Purpose

¹²⁹Xe MR spectroscopy provides a direct measurement of the pulmonary and pulmonary capillary environment. As ¹²⁹Xe diffuses from airspaces into the capillaries it exhibits distinct chemical shifts corresponding to three gas exchange compartments: gas, barrier tissue, and red blood cells (RBCs). The RBC:barrier ratio is a clear marker of gas exchange efficiency, while the RBC chemical shift appears sensitive to capillary blood oxygenation. Additionally, variations of the RBC spectral parameters with the cardiac cycle have demonstrated the ability to differentiate between pre- and postcapillary pulmonary hypertension (PH)[1]. Thus, an increasing array of static and dynamic ¹²⁹Xe spectroscopic parameters have the potential to play an integral role in the detection and characterization of various cardiopulmonary diseases. However, use of these parameters for clinical decision making and monitoring requires developing a better understanding of their repeatability. In this study we address this gap by reporting on the intra-session repeatability ¹²⁹Xe static and dynamic spectroscopic parameters.

Methods

Paired sets of ¹²⁹Xe MRS scans were acquired on total of 155 subjects with a range of cardiopulmonary disorders. Subjects underwent two separate ¹²⁹Xe dynamic spectroscopy scans during a single ¹²⁹Xe study. From FRC subjects inhaled a 500-1000mL volume containing ~70 mL dose equivalent of hyperpolarized ¹²⁹Xe and held their breath. During this breath hold, ¹²⁹Xe free induction decays (FIDs) were acquired every 15 or 20 ms at the dissolved-phase ¹²⁹Xe frequency (TE=0.45ms, flip angle≈20°, dwell time=20µs, 512 points, 500-800 FIDs).

The first 100 FIDs were discarded to ensure steady-state ¹²⁹Xe magnetization was confined to the pulmonary gas exchange regions. FIDs 100-150 were averaged together create a high SNR static spectrum before time domain fitting to a model with 1 Voigt lineshape (barrier) and 2 Lorentzian (RBC and gas) peaks. The static spectroscopy parameters quantified were RBC:barrier ratio, RBC chemical shift, RBC spectral width (FWHM), RBC phase (°, relative to barrier), barrier chemical shift, and 2 barrier spectral width parameters (FWHM_L and FWHM_G). For dynamic spectroscopy, all FIDs after the first 100 were fit in the time domain to the barrier Voigt model. Dynamic oscillations in the RBC signal amplitude, chemical shift, and phase were quantified by their peak-to-peak amplitudes after detrending[2].

Scans were excluded from repeatability analysis if the absolute difference in heart rate (HR) between scans was above 15 bpm (19 subjects excluded), if either static RBC SNR was below 6 (14 additional subjects excluded), or if either RBC amplitude oscillation R-square (goodness-of-fit parameter) was below 0.2 (5 additional subjects excluded). All statistical analysis was performed using the 117 subjects in the filtered data set. Repeatability was visualized using Bland-Altman plots. Intra-scan repeatability was quantified using the, coefficient of repeatability (CR), coefficient of variation (CV), and the intraclass correlation coefficient (ICC), defined as follows: $${CR= 1.96\times SD(scan_1-scan_2)}$$ $${CV=\frac{SD(scan_1-scan_2)}{mean(scan_1-scan_2)}}$$ $${ICC= \frac{MS_{subj}-MS_{error}}{MS_{subj} + MS_{error}} \approx \frac{ (SD\_subject's\_ true\_values)^2}{(SD\_subject's\_true\_values)^2 + (SD\_measurement\_error)^2} }$$ where MS is the mean square from an analysis of variance (ANOVA).

Results

The exclusion process is visualized in Figure 2, which depicts the relationship between the exclusion criteria and the repeatability of the RBC amplitude oscillations and RBC:barrier for all subjects. Figure 3 provides a table of all repeatability metrics calculated from this study. The repeatability of the static spectroscopic measurements is shown in Figure 4.

None of the measurements had a significant bias indicating that there is no difference in spectroscopic measurements taken at the beginning or end of MRS session. In the static parameters, RBC:barrier ratio had the highest ICC of 0.97 with CR=0.7.

The Bland-Altman plots of the RBC dynamic parameters are shown in Figure 5. Again, none of the measurements had significant bias. The RBC amplitude oscillations had CR=4.2% was more reproducible than either the chemical shift or phase oscillations, showing ICCs of 0.88, 0.64, and 0.73, respectively.

Discussion

These results indicate that ¹²⁹Xe static spectroscopy is very repeatable with all ICCs, except barrier chemical shift, at or above 0.90, and that dynamic spectroscopy is moderately repeatable. The higher repeatability in static spectroscopy is expected because the time-dimension is removed during FID averaging which creates a much higher SNR data set. This suggests that even if a subject does not have adequate SNR for a repeatable dynamic scan, there may be enough signal to reliably calculate static spectral parameters.

The repeatability of these measurements can be greatly affected by a change in in physiology. To reduce variation due to change in cardiac output, which could change RBC amplitude oscillation, we limited analysis to subjects that had similar HRs across both scans. However, there is a known dependence of RBC:barrier ratio on lung inflation level[3] and unfortunately, spectroscopy cannot measure lung inflation. In the future it may be possible to use ¹²⁹Xe oscillation imaging[4] to correct the global RBC amplitude oscillations for a change in lung inflation volume. Another potential source of physiological variation could be subjects inadvertently performing either a Muller or Valsalva maneuver and introducing a change in intrapleural pressures that could affect both static and dynamic spectral parameters.

Ultimately, this repeatability study for ¹²⁹Xe MRS measurements helps quantify meaningful variations in spectroscopy metrics and is an important step forward in moving ¹²⁹Xe MRS into the clinical setting.

Acknowledgements

2R01HL105643-06, R01HL126771, R01HL126771, Genentech