Investigating the impact of RF saturation-pulse parameters on compartment-selective gas-phase depolarization with xenon polarization transfer contrast (XTC) MRI

Abstract

Purpose:

To demonstrate the utility of continuous-wave (CW) saturation pulses in xenon-polarization transfer contrast (XTC) MR imaging and spectroscopy, to investigate the selectivity of CW pulses applied to dissolved-phase resonances, and to develop a correction method for measurement biases from saturation of the non-targeted dissolved-phase compartment.

Methods:

Studies were performed in six healthy Sprague-Dawley rats over a series of end-exhale breath holds. Discrete saturation schemes included a series of 30 Gaussian pulses (8 ms full width at half maximum), spaced 25 ms apart; CW saturation schemes included single block pulses, with variable flip angle and duration. In XTC imaging, saturation pulses were applied on both DP resonance frequencies, as well as off-resonance to correct for other sources of signal loss and compromised selectivity. In spectroscopy experiments, saturation pulses were applied at a set of 19 frequencies spread out between 185 and 200 ppm to map out modified z-spectra.

Results:

Both modified z-spectra and imaging results showed that CW RF pulses offer sufficient depolarization and improved selectivity for generating contrast between pre- and post-saturation acquisitions. A comparison of results obtained using a variety of saturation parameters confirm that saturation pulses applied at higher powers exhibit increased cross-contamination between dissolved-phase resonances.

Conclusion:

Using CW RF saturation pulses in XTC contrast preparation, with the proposed correction method, offers a potentially more selective alternative to traditional discrete saturation. The suppression of the RBC contribution to the GP depolarization opens the door to a novel way of quantifying exchange time between alveolar volume and hemoglobin.

Introduction

Hyperpolarized xenon-129 (HXe) MRI provides a powerful tool for evaluating pulmonary function and structure. Upon inhalation, xenon gas fills the ventilated airspaces, dissolves in the surrounding lung parenchyma, and binds to hemoglobin—with these reservoirs rapidly exchanging with each other. Due to its large electron cloud, xenon-129 is exquisitely sensitive to its chemical environment inside the lungs and exhibits distinct chemical shifts according to the compartment it resides in: at 0 ppm in the gas phase (GP), at ~198 ppm in the lung tissue and blood plasma (TP), and at a species and blood oxygenation-level dependent shift of ~200–225 ppm when bound to hemoglobin inside red blood cells (RBC).^1–7 Numerous imaging techniques have been developed that exploit these unique features of HXe MRI, and have been shown to be sensitive to some of the pathological changes associated with lung diseases such as asthma, chronic obstructive pulmonary disease (COPD), cystic fibrosis, and interstitial lung disease.^8–19

In the case of disease, the xenon exchange between gas volumes and lung tissue is of particular interest, as it reveals physiological parameters that also affect the vital exchange processes for oxygen and carbon dioxide. However, only about 2% of the xenon in the lung is dissolved in the lung parenchyma, and the detection of its MR signal is further hampered by a T₂* of just 1–2 ms at common clinical field strengths.^20,21 Nevertheless, imaging techniques based on simultaneous dissolved phase (DP)-GP imaging^4,22, 1-point Dixon imaging^23–26, chemical shift imaging^27–28, or multi-point Dixon imaging with hierarchical IDEAL (iterative decomposition of water and fat with echo asymmetry and least-squares estimation) reconstruction^{3, 29–32} have been successfully employed to image this DP xenon signal directly. However, all of these measurement techniques, in addition to characterizing xenon gas exchange, are also affected by the pulmonary circulation’s gas transport processes, making the exact nature of the assessed physiologic metric somewhat ill defined.

To exclusively quantify pulmonary xenon gas exchange, the xenon-polarization transfer contrast (XTC) MRI technique takes the reverse approach, using narrow-bandwidth radio-frequency (RF) pulses to selectively saturate or invert the DP xenon magnetization, measuring gas exchange as the resulting decrease in GP signal.^33–39 In current implementations, the manipulation of DP magnetization is always performed via a series of discrete RF pulses separated by a delay time ranging from several milliseconds to a few tens of milliseconds, during which depolarized DP xenon exchanges with the GP. Nevertheless, the impact of the saturation pulse parameters on GP signal depolarization has not been fully explored. Previous in vitro studies using conceptually similar chemical exchange saturation transfer with hyperpolarized xenon (Hyper-CEST) measurements indicate that discrete RF pulses might indeed be superior with regard to energy deposition and RF power limitations albeit less efficient than the alternative of continuous-wave (CW) saturation pulses.^40,41 However, it is unclear to what extent these findings are transferable to in vivo XTC MRI studies of the lung due to the short T2*, the presence of physical transport processes in addition to pure chemical exchange and the resulting mixing of contributions to the total GP depolarization from the two DP compartments. The latter factor, in particular, allows for the intriguing possibility that the dependence of the GP depolarization on the center frequency and power of applied CW saturation pulses could be a function of lung morphometry.

In this work, we investigated the utility of CW saturation pulses for contrast preparation in the XTC MRI technique and compared their compartment-selective saturation properties at different RF powers to those of the traditionally used discrete RF pulses. We also evaluated the feasibility of selectively saturating only the TP or RBC compartments using either discrete or CW-RF pulses and the possibility of correcting for cross contamination from the non-targeted DP resonance.

Methods

Animal Preparation

Experiments were performed in 6 healthy male Sprague-Dawley rats (250–400 g each). Rats were injected intraperitoneally with a mixture of 75–100 mg/kg ketamine and 10 mg/kg xylazine to induce anesthesia. Sedation was maintained during imaging with subsequent doses of ketamine equivalent to half the initial dose. Anesthetized rats were intubated orotracheally using a 14-gauge catheter (Terumo, Tokyo, Japan), sealing the surrounding area with putty (3M Center, St Paul, MN) to prevent gas leakage. Rats then breathed room air until connected to a custom-built, MRI-compatible mechanical ventilator which delivered a mixture of O2 (30%) and N2 (70%) at a breathing rate (BR) of 51 breaths/min, a tidal volume (TV) of 10 ml/kg, and an inspiratory:expiratory ratio of 1/2.45.

All imaging was performed in a horizontal bore 3T system (BioSpec, Bruker Inc, Billerica, MA). Body temperature was monitored using an endorectal probe and regulated by adjusting either the flow rate or water temperature of the heater (ThermoScientific, Waltham, MA). A respiratory pillow was taped onto the body of the rat to detect chest motion and produce a respiratory trace. Peak inspiratory pressure was also monitored using a fiber optic pressure transducer (FISO FPI-LS Series, Harvard Apparatus, Holliston, MA) connected to the ventilator’s exhale port. Animals were euthanized immediately after imaging. All imaging protocols were approved by the Institutional Animal Care and Use Committee at the University of Pennsylvania.

Gas Polarization and Administration

Natural abundance (for spectroscopy and chemical shift imaging (CSI)) and isotopically enriched xenon (87% xenon-129; for XTC imaging) were polarized through collisional spin exchange with optically pumped rubidium vapor in a prototype commercial system (XeBox-E10; Xemed, Durham, NH), which provided polarizations of 40%–50%. HXe gas was dispensed into a Tedlar bag (Jensen Inert Products, Coral Springs, FL) inside a cylinder that was subsequently pressurized and connected to the ventilator.

Prior to the HXe MRI studies, low resolution proton images were acquired to adjust the rotation and horizontal placement of the rat. For HXe measurements, ventilator settings were switched to a 30%/70% O2/Xe mixture. Ventilator pressure was calibrated to prevent significant differences in TV between delivery of Xe/O2 versus N2/O2 gas. Breathing was monitored during Xe administration, with triggering initiated based on continuous acquisition of the inspiratory pressure using the fiber optic transducer. All spectra were acquired over 1.2 s end-expiratory breath-holds regardless of saturation-scheme durations. Two wash-in breaths were implemented in between breath-holds at a rate of 37 bpm. The number of breath-holds was equivalent to the number of repetitions applied in the sequence. For imaging, three wash-in O2/Xe breaths and two wash-out O2/N2 breaths were applied for each breath-hold. Two images were obtained, pre- and post-saturation, over the course of a single breath-hold triggered at end exhale and held for 3.9 s.

MR Imaging and Spectroscopy

The GP depolarization was measured spectroscopically and with XTC MRI for a variety of DP saturation schemes. Spectroscopic measurements consisted of two free induction decay (FID) acquisitions with the excitation centered at the GP resonance (flip angle 15°, 101 ms sampling duration, 5757 sampling points) separated by a saturation segment during which either a single CW RF pulse or a series of 30 discrete Gaussian RF pulses (25 ms apart) with 19 center frequencies between 180 and 225 ppm applied to saturate the DP magnetization (Fig. 1A). The frequency difference Δ between TP and RBC DP resonances is of particular interest for assessing the compartment selectivity of the RF saturation by comparing the induced GP depolarization for the two characteristic on-resonance (f_TP and f_RBC) and off-resonance (Δ_TP and Δ_RBC) frequencies, as illustrated in Fig. 1B. The Gaussian RF saturation pulses had a full-width at half-maximum (FWHM) of 8 ms, a bandwidth (BW) of 137 Hz, and a B_1,rms ranging from 1.5 to 27.5 μT. The CW saturation pulses had total flip angles of either 5,400° or 10,000° and durations of 75, 90, 200, 300, 525, and 750 ms, resulting in applied flip angle rates ranging from 7.2°/ms (1.7 μT) to 133°/ms (31.4 μT).

Figure 1.

(A) Schematic of RF pulse sequences applied during end-exhalation breath-holds for both spectroscopy and XTC imaging. Following n wash-in breaths (n = 2 for spectroscopy and n = 3 for imaging) with Xe/O2 gas mixture, an initial data acquisition is followed by a saturation block and then a second acquisition; this scheme is repeated N times depending on the number of measured saturation frequencies and averages. (B) Characteristic saturation frequencies used for quantifying the compartment specificity of the investigated RF saturation schemes and for correcting biases incurred by the inadvertent saturation of the non-targeted DP compartment: TP resonance frequency (f_TP), RBC resonance frequency (f_RBC), frequency separation between TP and RBC compartments (Δ = f_RBC − f_TP), TP off-resonance frequency (Δ_TP = f_TP - Δ), and RBC off-resonance frequency (Δ_RBC = f_RBC + Δ).

An implementation the XTC MRI technique is structurally equivalent to the spectroscopic measurements (Fig. 1A), consisting of two image acquisitions separated by a saturation segment. 2D coronal multi-slice images were obtained using a 2-slice FLASH sequence with the following image parameters: matrix size 64 × 64; TR/TE = 9.2/1.8 ms; FOV = 60 × 60 mm²; slice thickness 14 mm). The excitation flip angles were 5.7° and 7.8° for the first and second image acquisitions, respectively, to partially compensate for the reduced GP magnetization available for the second measurement⁴⁴. The saturation segment consisted of either the same series of Gaussian RF pulses used for the spectroscopic measurements with B_1,rms of 2.5 or 4.6 μT, or a CW RF saturation pulse with one of the following parameter sets: a) flip angle 5,400°, duration 750ms (flip angle rate 7.2°/ms, B_1,rms 1.7 μT); b) 10,000°, 75 ms (133°/ms, 31.4 μT); c) 10,000°, 750 ms (13.3°/ms, 3.1 μT).

A dynamic CSI measurement with concentric center-out acquisition order was performed in one animal to obtain DP reference data based on xenon gas uptake by the lung parenchyma. Each line of k-space was measured 16 times during the respiratory cycle, requiring 576 breaths to fill a 24 × 24 k-space matrix. Here, only those 576 k-space lines corresponding to an end-inspiration data set were reconstructed, analyzed, and compared to XTC MRI. A Gaussian RF pulse (flip angle 40°, BW 8 kHz) centered at (f_TP + f_RBC)/2 was used for excitation. Other parameters included: coronal projection; TR/TE = 73/1.2 ms; FOV = 40 × 40 mm²; BW = 50 kHz.

Data Analysis

All image reconstruction and data analysis was performed using customized scripts developed in MATLAB (R2019A, MathWorks, Inc., Natick, MA). The sampled spectroscopy data was line-broadened by 100 Hz, Fourier transformed, and phased to first order. GP depolarization (d) was calculated as:

$$d=1-\frac{S_1}{S_0}$$ (1)

with S₀ and S₁ representing the integrated GP peak before and after saturation, respectively. Depolarization ratios were found by dividing the d value for the resonance frequency of interest by the d value obtained from the control saturation frequencies.

Modified z-spectra (hereafter simply referred to as “z-spectra”) that quantify the GP depolarization rate at each saturation center frequency, instead of being normalized to 1 at peak depolarization, were analyzed by first assuming a Lorentzian-shaped fit to the rate of change of the signal⁴⁴.

$$\frac{\dot{S}}{S}\left(f,t\right)=-\left({\Gamma }_{\text{TP}}\left(f\right)+{\Gamma }_{\text{RBC}}\left(f\right)+k\right)$$ (2)

where the measured signal, S, is a function of the saturation frequency, f and time, t. A constant term, k, represents an offset encompassing saturation-independent signal loss due primarily to O2 collisions and uptake into tissue and blood. The Lorentzian functions were defined as

$${\Gamma }_{\text{TP}}\left(f\right)=\frac{{\Gamma }_0\left(f_0\right)}{1+{\left(\frac{f-f_0}{2{\delta }_{\text{TP}}}\right)}^2}$$ (3)

and similarly for Γ_RBC, where f₀ represents the line center frequency, δ the Lorentzian FWHM, and Γ₀(f₀) the amplitude. Assuming that the depolarization rate is constant during the saturation period of length τ, the left-hand side of Eq. 2 can be rewritten in terms of measured signal before (S₀) and after saturation (S₁). Integrating Eq. 2 over the time boundaries of t = 0 to t = τ, we arrive at:

$$\frac{1}{\tau }\left[ln\left(S_0\left(f\right)\right)-ln\left(S_1\left(f\right)\right)\right]={-\Gamma }_{TP}\left(f\right)-{\Gamma }_{RBC}\left(f\right)-k$$ (4)

The measured values S₀(f) and S₁(f) are therefore combined as in the left side of Eq. 4 before being fitted to the right side of Eq. 4, allowing the δ, f₀, and Γ₀(f) to vary.

For image analysis in this study, slices were analyzed separately except for comparison with the CSI measurement for which the slices were added into a single projection image. The lung was segmented by thresholding and elimination of disconnected pixels. When using two different excitation flip angles (α₀, α ₁) the depolarization ratio needs to be corrected for the fact that both the excitation flip angle and the RF-induced GP signal decay are higher for the second acquisition than if both flip angles were the same. Therefore, for the employed Cartesian acquisition with M phase encoding lines the GP depolarization maps were calculated from the pre- and post-saturation image intensities as^39,43:

$$d=[1-\frac{S_{1}sin\left({\alpha }_0\right)}{S_{0}sin{(\alpha }_1)}{\left(\frac{cos\left({\alpha }_0\right)}{cos\left({\alpha }_1\right)}\right)}^{M/2}]$$ (5)

To eliminate depolarization due to T₁, RF-induced signal loss and any compromised selectivity in the DP saturation pulses we propose the use of a Δ-frequency correction. As in conventional XTC MRI measurements, the depolarization for GP saturation at f_TP was corrected by dividing by control measurements at -f_TP³⁹. However, the saturation at Δ_TP = f_TP - Δ was used to correct the GP depolarization maps following saturation at f_RBC:

$$d_{\text{RBC},\text{corr}}=\frac{d_{\text{RBC}}}{d_{\mathrm{\Delta }\text{TP}}}.$$ (6)

The mean depolarization for the entire lung was calculated as a weighted average by multiplying each pixel by its pre-saturation intensity before adding all depolarization values. Figure 2 displays a schematic for generating RBC-GP and TP-GP exchange maps from initial spin-density maps and applying either Δ-frequency or control-study (CTRL) correction.

Figure 2.

A diagram of the image analysis process, starting from the initial acquired spin density maps to the RBC and TP exchange maps, correcting for imperfect selection of the targeted resonance. For each of the 5 saturation frequencies, two images are acquired during breath-holds following ventilation with Xe/O2 gas mixture. Post/pre-saturation ratios are taken for each frequency, and lastly, these ratio maps are divided by either the ratios maps of the opposing resonance saturation frequency (d_RBC,corr) or the control measurement to yield the corrected compartment-specific GP depolarization maps d_RBC,corr or d_TP,corr, respectively.

When comparing z-spectrum fits corresponding to different saturation intensities and saturation schemes, we found that continuous and pulsed schemes with the same average RF saturation power P_avg displayed common features. Since RF power is hardware specific we converted it into the hardware-independent B_1,rms. Noting the proportionality between P_avg and $B_{1,rms}^2$:

$$B_{1,\mathrm{r}\mathrm{m}\mathrm{s}}^2=k{P}_{\mathrm{a}\mathrm{v}\mathrm{g}}$$ (7)

and the general relationship between the flip angle rate α / t, the gyromagnetic ratio γ and B₁,

$$\frac{\alpha }{t}=\gamma B_1$$ (8)

we evaluated the proportionality constant k in the simple case of the flip-angle calibration power P_cal, which corresponds to the square pulse power delivering a 90-degree (¼ rotation) flip angle in 1ms:

$$B_{1,\mathrm{c}\mathrm{a}\mathrm{l}}=\frac{\left(1/4\right)}{\gamma \left(1\mathrm{m}\mathrm{s}\right)}=\sqrt{k{P}_{\mathrm{c}\mathrm{a}\mathrm{l}}}$$ (9)

Individual saturation pulse energies were calculated by numerically integrating the instantaneous power over the pulse length; we found that the Gaussian pulse shape used for discrete saturation contained 0.299 of the energy of a square pulse with the same length and maximum power. B_1,rms of these pulses was therefore

$$discrete:B_{1,\mathrm{r}\mathrm{m}\mathrm{s}}=\sqrt{k{P}_{\mathrm{a}\mathrm{v}\mathrm{g}}}=\sqrt{k{P}_{\mathrm{m}\mathrm{a}\mathrm{x}}\times 0.299\times \frac{\left(pulselength\right)}{TR}}$$ (10)

B_1,rms of the continuous pulses was calculated by direct application of Eq. 7:

$$continous:B_{1,\mathrm{r}\mathrm{m}\mathrm{s}}=\sqrt{k{P}_{\mathrm{m}\mathrm{a}\mathrm{x}}}$$ (11)

with P_max adjusted between saturation schemes on the scanner. In the case of continuous pulses, where B₁ is constant, the RF intensity can also be expressed as a constant flip angle rate (degrees / ms) using Eq. 8. Both are given as appropriate in the Results section to facilitate intuitive interpretation of saturation conditions.

Results

Figure 3 compares GP z-spectra of the DP region in a healthy rat for two CW RF pulses of different flip angle rates to those acquired using a series of discrete Gaussian RF inversion pulses with comparable B_1,rms. Despite the differences in saturation scheme, the z- spectra appear qualitatively similar. For the selected B_1,rms, the GP depolarization increases with RF power and both DP resonances are clearly discernible. There is some overlap between the peaks, however, and the peak widths also appear to widen slightly with B_1,rms, indicating reduced saturation pulse selectivity.

Figure 3.

Z-spectra for two CW saturation pulses applied with B_1,rms of 1.7 μT (7°/ms) and 3.1 μT (13°/ms), and for a series of Gaussian inversion pulses with a B_1,rms of 2.1 μT. Discrete and CW saturations resulted in qualitatively similar z-spectra, with the widths of both the TP peak centered at 198 ppm and RBC peak centered at 211 ppm broadening with increasing RF power.

We quantified the peak broadening observed in Fig. 3 by fitting Eq. 4 to the z-spectra for 9 different CW pulses with different flip angle – RF duration combinations and 9 discrete pulses of similar B_1,rms. Figure 4A displays modified GP z-spectra of the DP region in a healthy rat for CW saturation pulses with total flip angles of either 5,400° or 10,000° and durations ranging from 75 to 750 ms, expressed as an applied flip-angle-per-millisecond and the equivalent B_1,rms. Z-Spectra for discrete saturation pulses of similar B_1,rms are depicted in Fig. 4B. As B_1,rms increases, the depolarization rate and the widths of the two peaks in the z-spectra also increase, diverging more and more from the z-spectrum for low-power saturation pulses shown in Fig. 3. However, the measured GP depolarization rate at the RBC resonance begins to decline relative to the TP resonance for a B_1,rms—exceeding 3 μT—until a single, broad peak centered at the TP resonance predominates.

Figure 4.

Z-spectra of the DP regime as a function of B_1,rms for CW (A) and discrete Gaussian (B) saturation pulses. The spectra were fitted using Eq. 4: The blue curve represents the total fit, and the red and green lines represent the RBC and TP contributions to the fit, respectively. The final panel in (B) only shows the total fit as individual fitting of the compartmental contributions did not yield plausible results due to the disappearance of the RBC signal at that power. Note that, for the purpose of clarity, the top and bottom rows use different scales for the depolarization rates.

Figure 5 displays a quantitative analysis of the Lorentzian fits in Fig. 4 as a function of B_1,rms. For low-power saturation pulses the GP depolarization (Fig. 5A and 5B), FWHM (Fig. 5C), and the TP contribution to the GP depolarization (Fig. 5D) appear to increase in synchrony, mainly as a function of B_1,rms, and irrespective of pulse shape. However, above ~10 μT the fitting parameters, while still qualitatively similar, begin to diverge quantitatively for the CW and discrete saturation pulses. For the Gaussian pulse shape chosen for this study, discrete pulses generally induce a higher GP depolarization rate while being less selective than CW pulses of the same B_1,rms. For both types of saturation pulses, the linewidth of the two DP peaks in the z-spectra appears to increase with B_1,rms, most significantly in the high B_1,rms range (Fig. 5C). These quantitative findings confirm the qualitative observation of waning compartment selectivity for higher depolarization rates based on Figs. 3 and 4. The loss of compartment selectivity is not symmetric, however. Instead, the contribution to the GP depolarization due to exchange with the TP compartment when saturating the RBC resonance increases rapidly for B_1,rms larger than ~10 μT, although the RBC contribution when saturating the TP resonance does not change detectably (Fig. 5D).

Figure 5.

Logarithmic plots of the fitting parameters for the Lorentzians in Figure 4 as a function of B_1,rms for both discrete and CW saturation pulses. (A) Total GP depolarization rate measured for four different saturation frequencies: Δ_TP, f_TP, f_RBC, Δ_RBC. For low B_1,rms the GP depolarization rate for both saturation pulse types increases in parallel, but diverges beyond ~10 μT, at which point the depolarization rate for the discrete pulses begin to quickly outpace those of the CW pulses, at both on- and off-resonance frequencies. (B) Low-power region of the plot in panel (A). (C) FWHM of the z-spectra lines associated with the TP and RBC compartments. As B_1,rms increases, both lines broaden and the compartment selectivity of the saturation decreases, first gradually and then more rapidly for B_1,rms > ~10 μT. In the high B_1,rms regime, discrete pulses, particularly for the TP fit, exhibit greater widths than their CW counterparts. The low FWHM for the discrete saturation with a B_1,rms of 27.5 μT is an artefact of the unreliable fitting of the disappearing RBC peak at high saturation power. (D) Relative contribution of the TP compartment to the total GP depolarization at four different saturation frequencies. For all saturation powers, almost the entire GP depolarization is associated with the exchange of saturated TP magnetization with the GP at both f_TP and Δ_TP. On the other hand, while the TP contribution at f_RBC and Δ_RBC is low at low power, it begins to rapidly increase, and eventually dominate, at higher RF saturation power.

The characteristics of the global z-spectra in Fig. 4 also affect the spatially resolved GP depolarization maps employed by the XTC MRI technique. Figure 6 depicts such coronal depolarization maps for five different center frequencies (f) of the CW saturation pulses (see Fig. 1B) in a healthy rat: f_TP, f_RBC, Δ_TP, Δ_RBC, and a control frequency – f_TP. All saturation pulses had a total flip angle of 10,000°, with a duration of either 750 ms (13.3°/ms, 3.1 μT, Fig 6 top row) or 75 ms (133°/ms, 31.4 μT, Fig 6 bottom row). In agreement with the z-spectra in Fig. 4, the mean GP depolarization is highest for saturation at the TP resonance, followed by saturation at the RBC resonance—suggesting a higher volume of exchange between GP-TP than between GP-RBC. Centering the saturation pulses at the control frequency, at which negligible saturation of the DP magnetization occurs, yields mean GP depolarizations of 3.7% and 13% for the short and long saturation pulses, respectively, reflecting the inherent T₁ of the GP magnetization due to the presence of paramagnetic oxygen, the effect of the RF pulses, and xenon gas removal by the blood stream. For the 13.3°/ms (3.1 μT) saturation pulse, the GP depolarizations at Δ _TP and Δ_RBC are of similar magnitude and only slightly higher than in the control measurement. For the 133°/ms (31.4 μT) pulse, however, the on-resonance depolarization for the TP and the RBC compartments is lower, while the off-resonance depolarization at Δ_TP and Δ_RBC is much higher relative to the control measurement than for the 13.3°/ms pulse. While higher RF power depolarizes the GP at a higher rate, the total depolarization is reduced because the saturation transfer to the GP is limited by the gas exchange rate between the DP and GP compartments.

Figure 6.

CW-XTC MRI depolarization maps using two different saturation pulses with durations of 750 ms (3.1 μT, 13.3°/ms) and 75 ms (31.4 μT, 133°/ms) in the top and bottom rows, respectively. Saturation was applied at five different saturation frequencies: f_TP, f_RBC, Δ_TP, Δ_RBC, and CTRL. While the total GP depolarization for on-resonance saturation (f_TP, f_RBC) is higher for the longer, lower powered saturation pulse, it is lower per unit time. The reverse is true for off-resonance saturation (Δ_TP, Δ_RBC) due to the lower compartment-selectivity of the higher-powered pulse.

To further investigate the loss of saturation selectivity with increasing RF power, we measured the GP depolarization for different B_1,rms with the saturation center at Δ_TP (Fig. 7A) and Δ_RBC (Fig. 7B) in coronal images acquired in 5 different rats, averaged over 3 repetitions per rat. The lower the GP depolarization at these frequencies, the better the selectivity of the saturation scheme. As demonstrated in Fig. 7, the mean GP depolarization at the off-resonance frequencies is below ~10% for CW pulses up to ~3 μT, but exceeds 20% at Δ_RBC and 35% at Δ_TP at the highest B_1,rms of 31.4 μT. While these trends are clearly apparent for the mean GP depolarization values, the measurements for individual animals are considerably dispersed around those means due to variabilities in the ventilation volume and cardiogenic signal fluctuations.

Figure 7.

Average GP depolarization measured with the XTC MRI technique saturating off resonance at Δ_TP (A) and Δ_RBC (B) in five different rats. Five different CW and discrete saturation schemes were compared. Low off-resonance GP depolarization indicates good compartment selectivity. Across all saturation schemes, the higher the GP depolarization, the lower the compartment selectivity.

As a consequence of the observed decrease in frequency selectivity with increasing saturation power, the measured GP depolarization for one of the DP compartments is biased by the volume of the other. In principle, however, this effect could be partially corrected for by measuring the GP depolarization at symmetric off-resonance frequencies (Δ_TP and Δ_RBC). Nevertheless, a careful analysis of the measured z-spectra reveals the sequential nature of the xenon magnetization diffusing first into the TP compartment and only then to the RBC compartment. As the saturation becomes less selective at higher B_1,rms a correction of the TP saturation with the Δ_RBC saturation is not helpful, as the latter becomes dominated by the TP saturation as well while the RBC saturation remains negligible (Figs. 4 and 5). Figure 8 displays representative GP depolarization maps using either the control or the Δ_TP measurement for correction, as well as the derived (RBC/TP)_CORR ratio maps for several different CW and discrete RF pulse saturation schemes. Using the Δ_TP-frequency correction always results in a lower apparent GP depolarization value than the maps generated using the control measurement, and this difference increases with flip angle rate. At the highest applied RF saturation power (Fig. 8, bottom row), the corrected GP depolarization at the RBC frequency drops to zero, in agreement with the z-spectroscopy measurements, while the conventional control measurement correction erroneously yields over 40% GP depolarization. In combination, these maps illustrate the impact on XTC MR imaging results of both the TP line broadening and the disappearance of the RBC line in the corresponding z-spectra in Fig. 4.

Figure 8.

TP and RBC compartment-selective XTC MRI exchange maps generated using either a conventional control measurement (TP and RBC correction) or the proposed Δ-frequency correction method (RBC only) and their mean GP depolarization (Columns 1–3). Maps are based on several different saturation pulse parameters. The corrected RBC saturation maps diverge with increasing B_1,rms and the control-measurement corrected maps always exhibit higher depolarization values than the Δ_TP corrected maps. Even at the highest B_1,rms of 31.4 μT at which spectroscopic measurements have confirmed the absence of any significant RBC contributions, RBC/CTRL indicates approximately 42% GP depolarization while RBC/Δ_TP is in agreement with the z-spectroscopy results. The corrected RBC/TP ratio (right-most column) remains fairly constant until the applied RF saturation power is so high that the RBC contribution is suppressed by the saturation of the xenon magnetization in the TP compartment in transit to the hemoglobin binding sites.

Figure 9 shows an RBC/TP map derived from images acquired at end-exhale using a CSI sequence with a short TR and a high flip angle (TR_90°,eff = 313 ms⁴⁴) alongside both uncorrected and Δ_TP-frequency-corrected RBC/TP maps acquired with different saturation schemes using the XTC MRI technique. Because CSI quantifies the DP signal directly, the resulting maps depict the downstream accumulation of DP magnetization, while XTC MRI only measures the xenon signal in direct exchange with the alveolar volume. These two measurements are therefore similar but not equivalent for short TR_90°,eff. Despite this caveat, the mean values of the CSI- and XTC-based RBC/TP maps are quite similar, and converge even more when the Δ_TP-frequency correction is applied. The orange box highlights saturation parameter sets for which the most accurate (RBC/TP)_CORR measurements can be expected. At lower B_1,rms, the depolarization efficiency is too low and results in noisy measurements; at higher B_1,rms, the disappearing RBC contribution makes these measurements unsuitable for quantifying gas exchange between the alveolar volume and hemoglobin.

Figure 9.

RBC/TP ratio map from a short-TR_90°,eff CSI acquisition (far left) compared to ratio maps based on XTC MRI measurements using six different discrete and CW saturation schemes with B_1,rms values ranging from 1.7 to 31 μT with and without Δ-frequency correction for the RBC saturation, in a different rat than that shown in Fig. 8. The corrected RBC/TP ratio maps within the orange box are in good agreement with the CSI measurement. Lower B_1,rms values result in insufficient depolarization and noisy maps, while higher-powered saturations suppress the RBC component, skewing the ratio downward.

Discussion

In this work, we explored the use of CW saturation during the contrast-preparation phase of the XTC MRI technique (CW-XTC) in healthy rats, and compared it to the typically employed discrete RF saturation pulses.³⁹ Applying a series of discrete saturation pulses permits a well-controlled manipulation of the DP magnetization, yielding a subsequent GP depolarization that is a function of the RF frequency, the saturated DP volume, and the delay time between consecutive pulses. However, several dozen RF pulses and delay times of tens of milliseconds are usually required to induce significant GP depolarization, resulting in contrast preparation periods of a few seconds. In principle, this preparation time could be minimized by continuously irradiating the targeted DP frequency with sufficient power to induce significant DP depolarization on the time scale of the pulmonary xenon gas exchange time constant. The ongoing gas exchange between the lung tissue and the alveolar airspaces leads to a continuous depolarization of the GP magnetization that can then be quantified as a decrease in GP signal after contrast preparation, just as with the conventional XTC MRI technique. Despite the similarities, however, there are also clear differences between these two types of contrast preparation that we further investigated here.

To compare the GP depolarization for the various RF saturation pulses, we classified the pulses by their B_1,rms and, in the case of CW pulses, by the more intuitive quantity of an applied flip angle per millisecond. The effect of a given DP saturation scheme on the GP magnetization was assessed by measuring the GP depolarization for different saturation pulse center frequencies within the DP region in the form of z-spectra. In the low-power regime (< ~10 μT) the resulting z-spectra have a very similar B_1,rms dependence and qualitative appearance, clearly delineating the presence of separate RBC and TP compartments standing in exchange with the GP. In addition to the effective GP depolarization rate, the compartmental selectivity of a given RF saturation scheme is also of great importance when interpreting the measurement results: i.e., to what extent the resulting GP depolarization for a saturation pulse centered at the TP resonance is affected by the volume of the RBC compartment, and vice versa. As expected, increasing the RF power yields an increase in GP depolarization rate; it also reduces saturation selectivity, as reflected in the broadening of the peaks in the z-spectra.

However, these general trends do not affect the two DP compartments equally. Based on the peak fits in the acquired z-spectra, the on-resonance (f_TP) and off-resonance (Δ_TP) TP saturation at all RF powers yields a higher GP saturation than the corresponding RBC on-resonance (f_RBC) and off-resonance (Δ_RBC) saturations. When saturating the TP on- and off-resonance frequencies, approximately 90–95% of the GP depolarization can be attributed to exchange between the TP and GP compartments, with the rest being attributed to off-resonance contributions from the RBC compartment. However, when saturating the RBC on- and off-resonance frequencies, the contribution of the exchange between the RBC and GP compartments is RF power dependent. While 80–90% of the GP depolarization is associated with RBC-GP gas exchange at low RF power, this contribution quickly decreases for B_1,rms exceeding 10 μT. (Note that the GP depolarization followed a similar trend when saturating at Δ_RBC as at f_RBC, but the measurement was less reliable due to the low absolute depolarization rate at low B_1,rms.) This RF power dependence mirrors a rapidly increasing GP depolarization at the TP off-resonance frequency, indicating an increased TP saturation when irradiating at f_RBC.

Dissolved gases subject to the XTC effect traverse the alveolar membrane which separates the alveolar volume from the blood vessels, as well as traversing the blood plasma before and after exchanging with hemoglobin inside the RBCs. The xenon is therefore affected by off-resonance TP saturation effects during both periods of TP dissolution, regardless of whether a hemoglobin binding event takes place. If these effects, on average, lead to complete depolarization of the xenon, then the presence or absence of an RBC binding event and the associated interaction with the saturating RF cannot contribute additional depolarization and is undetectable in the z-spectra. The RBC peak is therefore suppressed at high power, where this condition is met for saturation frequencies near f_RBC. Interestingly, this phenomenon might offer a unique way to quantify the xenon transit time from the alveolar volume to the RBC compartment, which has proven too short to be accurately measured with existing techniques. This concept will be investigated in the future.

Our spectroscopic results directly translate to the GP depolarization maps obtained with the CW-XTC MRI technique. Depending on the pulse duration, the application of CW saturation pulses with identical total flip angles can result in very different GP depolarization maps. Shorter CW saturation pulses yield lower GP depolarization than longer pulses—albeit with higher depolarization per unit time—because the total depolarization rate becomes limited by the xenon dissolution rate. For the same reason, at higher RF power a series of discrete inversion pulses causes a higher GP depolarization than a CW pulse of the same duration and identical B_1,rms, as periods of more rapid longitudinal magnetization saturation are interspersed with pauses during which the imprinted DP saturation information is transferred to the GP via exchange, instead of being partially erased again by the continuation of the saturation pulse.

As discussed for the z-spectra measurements, the inadvertent saturation of one DP compartment while the saturation center frequency is centered on the other causes gas exchange information from the untargeted compartment to bias the measured exchange information from the interrogated one. As indicated by the fits of the z-spectra, the contribution of the RBC compartment to the GP depolarization when saturating at f_TP is small at ~10%; indeed, the fits are likely overestimating the effect, since the xenon gas is already partially depolarized within the alveolar wall on its transit to the hemoglobin binding sites. On the other hand, the contribution of the TP compartment to the GP depolarization when saturating at f_RBC is always at least 20%—and approaching 100% at high flip angle rates—necessitating a measurement correction. We propose to use GP depolarization measurements at Δ_TP as an estimate for the impact on the TP compartment when saturating at f_RBC. This correction scheme is effective if a symmetric TP lineshape can be assumed, which may be reasonable—given qualitatively symmetric lineshapes observed in human spectroscopy, for example, where the dissolved phase peaks are well-separated. Based on our empirical observations, the correction allows the quantification of the gas exchange between the alveolar volume and hemoglobin until the off-resonance saturation of the TP compartment destroys most of the xenon magnetization before it can bind to hemoglobin. In healthy rats this effect becomes dominant when the B_1,rms approaches the 5–10 μT regime. Up to that point, despite large quantitative differences in the individual depolarization maps, the corrected RBC-to-TP ratio remains approximately constant. For higher B_1,rms, the contribution of the RBC compartment to the measured GP depolarization is diminished by the mechanism discussed above, resulting in inaccurate GP depolarization maps for RBC saturation. The corrected RBC-to-TP ratio in the low RF power regime was found to be in good agreement with approximately equivalent CSI measurements, further confirming the consistency of CW saturation with existing techniques.

In all our studies, at low RF pulse power up to a B_1,rms of approximately 5–10 μT, CW and discrete Gaussian saturation pulses exhibited qualitatively and quantitatively similar saturation behavior. However, with further increases in power the discrete pulses became more depolarizing but less selective in their saturation profile than equivalent CW pulses. While the lower selectivity of our Gaussian pulses seems to contradict previous in vitro Hyper-CEST studies⁴⁴, this effect could be caused by the short DP T2* and more complex gas exchange mechanisms in the in vivo lung or be an indication that an optimized pulse profile could significantly improve the selectivity of discrete RF saturation pulses. We also did not investigate the practical implications of these characteristics for human studies. While limitations on achievable peak B₁ strength might allow for more efficient GP depolarization with CW saturation in human-size whole body MRI systems this benefit could be at least partially negated by the higher associated SAR.

To improve the comparability of our RF saturation parameters, we kept the total duration of the saturation segment at 750 ms, and as many studies as possible were conducted in the same animal during the same imaging session. The downside of this approach was that the GP depolarization was not optimized for contrast-to-noise⁴³ and there were unavoidable drifts in some physiological parameters due to the large number of different studies spread out across multi-hour imaging sessions, with both factors contributing to the observed measurement variability.

With the validity of CW-XTC MRI supported by conventional XTC and CSI acquisitions, future measurements of actual lung function using this technique with optimized acquisition parameters will be much more robust, permitting the same or better regional analysis capabilities compared to conventional XTC MRI methods. We further anticipate that, at 3T, a translation to human imaging will benefit from the larger frequency separation of the DP compartments in humans (~20 ppm vs ~13 ppm in rats), with acceptable compartment selectivity even without our proposed off-resonance correction method. Further, both the high-power demands on the broadband amplifier when driving a human-size chest coil and the total SAR impose inherent limitations on the achievable flip angle rates, so that the RBC suppression found in rats may be less easily observable.

Conclusions

In the low power regime of less than 5–10 μT the use of CW RF saturation pulses during the contrast preparation phase of XTC MR imaging yields GP depolarization rates similar to conventionally-used discrete Gaussian RF pulses of equivalent B_1,rms. While the lower peak B1 amplitudes of CW pulses for a given B_1,rms may permit for more rapid contrast preparation, which would be of great importance for imaging both small animals and uncooperative or sick human patients, this benefit could be partially negated by the associated higher SAR. We found the compartment-selectivity of both saturation schemes to be power dependent, although we propose an off-resonance correction method that can drastically reduce the measurement error for compartment-selective RBC saturation. In addition, the sensitivity of XTC MRI to the volume of the RBC compartment diminishes at high RF power, as the xenon magnetization is already saturated in transit from the alveolar GP through the TP compartment; however this effect may also provide the opportunity to quantify the gas exchange time constant between alveolar volume and hemoglobin.