Recently, efforts have intensified to exploit the chemical shifts and solubility of ¹²⁹Xe to characterize pulmonary function and pathology¹. Because diseases such as idiopathic pulmonary fibrosis (IPF) restrict the diffusion of gases between the airspaces, interstitial tissue, and capillary beds, identifying regional variability among these compartments could provide urgently needed metrics to detect and diagnose disease, as well as to assess progression or therapy response^2,3. To date, all of this work has assumed that the dissolved-phase signal consists of two resonances – one at ~198 ppm that represents ¹²⁹Xe in the barrier tissue and plasma, and one at ~217 ppm that represents ¹²⁹Xe transiently bound to hemoglobin molecules in red blood cells (RBCs)³. However, here we demonstrate with non-linear curve fitting of the complex ¹²⁹Xe FIDs that the dissolved phase consists of 3 resonances. Furthermore, we use a more robust gas-phase reference⁴ to report the frequencies and linewidths of these resonances from a cohort of healthy normal subjects.

Spectra from 7 healthy normal subjects (38±16.9 years old) were acquired on a 1.5T GE Healthcare 15M4 EXCITE MRI scanner with a Quadrature ¹²⁹Xe vest coil (Clinical MR Solutions, Brookfield, WI). Each subject inhaled ~200 mL of isotopically enriched ¹²⁹Xe (85%), hyperpolarized to ~20% via spin-exchange optical pumping (Model 9800, Polarean, Inc., Durham, NC). Subjects held their breath for ~15 seconds while 200 FIDs were acquired with 512 samples/FID, TE/TR = 0.875/20 ms, BW = 8.06 kHz, 1.2 ms 3-lobe sinc pulse, flip angle ≈ 18°. The first 100 frames were discarded to avoid measuring magnetization that had accumulated downstream of the capillary beds. The remaining 100 frames were averaged and then decomposed into a series of exponentially decaying FIDs. Each component signal (denoted by $$$n$$$) was described by four parameters: an amplitude ($$$a_n$$$), starting phase ($$$\phi_n$$$), resonant frequency ($$$f_n$$$), and linewidth ($$$w_n$$$), then summed according to

$$s_{fit}\left(t\right)=\sum_{n=0}^{n_{components}} {a_ne^{i{\phi}_n}e^{2{\pi}if_nt}e^{-{\pi}w_nt}}$$

We used the trust-region-reflect algorithm⁵ to minimize the complex least-squares residual error between the measured time domain FID and fitted result. Our open-source MATLAB toolkit can be downloaded from www.civm.duhs.duke.edu/NmrSpectroscopy for academic and research use. Resonant frequencies were reported in ppm units using the alveolar component of the gas-phase resonance. This resonance was determined to arise at -2.90 ppm relative to true zero when corrected for Xe-O_2, Xe -Xe shifts, as well bulk magnetic susceptibility shifts^6,7,8.

Robust spectroscopic fitting was achieved across all 7 subjects. Because the fitting used the complex signal and was performed in the time domain, no linebroadening, zeropadding, or constraints were required in any subject. Figure 1 illustrates a representative spectrum and fits using 2 (left) and 3 (right) dissolved-phase peaks. By including an additional dissolved-phase peak, the residual error was reduced to the level of the noise and became unpatterned. Table 1 provides the mean±stdev of the frequencies and linewidths for each dissolved-phase peak. Complex spectral curve fitting makes clear that the dissolved phase of ¹²⁹Xe in the human lung consists of three stable Lorentzian-shaped resonances. Including the additional peak only minimally affects the frequency and linewidth of the RBC peak; however the two “barrier” peaks are shifted from and broader than the previous single barrier peak. We hypothesize that these two “barrier” peaks are affected by inflammatory and fibrotic process, and could serve as useful indicators of disease progression in diseases such as IPF. Finally, we hope to account for the 3^rd dissolved-phase peak in future imaging techniques in order to spatially localize and identify their source.