196

Establishing a Standardized Healthy Reference Distribution for Multi-Site 129Xe Gas Exchange MRI Across Major Scanner Platforms
Suphachart Leewiwatwong1, Aryil Bechtel2, David Mummy2, Shuo Zhang2, Junlan Lu3, Zackary Cleveland4, Matthew Willmering4, Juan Parra-Robles4, Sean Fain5, Andrew D Hahn5, and Bastiaan Driehuys1,2,3
1Biomedical Engineering, Duke University, DURHAM, NC, United States, 2Radiology, Duke University, DURHAM, NC, United States, 3Medical Physics Graduate Program, Duke University, DURHAM, NC, United States, 4Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States, 5Radiology, University of Iowa, Iowa City, IA, United States

Synopsis

Keywords: Hyperpolarized MR (Gas), Hyperpolarized MR (Gas)

Motivation: Quantitative 129Xe gas exchange MRI, conducted across different imaging centers and scanner platforms, requires consistent healthy reference distributions.

Goal(s): To establish standardized reference values from an 18-30yr old multicenter healthy cohort for 129Xe gas exchange MRI.

Approach: Participants from three research centers underwent pulmonary function tests and a standardized 129Xe MRI/MRS protocol. Data were processed centrally and corrected for T2*.

Results: A balanced multicenter dataset revealed minimal variability between combined and site-specific reference distributions, validating the combined values for cross-center use. The distribution for 208-ppm excitation could be reliably transformed for 218-ppm excitation.

Impact: This study provides robust cross-platform reference distributions for 129Xe gas exchange MRI, facilitating comparison of quantitative imaging in multi-center respiratory research.

INTRODUCTION

Establishing robust healthy reference values is crucial for the accurate quantitative 129Xe gas exchange (GX) MRI[1], as xenon enables the distinct visualization of each gas exchange compartment—alveolar gas phase, membrane, and red blood cells (RBC). Currently available reference values have been limited, generated from small, center-specific populations, and are MR platform-specific. This has introduced uncertainty and the potential for bias when comparing patient images across centers. This work seeks to establish standardized values that are derived from three geographically distinct imaging centers and encompass all three major MRI scanner platforms.

METHODS

Three separate centers, Duke University[1] (Siemens), Cincinnati Children’s Hospital Medical Center[2] (Philips), and the University of Iowa[3] (General Electric), recruited non-smoking, healthy participants aged 18-30. Participants underwent pulmonary function tests (PFTs) and 129Xe MRS and 1-point Dixon GX MRI protocol based on the 129Xe MRI clinical trials consortium recommendations[4], but modified to reduce scan time to 10s (TR=8.5ms, flip=15˚) and with RF dissolved-phase excitation at 208 ppm, centered between the membrane and red blood cell (RBC) resonances. Raw MR data were converted to an ISMRMRD[5] format adapted for 129Xe MRI/MRS and processed with a single MRI/MRS analysis pipeline. Subjects were excluded if they had abnormal PFTs, image artifacts, RBC SNR<5, or RBC spectral peaks with FWHM >310 Hz[6, 7]. Images were corrected for T2*, which for the gas phase was approximated as 18ms[8] and for the dissolved-phase was the average T2* estimated from fits to the membrane and RBC spectral peaks for all subjects. All T2* values were adjusted to a B0 of 3T [9]. Data from the three centers were utilized to construct both site-specific and combined reference distributions and color-binning thresholds for each gas exchange compartment, as done previously[10]. To estimate the differences between using a site’s own reference distribution vs the combined one, the voxel percentage within one standard deviation (std) of the reference mean was assessed using both approaches[11] and compared using a Wilcoxon signed-rank test[12]. Cohen's d[13] was employed to determine the effect size for these Box-cox transformed[14] comparisons. Cohen's d values are typically interpreted as small (0.2), medium (0.5), and large (0.8) effect sizes. Moreover, we sought to assess how reference distributions acquired here at 208 ppm compare to those acquired at the consortium recommendation of 218 ppm. To do this, the multi-site combined 208-ppm distributions established here were rescaled by accounting for the frequency-dependent excitation at 208 vs 218 ppm; they were then compared with previously established 218-ppm distributions using Cohen’s d. To derive scaling factors, a Hanning-windowed 0.69 ms sinc RF pulse was simulated at both frequencies using the Bloch equations by applying a series of rotations about the time-varying B1 vector. The transverse magnetization (Mxy) distribution across frequencies was visualized to depict the signal strength at the membrane and RBC resonances, which was then utilized to calculate the 208-to-218-ppm scaling factors.

RESULTS

The curated dataset from the three centers consisted of Duke/Siemens (N=7, 4M/3F), Cincinnati/Philips (N=18, 10M/8F), and Iowa/General Electric (N=7, 3M/4F), and were utilized to construct reference distributions for each gas exchange compartment. Combined and site-specific reference distributions were qualitatively similar (Figure 1). Using both combined and site-specific distributions, the color-binned images from each site demonstrated typical GX transfer for healthy subjects (Figure 2). Absolute Cohen’s d from each GX compartment between combined and site-specific distributions ranged from 0.004 to 0.304, indicating small effect sizes. For the three compartments, the proportion of voxels within one std of the reference mean varied slightly, ranging from 66.8% to 74.6% (Figure 3), in line with the expected 68.2%. When these percentages were recalculated using each site-specific distribution, no significant variations were observed (p = 0.5). The combined 208-ppm distribution, when transformed and compared to a measured 218-ppm distribution (Figure 4) was quite similar (Figure 5), with a maximum Cohen’s d of 0.296 for the ventilation compartment, indicating a small effect size.

DISCUSSION

The small Cohen’s d effect sizes and similar proportion of voxels near the mean when using a site-specific vs combined distribution suggest it can serve as a reliable benchmark for cross-center comparisons. Moreover, the analysis of 208 vs 218-ppm excitation, both theoretical and empirical, suggest that the combined distribution, acquired with 208-ppm excitation, can be reliably used to estimate the reference distribution for the current consortium standard 218-ppm excitation. Transformation of the 208-ppm distribution to the consortium-standard 218 ppm requires only scaling the membrane signal by 0.92 and RBC by 1.03. With continued recruitment of additional healthy subjects, these reference values may be further refined to test for sex differences.

Acknowledgements

Acknowledgements: R01HL105643, R01HL12677, NSF GRFP DGE-2139754

References

[1] Z. Wang et al., "Quantitative analysis of hyperpolarized 129Xe gas transfer MRI," Med Phys, vol. 44, no. 6, pp. 2415-2428, Jun 2017, doi: https://doi.org/10.1002/mp.12264.

[2] M. M. Willmering et al., "Improved pulmonary 129Xe ventilation imaging via 3D-spiral UTE MRI," Magn Reson Med, vol. 84, no. 1, pp. 312-320, Jul 2020, doi: https://doi.org/10.1002/mrm.28114.

[3] A. D. Hahn et al., "Functional xenon-129 magnetic resonance imaging response to antifibrotic treatment in idiopathic pulmonary fibrosis," ERJ Open Res, vol. 9, no. 3, May 2023, doi: 10.1183/23120541.00080-2023.

[4] P. J. Niedbalski et al., "Protocols for multi-site trials using hyperpolarized 129Xe MRI for imaging of ventilation, alveolar-airspace size, and gas exchange: A position paper from the 129Xe MRI clinical trials consortium," Magnetic Resonance in Medicine, vol. 86, no. 6, pp. 2966-2986, 2021, doi: https://doi.org/10.1002/mrm.28985.

[5] S. J. Inati et al., "ISMRM Raw data format: A proposed standard for MRI raw datasets," Magn Reson Med, vol. 77, no. 1, pp. 411-421, Jan 2017, doi: 10.1002/mrm.26089.

[6] A. Rose, "The Sensitivity Performance of the Human Eye on an Absolute Scale*," J. Opt. Soc. Am., vol. 38, no. 2, pp. 196-208, 1948/02/01 1948, doi: 10.1364/JOSA.38.000196.

[7] E. A. Bier et al., "A protocol for quantifying cardiogenic oscillations in dynamic 129Xe gas exchange spectroscopy: The effects of idiopathic pulmonary fibrosis," NMR Biomed, vol. 32, no. 1, p. e4029, Jan 2019, doi: 10.1002/nbm.4029.

[8] X. Xu et al., "Hyperpolarized 129Xe gas lung MRI-SNR and T2* comparisons at 1.5 T and 3 T," Magn Reson Med, vol. 68, no. 6, pp. 1900-4, Dec 2012, doi: 10.1002/mrm.24190.

[9] E. M. Haacke, Magnetic resonance imaging : physical principles and sequence design. J. Wiley & Sons, 1999.

[10] M. H. Ziyi Wang, Rohan Virgincar, Elianna A Bier, Sheng Luo, and Bastiaan Driehuys, "Quantifying Hyperpolarized 129Xe Gas Exchange MRI Across Platforms, Field Strength, and Acquisition Parameters," ISMRM, 2019.

[11] M. He et al., "Generalized Linear Binning to Compare Hyperpolarized 129Xe Ventilation Maps Derived from 3D Radial Gas Exchange Versus Dedicated Multislice Gradient Echo MRI,"Acad Radiol, vol. 27, no. 8, pp. e193-e203, Aug 2020, doi: 10.1016/j.acra.2019.10.016.

[12] D. Rey and M. Neuhäuser, "Wilcoxon-Signed-Rank Test," in International Encyclopedia of Statistical Science, M. Lovric Ed. Berlin, Heidelberg: Springer Berlin Heidelberg, 2011, pp. 1658-1659.

[13] D. Lakens, "Calculating and reporting effect sizes to facilitate cumulative science: a practical primer for t-tests and ANOVAs", Front Psychol, vol. 4, p. 863, Nov 26 2013, doi: 10.3389/fpsyg.2013.00863.

[14] G. E. P. Box and D. R. Cox, "An Analysis of Transformations," Journal of the Royal Statistical Society. Series B (Methodological), vol. 26, no. 2, pp. 211-252, 1964.

Figures

Figure 1. Healthy reference distributions for ventilation membrane-to-gas, and RBC-to-gas categories compartments from each site vs the combined distribution. Note, these reference distributions are for RF excitation between the membrane and RBC resonances at 208 ppm from the gas-phase signal. Mean values for RBC-to-gas and membrane-to-gas at each of the three centers were comparable to the combined reference mean, with differences ranging from 4.80% to 15.19%.

Figure 2. Representative healthy subjects from each of the three sites processed with site-specific vs the combined reference distribution. Pixels in dark and light green are those that fall within one standard deviation (std) below and above the reference mean. The color bar indicates standard color binning, with each separated by one std of the reference distribution. 'H' stands for High, 'L' for Low, and 'D' for Defect. The montage displays a uniform 'healthy green' across subjects regardless of using the site’s own or combined reference distribution.

Figure 3. Boxplots for each gas exchange compartment showing the percentage of voxels within one standard deviation of the reference mean when using a site’s vs combined reference distribution. The inset table shows the mean±std of each boxplot. Green diamonds are mean values; red lines represent medians. The dashed line at 68.2% is the expected percentage within one std of a normal distribution. The transparent red and green lines represents each subject’s decrease/increase when transitioning from the site-specific to combined distribution.

Figure 4. A. 129Xe spectrum showing the RBC and membrane peaks at 218 and 198 ppm, relative to ventilation (0 ppm). B. Real, time-domain profile of the Hanning-windowed 0.69ms sinc RF pulse at 208 vs 218 ppm. C. Tranverse magnetization Mxy simulations showing for 208-ppm excitation, both membrane and RBC are excited with equal magnitudes of 0.97. At 218 ppm, the RBC is fully excited while the membrane is only 0.89. This results in correction factors to transform 208-ppm excitation to 208-ppm of 0.92 for membrane and 1.03 for RBC.

Figure 5. Applying the RF pulse scaling factors to account for 208-ppm excitation, allows the combined 3-site reference distribution to be used to estimate the distribution for the current consortium recommended 218-ppm excitation. This could be compared to the directly measured 218-ppm distribution (n=15, Duke), and showed good agreement. Cohen's d also indicates a small effect size (effect size < 0.5), suggesting that the estimated values are good approximations of the 218-ppm reference distribution.

Proc. Intl. Soc. Mag. Reson. Med. 32 (2024)
3196
DOI: https://doi.org/10.58530/2024/3196