MR properties of 19F C3F8 gas in the lungs of healthy volunteers: T2∗ and apparent diffusion coefficient at 1.5T and T2∗ at 3T

To measure the transverse relaxation time ( T2∗ ) and apparent diffusion coefficient (ADC) of 19F‐C3F8 gas in vivo in human lungs at 1.5T and 3T, and to determine the representative distribution of values of these parameters in a cohort of healthy volunteers.


| INTRODUCTION
Currently, lung imaging with fluorinated gases (SF 6 , C 2 F 6 , C 3 F 8 , C 4 F 8 1 ) MRI is not as well-characterized as hyperpolarized (HP) gas MRI, with a relative paucity in the literature. For example, there have already been numerous longitudinal and clinical studies performed with 3 He and 129 Xe gases. [2][3][4] In addition, typical values of MR measurable parameters for gas phase 3 He and 129 Xe have been characterized in vivo, such as T * 2 , 5-7 T 1 , 8 T 2 , 9 and the apparent diffusion coefficient (ADC). [10][11][12][13] These values have been used to optimize pulse sequence design for improved ventilation image quality, [14][15][16] and also to inform diffusion-weighted imaging (DWI) acquisition strategies for quantitative microstructural imaging with 3 He and 129 Xe. 11,12,17 The inherently low MR signal and short T * 2 of fluorinated gases results in lower signal-to-noise ratio (SNR) and necessitates lower image resolution when compared with HP gas imaging. 18 Recently, there have been advances in sequence optimization for fluorinated gas imaging using ultrashort echo time and steady-state free precession methods. 19,20 However, to date, there has only been preliminary investigation on whether fluorinated gas imaging can be used routinely to provide suitably robust quantitative measures of lung microstructure and function. [21][22][23] 1.1 | Transverse relaxation-T *

2
The T * 2 relaxation parameter has been shown to depend on physiological changes in different tissues/organs with 1 H MRI, 24,25 and is, therefore, an important parameter for quantitative imaging. For C 3 F 8 in phantoms, T 1, T 2 , and T * 2 is approximately 6-8 ms when diluted in nearly 100% O 2 and approximately 18-20 ms for undiluted (100%) C 3 F 8 at 95.2 kPa. In contrast, for 129 Xe and 3 He the T 1 reduces from hours to less than 30 s when mixed with O 2 28,29 in the lungs, whereas the T 2 is lower than 3 s. When measured in human lungs T * 2 is 28 ms and 14 ms at 1.5T and 3 T for 3 He 30,31 , respectively, and 52 ms and 24 ms at 1.5 T and 3T for 129 Xe, 6 respectively. The T * 2 of HP gases has also been shown to change with lung inflation level and decreases at distinct physical susceptibility interfaces, such as around the major blood vessels and at the diaphragm, 6 though correlation with disease pathologies has not yet been studied. The T * 2 for C 3 F 8 (measured through nonlocalized lung spectroscopy) has been shown to be sensitive to modulation of tissue magnetic susceptibility, 23 thus the T * 2 may also be a sensitive marker of lung microstructure variation.

| Apparent diffusion coefficient
In lung imaging with HP 3 He and 129 Xe, DWI is routinely used to probe the lung microstructure using the measurement of ADC and theoretical models of multiple b-value HP gas DWI. 10,[32][33][34] The measured ADC is sensitive to changes in alveolar dimensions with diseases, such as emphysema, 11 idiopathic pulmonary fibrosis, 35,36 and chronic obstructive pulmonary disease. 37,38 Furthermore, even relatively small ADC changes related to lung inflation level, 39,40 age, 41 and physiological distribution within the lungs 42 are observable. ADC measurements with fluorinated gases have been performed in rats with C 2 F 6 43,44 and SF 6 , 45 demonstrating that there is restricted diffusion and that the ADC is larger in emphysematous lungs. In contrast to measurements made in excised lungs with 100% C 2 F 6 46 and C 3 F 8 , 47 performing in vivo ADC measurements with 79% C 3 F 8 + 21% O 2 will accurately provide a normative range of values and distribution across healthy subjects. Furthermore, such a study will establish the feasibility of performing in vivo C 3 F 8 ADC studies with the constraints imposed by the sensitivity of a thoracic radiofrequency (RF) coil, breath-hold limitations on image acquisition time, and the variability of gas concentration through voluntary continual breathing rather than controlled pumping.

| Overview
Determining the relative sensitivity and achievable quality of DWI with C 3 F 8 in relation to 129 Xe was one aim of this study. Furthermore, the value and distribution of T * 2 in vivo is also unknown. Therefore, in this study theT * 2 and ADC with 19 F imaging of 79% C 3 F 8 + 21% O 2 was investigated in the lungs of healthy volunteers. In the same eight volunteers, T * 2 mapping was carried out and the change from TLC to FRC was evaluated at 1.5T. In addition, T * 2 mapping at TLC was performed at 3T in seven of the volunteers to evaluate the field strength dependence of T * 2 . To determine the sensitivity of C 3 F 8 ADC to changes in lung microstructural length scales, the differences obtained at FRC or TLC, and the regional distribution within the lungs, was investigated in eight healthy volunteers. ADC mapping with 129 Xe was carried out in six of the volunteers as a means of comparison with the equivalent established and higher SNR HP gas techniques.

| Overview
In total, eight subjects, seven male and one female (S1-S8, aged 29 ± 4 years), were imaged following informed consent. All in vivo MRI experiments were performed under the approval of the UK National Research Ethics Committee and the local National Health Service research office. The clinical grade 79% C 3 F 8 /21% O 2 gas mixture (BOC Special Products, Guildford, UK) was inhaled from a 25-L reservoir bag via a mouthpiece and three-way valve and mouthpiece (Hans Rudolf, Shawnee, KS). Hyperpolarization (~30%-40%) of 86% enriched 129 Xe gas was performed in house using the spin-exchange optical pumping method 48  Philips, Andover, MA) using an elliptical transmit/receive quadrature birdcage coil (RAPID Biomedical, Rimpar, Germany). Experiments at 1.5T (GE HDx; GE Medical Systems, Milwaukee, WI) with 19 F were performed with an in-house constructed transceiver array, 49,50 which improves the average SNR by a factor of approximately 5 throughout the lung region when compared with a single transceiver vest coil. 129 Xe imaging at 1.5T was performed with a flexible transceiver vest coil (Clinical MR Solutions [CMRS], Brookfield, WI). Table 1 lists the various imaging acquisition parameters for both C 3 F 8 and 129 Xe scanning. In vivo 19 F-C 3 F 8 T * 2 measurements were performed at 1.5T (FRC and TLC for eight subjects) and at 3T (TLC for seven subjects). In addition, in vivo ADC measurements at 1.5T with 19 F-C 3 F 8 (FRC and TLC for eight subjects) and 129 Xe (FRC and TLC for six subjects) were performed and compared. Details of sequence, parameter choice, and scan procedures used in this work are included in following sections.

| T * 2 mapping
At 1.5T, 19 F T * 2 mapping was performed at lung-inflation levels of TLC and FRC, with the following sequence of breathing maneuvers: (1) Four deep breaths were taken of the gas mixture via a three-way valve from a 25-L Douglas bag to fully saturate the lungs; (2) imaging was then performed under breath-hold apnea at TLC (22 s); (3) the volunteers then exhaled through the three-way valve and continued to breath normally with inhaled gas coming from the Douglas bag; and (4) once the volunteer signaled they were able to commence a second breath-hold, imaging was repeated after exhalation to FRC.
From multiecho SPGR acquisition sequences the signal for each echo time (S n ℎ ) was fit voxel-wise according to: where ΔTE is the spacing between echoes, n ℎ is the echo number and S 1 is the amplitude of the first echo image. The fitting was performed only on pixels with an SNR>10 for the first echo at 1.5T (Δ = 2.3 ms) and at 3T (Δ = 1.5 ms). This corresponds to at least ≥2.5 noise SD for n ℎ = 2, the recommended SNR threshold for pixel-wise truncation of measurements, 51 for T * 2 >1.7 ms at 1.5T and T * 2 > 1.1 ms at 3T. To evaluate the distribution of T * 2 within the lungs, averaged histograms of the T * 2 values from all slices and axial, sagittal, and coronal plots of the maps were produced.

| Apparent diffusion coefficient
The signal after an applied trapezoidal bipolar gradient (S b ) is characterized by: where S 0 is the signal without diffusion gradients, the ADC is the apparent diffusion coefficient, and the b-value and the diffusion time (Δ) of the applied pulse are described in the work by Al and Da. 34 For effective lung DWI, the length scale of the confining structure (l s ) of the alveoli must be of the same magnitude as the free diffusion length (l d = √ 2D 0 Δ) or the gradient dephasing length (l g = D 0 ∕ G 1∕3 ), which is the average length that a spin must diffuse to dephase by 2π radians. 52 Figure 1 shows the different length scale regimes in relation to potential DWI conditions typically achieved with Xe (D 0 = 14 mm 2 /s 9 ) and 3 He (D 0 = 86 mm 2 /s 9 ) in air, and C 3 F 8 mixed with 79% O 2 (D 0 = 2.7 mm 2 /s 26 ) for an average alveolar diameter of l s at approximately 250 µm. 53 For DWI with 129 Xe Δ = 8.5 ms 54 and b = 0.12 s/mm 255 the geometrical parameters derived from models of the acinar airway closely match those obtained with 3 He Δ = 1.6 ms, that has been shown to be effective for characterizing lung microstructure 54 ; therefore, this diffusion time was used for 129 Xe DWI in this study. For C 3 F 8 DWI, Δ = 2.2 ms and b = 0.18 s/mm 2 with a gradient echo sequence was used, matching that used previously with C 2 F 6 . 46 This was expected to put the measurements in the localization regime (see Figure 1), where the ADC signal is dominated by diffusional restriction at the boundaries of the lung alveolar structure. 56 To determine the sensitivity of C 3 F 8 ADC to changes in airway microstructural dimensions caused by lung inflation, the ADC was measured at both FRC and TLC and compared with the equivalent 129 Xe ADC measurements.
For 19 F ADC imaging, the same breathing maneuvers were followed as for T * 2 imaging, except that two additional images were acquired at breath-holds of TLC and FRC (22 s each) obtained sequentially while breathing from the same 25-L Douglas bag. The two images obtained at the same inflation level were averaged together for increased SNR. To perform an independent measurement of the D 0 the same ADC measurement was performed with the Douglas bag on three separate occasions. For 129 Xe imaging, a 1-L bag of gas was inhaled from FRC consisting of 400-mL N 2 gas mixed with 600-mL 129 Xe. 48 The volunteers then either breathed in room air to TLC or exhaled to FRC prior to imaging during breath-hold (16 s).
All C 3 F 8 and 129 Xe DW images were thresholded so that only voxels with SNR >15 57 were used in the calculation of ADC. To evaluate the distribution of ADC values at FRC and TLC, histograms of 129 Xe and 19 F ADC averaged over all slices were plotted for all volunteers. Furthermore, similar to the process carried out in Fichele et al, 42 the ADC gradient in the anteroposterior direction was calculated by first visually identifying the center of the lungs and then plotting the average ADC for each of the slices/pixels relative to the center for all volunteers together.

2
Maps of T * 2 in central axial, coronal, and sagittal slices for volunteer S1 are shown at 1.5T at FRC in Figure 2A, at TLC in Figure 2B, and at 3T at TLC in Figure 2C. The T * 2 values are much lower than those found in phantoms where T * 2 T 2~T1 = 18-22 ms. 20 Also, a clear decrease in T * 2 is observed around the intrapulmonary vessels and the diaphragm, where tissue-air bulk magnetic susceptibility gradients are highest. The recorded mean values for all volunteers are listed in Table 2 along with the p value for the paired t test comparing changes between the mean T * 2 at FRC and TLC (1.5T) and also between TLC at 1.5T and 3T, which is demonstrated clearly in the histograms of the T * 2 maps shown in Figure 2D.

| Apparent diffusion coefficient
ADC measurements made in the Douglas bag alone determined a D 0 of 2.54 ± 0.06 mm 2 /s for the C 3 F 8 /O 2 mixture. ADC maps generated from C 3 F 8 imaging in volunteer S5 are shown in Figure 3A (at FRC) and Figure 3B (at TLC). The mean 19 F-C 3 F 8 ADC histograms from all volunteers are shown in Figure 3C. Because of our chosen rejection criterion of SNR <15 on voxels when mapping ADC, there was a consistent exclusion of areas around the major pulmonary vessels, and in some regions around the diaphragm of volunteers in C 3 F 8 imaging. This was caused by the reduced signal from lower T * 2 and partial voluming in these regions, as observed in Figure 1, and also the longer TE required for the ADC sequence. Figure 3D-F shows equivalent maps generated from 129 Xe imaging in the same volunteer. The ADC maps in Figure 3  show regions of heterogeneous ADC near the heart and to the inferior of the lungs, as well as localized regions of lower than average ADC.  Abbreviations: ASC, apparent diffusion coefficient; FRC, functional residual capacity; TLC, total lung capacity; SNR, signal-to-noise ratio.
The results from linear regression of the anteroposterior anatomical gradients in ADC are presented in Table 2. Plots of the linear variation can be viewed in Supporting Information Figure S1.
The mean T * 2 of C 3 F 8 in lungs of volunteers was found to be higher than previously reported (1.5-2.2 ms 23,58 ). These previous measurements were performed as global whole lung spectroscopy and the returned T * 2 values are expected to be lower because of the wider B 0 inhomogeneity across the entire lung when compared with an imaging voxel. The variation of T * 2 between volunteers is predicted to be primarily dependent on the normal variations in alveolar dimensions within the population 59 and the susceptibility effects from the inhomogeneity of the tissue interfaces (differences in the bulk magnetic susceptibility 60 at the air-tissue interfaces of alveoli 61 ). Therefore, it is expected that microscopic susceptibility differences associated with different disease pathologies may also show changes in T * 2 . Our work indicates that 19 F T * 2 mapping at 1.5T is less technically challenging than at 3T because a longer T * 2 is observed at 1.5T, which is consistent with previous results obtained with HP gases. 5,6,31

| Apparent diffusion coefficient
For C 3 F 8 , longer diffusion times are required to match the same length scale as those sensitized in 3 He and 129 Xe DWI; achieving these is hindered by the low T * 2 and SNR. Although a spin-echo sequence could potentially be used to mitigate this, 19 F-C 3 F 8 DWI with a spin-echo-based sequence would result in unfeasible breath-hold times because of specific absorption rate constraints and RF power restrictions on RF-pulse duration and B 1 amplitude. In addition, any further gains in SNR are predicted to be limited because of the transmit homogeneity of the vest RF coil and the longer sequence TR of a spin-echo mandating reduced averaging. In future studies, spin-echo-based sequences could potentially be applied for the benefit of increased diffusion times. The measured in vivo ADC values are lower than the measured (D 0 = ~2.54 mm 2 /s) and previously published (D 0 = ~2.7 mm 2 /s 26 ) free diffusion coefficients of C 3 F 8 mixed with 21% O 2 , showing some sensitivity to acinar diffusion restriction. The in vivo healthy volunteer C 3 F 8 ADC values are similar to those acquired from excised healthy lungs with C 2 F 6 (1.8 mm 2 /s 46 ). In addition, clear changes in ADC between FRC and TLC were observed, as well as regional differences caused by the gravitational gradient at FRC, but not at TLC. In previous work with 3 He, a similar gradient in ADC was observed in the anteroposterior direction, 37,42 that was reduced or not observable at TLC. 39 Furthermore, previously with 129 Xe in healthy volunteers, a 22% decrease in the mean ADC was found from the anterior to the posterior of the lungs in healthy volunteers, which was not observed in patients with chronic obstructive pulmonary disease. 38 A decreasing gradient in the superoinferior direction has also been reported, 37,38,42 but was not observed in this study. Two factors may have masked the measurement of this gradient: (1) the gradient depends on the posture of the imaging subject, 42 and (2) regions of the lung next to the heart experience compression, which results in regional changes in ADC that have been observed in HP gas-diffusion imaging. 62 Based on the observed changes with lung inflation, there is a strong indication from this work that the DWI parameters used here for in vivo 19 F-C 3 F 8 ADC mapping will be able to detect changes in lung microstructure in different pathologies where changes are larger, such as in emphysema where the measured 3 He ADC can increase by a factor of two to three when compared with healthy lungs, 63 or in idiopathic pulmonary fibrosis where the 3 He ADC can increase by a factor of three to five in regions of fibrotic tissue. 36 Previous attempts at in vivo ADC measurements of C 3 F 8 in experiments with a single volunteer resulted in a maximum image SNR of approximately 15, 58,64 which is below the threshold set here for inclusion of voxels in the ADC calculation. In addition, these previous studies used shorter diffusion times (Δ = 1 ms) and smaller b-values (0.0959 s/mm 2 58 and 0.0133 s/cm 2 , 64 which places those measurements in the free diffusion regime. The reported ADC values in some regions were ≥6 mm 2 /s, which far exceeds the free diffusion coefficient and may have been a result of the low SNR and the weak b-values used in that work. In future work, ensuring that the gas mixture concentration in the lungs reaches full saturation of 79% perfluoropropane per 21% O 2 is necessary because the partial pressure strongly influences the free diffusion coefficient (approximately D 0 = ~2.3-7.7 mm 2 /s for 100%-0% partial pressure with O 2 ).

| CONCLUSIONS
By utilizing improvements in receiver design, optimized imaging parameters, and breathing maneuvers, three-dimensional in vivo ADC mapping with C 3 F 8 in the human lungs was found to be feasible with a greater resolution than previously attempted. Thus, for the first time, systematic in vivo mapping of ADC at 1.5T and T * 2 at the two clinically relevant MRI field strengths (3T and 1.5T) is presented for C 3 F 8 in the lungs of healthy volunteers, indicating sensitivity to change in acinar airways dimensions. These results show promise for future studies in lung diseases that exhibit microstructural airway changes.