Hyperpolarized 3He and 129Xe magnetic resonance imaging apparent diffusion coefficients: physiological relevance in older never‐ and ex‐smokers

Abstract Noble gas pulmonary magnetic resonance imaging (MRI) is transitioning away from 3He to 129Xe gas, but the physiological/clinical relevance of 129Xe apparent diffusion coefficient (ADC) parenchyma measurements is not well understood. Therefore, our objective was to generate 129Xe MRI ADC for comparison with 3He ADC and with well‐established measurements of alveolar structure and function in older never‐smokers and ex‐smokers with chronic obstructive pulmonary disease (COPD). In four never‐smokers and 10 COPD ex‐smokers, 3He (b = 1.6 sec/cm2) and 129Xe (b = 12, 20, and 30 sec/cm2) ADC, computed tomography (CT) density‐threshold measurements, and the diffusing capacity for carbon monoxide (DLCO) were measured. To understand regional differences, the anterior–posterior (APG) and superior–inferior (∆SI) ADC differences were evaluated. Compared to never‐smokers, COPD ex‐smokers showed greater 3He ADC (P = 0.006), 129Xe ADCb12 (P = 0.006), and ADCb20 (P = 0.006), but not for ADCb30 (P > 0.05). Never‐smokers and COPD ex‐smokers had significantly different APG for 3He ADC (P = 0.02), 129Xe ADCb12 (P = 0.006), and ADCb20 (P = 0.01), but not for ADCb30 (P > 0.05). ∆SI for never‐ and ex‐smokers was significantly different for 3He ADC (P = 0.046), but not for 129Xe ADC (P > 0.05). There were strong correlations for DLCO with 3He ADC and 129Xe ADCb12 (both r = −0.95, P < 0.05); in a multivariate model 129Xe ADCb12 was the only significant predictor of DLCO (P = 0.049). For COPD ex‐smokers, CT relative area <−950 HU (RA950) correlated with 3He ADC (r = 0.90, P = 0.008) and 129Xe ADCb12 (r = 0.85, P = 0.03). In conclusion, while 129Xe ADCb30 may be appropriate for evaluating subclinical or mild emphysema, in this small group of never‐smokers and ex‐smokers with moderate‐to‐severe emphysema, 129Xe ADCb12 provided a physiologically appropriate estimate of gas exchange abnormalities and alveolar microstructure.


Introduction
Magnetic resonance imaging (MRI) using hyperpolarized noble gases 3 He and 129 Xe provides high spatial and temporal resolution images of pulmonary gas distribution (Kauczor et al. 1997; de Lange et al. 1999;Moller et al. 2002). Parenchymal microstructure can also be probed using diffusion-weighted (DW) imaging. DW imaging takes advantage of the gases rapid molecular diffusion to generate 3 He and 129 Xe apparent diffusion coefficients (ADCs; Saam et al. 2000;Salerno et al. 2002). Although the first demonstrations of inhaled hyperpolarized gas MRI used 129 Xe gas (Albert et al. 1994;Mugler et al. 1997), until recently, most examinations in subjects with chronic obstructive pulmonary disease (COPD) used 3 He gas mainly due to the nearly threefold higher gyromagnetic ratio and higher polarizationsboth of which independently contribute to greater image quality for 3 He MRI. These previous studies showed that 3 He ADC values are reproducible in emphysematous subjects (Morbach et al. 2005;Diaz et al. 2008;Mathew et al. 2008), sensitive to lung microstructure and airspace size (Saam et al. 2000;Salerno et al. 2002), and correlate with spirometry (Salerno et al. 2002;Diaz et al. 2009), diffusing capacity of carbon monoxide (DL CO ; Fain et al. 2006), and x-ray computed tomography (CT) measurements of emphysema (Diaz et al. 2009). Importantly, 3 He MRI ADC measurements were directly compared with stereological measurements of fixed lung tissue and this showed the high sensitivity of 3 He MRI ADC to lung surface area abnormalities in emphysematous tissue (Woods et al. 2006). Despite these important findings, 3 He MRI is unlikely to be used clinically because of its relatively high cost and low abundance (Fain et al. 2010). On the other hand, 129 Xe gas is substantially more abundant, relatively inexpensive, and clinical quantities may be polarized in 30-60 min (Mugler et al. 1997(Mugler et al. , 2004. Because of this, there is considerable interest in transitioning to 129 Xe MRI. Although 129 Xe gas is capable of transmembrane diffusion (Cleveland et al. 2010) this leads to adverse systemic effects when used in high concentrations; Xe has been widely and safely used in pulmonary CT (Chae et al. 2008;Honda et al. 2012), cerebral blood flow imaging (Carlson et al. 2011), and lung scintigraphy (Suga 2002). Recent 129 Xe MRI safety studies demonstrated that the gas concentrations compatible with MRI are well tolerated in healthy subjects and those with respiratory disease (Driehuys et al. 2012;Shukla et al. 2012). Previous work showed that 129 Xe ADC (b = 12 sec/cm 2 ; Kaushik et al. 2011) and 129 Xe ADC morphometry measurements (Ouriadov et al. 2013) provide a way to distinguish between age-matched subjects with and without COPD. Moreover, although strong and significant correlations for 3 He ADC with 129 Xe ADC (b = 12 sec/cm 2 ) were reported (Kirby et al. 2012d), significantly lower 129 Xe as compared to 3 He MRI ventilation was observed in COPD ex-smokers ) and asthmatics . Regions of diminished 129 Xe as compared to 3 He ventilation are spatially related to emphysematous regions  in COPD subjects.
It is important to consider the different self-diffusion coefficients of 3 He and 129 Xe gases. For example, 129 Xe gas has a much lower diffusion coefficient than 3 He and, therefore, in order to achieve sufficient diffusion weighting for 129 Xe MRI to derive meaningful lung microstructural information, appropriately large b values and diffusion times must be applied (Sukstanskii and Yablonskiy 2012). However, increasing the diffusion weighting is coupled with a reduction in image signal-to-noise ratio (SNR); therefore it is important to experimentally demonstrate the 129 Xe ADC b value that provides the optimal balance between sensitivity to the lung microstructural abnormalities and feasibility with respect to image SNR. It is also essential to investigate the different 129 Xe MRI ADC b values and their associations with other emphysema measurements in order to better understand how imaging estimates are related to physiological measurements in COPD patients. Previous studies have focused on 129 Xe ADC generated with a b value of 12 sec/cm 2 (Kaushik et al. 2011;Kirby et al. 2012d), and therefore here we aimed to (i) determine whether 129 Xe ADC generated using b values of 12, 20, and 30 sec/cm 2 and contemporaneously acquired 3 He ADC , distinguishes older never-smokers from ex-smokers with COPD, (ii) compare 3 He and 129 Xe ADC in the superiorinferior (SI) and anterior-posterior (AP) lung regions to better understand potential dependencies for b value with regional ADC, and (iii) determine the significant associations of 129 Xe ADC with b values of 12, 20, and 30 sec/ cm 2 with DL CO and CT measurements of emphysema, similar to those recently used in the COPDGene (Regan et al. 2010), CanCOLD (Bourbeau and Saad 2013), ECLIPSE (Vestbo and Anderson 2008), and SPIROMICS (Couper et al. 2014) studies.

Subjects
All subjects provided written informed consent to the study protocol that was approved by the local research ethics board and Health Canada. COPD subjects were exsmokers with a smoking history of at least 10 pack-years between 50 and 85 years of age and categorized according to the Global Initiative for Chronic Lung Disease (2013) (GOLD) spirometry criteria. Healthy older never-smokers (<1 pack-year over their lifetime and no smoking in last 20 years) were enrolled who had no history of previous chronic or current respiratory disease.

Pulmonary function tests
A body plethysmograph (MedGraphics Corporation, Saint Paul, MN) with the attached gas analyzer was used to measure airflow, lung volumes, and DL CO according to ATS/ERS guidelines (Macintyre et al. 2005;Miller et al. 2005;Wanger et al. 2005).

Image acquisition
Magnetic resonance imaging was performed on a whole body 3.0 Tesla Discovery 750MR (General Electric Health Care, Milwaukee, WI) MRI system with broadband imaging capability. Subjects were instructed to inhale a gas mixture from a 1.0L Tedlar â bag (Jensen Inert Products, Coral Springs, FL) from functional residual capacity (FRC) and image acquisition was performed under breath-hold conditions .
Computed tomography was performed on a 64-slice Lightspeed VCT scanner (GEHC, Milwaukee, WI). A single spiral acquisition of the entire lung was acquired from apex to base with subjects in breath-hold condition after inhalation of N 2 from a 1.0L Tedlar â bag (detector configuration of 64 9 0.625 mm, 120 kVp, 100 effective mA, tube rotation time of 500 msec, and a pitch of 1.0). Reconstruction of the data was performed using a standard convolution kernel and a slice thickness of 1.25 mm.

Image analysis
3 He ADC and 129 Xe ADC with b values of 12 sec/cm 2 (ADC b12 ), 20 sec/cm 2 (ADC b20 ), and 30 sec/cm 2 (ADC b30 ) were generated as described previously (Kirby et al. 2012b). Briefly, we first segmented the 3 He MRI non-DW images using k-means cluster algorithm as described previously for 3 He MRI segmentation (Kirby et al. 2012a). The segmented binary mask for the non-DW images was applied to the DW images. The ADC maps were generated on a voxel-by-voxel basis according to equation (1): where b = 1.6 sec/cm 2 , S 0 = the segmented non-DW image, and S = the diffusion-weighted image. The regional differences in ADC were evaluated for all AP coronal slices and for each coronal slice in three SI regions of interest (ROIs). The AP gradient (AP G ) was defined as the slope of the line of best fit that described the change in ADC as a function of distance in centimeters for all slices in the anterior to posterior direction. The three SI ROI were generated by dividing the center coronal slice into three equivalent ROIs from apex to base and applying those boundaries to all slices. ΔSI was the mean change in ADC in the superior and inferior ROI. To compare 3 He and 129 Xe ADC ΔSI and AP G to better understand any dependency on b value for regional ADC differences in the respective gases, we compared the older never-smokers to the COPD subjects, as well as to only the COPD ex-smokers with emphysema. COPD subjects without emphysema are unlikely to show an increased 3 He or 129 Xe ADC ΔSI and AP G difference compared to the older never-smokers. The SNR for 3 He and 129 Xe DW images and nondiffusion-weighted (NDW) images were determined by calculating the mean voxel value within a 5 9 5 voxel ROI for four representative ROI within the lung parenchyma, and dividing by the standard deviation of the mean voxel values for four representative 5 9 5 voxel ROI within the image background where there was no lung structure, for each slice. SNR was determined for each slice and then averaged to obtain a single SNR value for each subject image. Quantitative CT measurements were performed using the Pulmonary Workstation 2.0 (VIDA Diagnostics, Inc., Coralville, IA); the relative area with attenuation values below À950 Hounsfield Units (HU; RA 950 ), À910 HU (RA 910 ), and À856 HU (RA 856 ) were generated  as well as the CT HU value at which 15% of the voxels in the frequency distribution histogram have a lower density (HU 15% ). A CT density threshold for emphysema of RA 950 > 6.8% was adopted based on a previous study that reported RA 950 = 6.8% as the upper 95% limit of predicted normal values .
Qualitative visual scoring was also performed by an expert chest radiologist (R. E.-R.) using a method adapted from Bankier et al. (1999). Emphysema scores were based on the percentage of low attenuation, tissue destruction, and vascular disruption.

Statistical methods
A Mann-Whitney unpaired two-tailed t-test was performed for statistical comparison of all pulmonary function measurements, 3 He and 129 Xe ADC,3 He and 129 Xe ADC AP G , and 3 He and 129 Xe ADC SI using GraphPad Prism version 4.00 (GraphPad Software Inc., San Diego, CA). A one-way analysis of variance (ANOVA) was used for statistical comparison of the mean SNR values for ADC DW images for 3 He and 129 Xe with all three b values using IBM SPSS Statistics 20.0 (SPSS Inc., Chicago, IL). Linear regression (r 2 ) and Pearson correlation coefficients (r) were used to determine the relationships between imaging and other measurements using Graph-Pad. A Holm-Bonferroni correction (Van Bell et al. 2004) was used to correct for multiple tests. Multivariate linear regression modeling was used to evaluate the relationship between DL CO and 3 He ADC, 129 Xe ADC b12 , 129 Xe ADC b20 , and 129 Xe ADC b30 as well as between CT density measurements and 3 He ADC, 129 Xe ADC b12 , 129 Xe ADC b20 , and 129 Xe ADC b30 using PROC REG in SAS â 9.2 Software (SAS Institute Inc., Cary, NC). In all statistical analyses, results were considered significant when the probability of making a Type I error was <5% (P < 0.05). Table 1 shows all demographic and pulmonary function measurements for four older never-smokers and 10 COPD ex-smokers (GOLD Class I: n = 1; GOLD Class II: n = 6; GOLD Class III: n = 2; GOLD Class IV: n = 1). Although there were no significant differences in age, sex, or BMI, the never-smokers and COPD ex-smokers were significantly different with respect to FEV 1 (P = 0.01), FEV 1 /FVC (P = 0.002), TLC (P = 0.02), RV (P = 0.02), FRC (P = 0.01), and DL CO (P = 0.006).   129 Xe ADC b = 12 (cm 2 /sec) 129 Xe ADC b = 20 (cm 2 /sec) 129 Xe ADC  Hyperpolarized MRI ADC Figure 1 shows the central coronal 3 He ADC, 129 Xe ADC b12 , 129 Xe ADC b20 , and 129 Xe ADC b30 maps for an older never-smoker and two COPD ex-smokers. ADC maps appear brighter and more heterogeneous for the COPD ex-smokers in comparison with the healthy neversmoker for 3 He MRI as well as for 129 Xe MRI for all b values. As shown in Figure 2, there were statistically significantly greater ADC values for all COPD ex-smokers as well as only those COPD ex-smokers with emphysema in comparison to the older never-smokers for 3 He ADC (all COPD: P = 0.002; COPD with emphysema: P = 0.006), 129 Xe ADC b12 (all COPD: P = 0.002; COPD with emphysema: P = 0.006), and 129 Xe ADC b20 (all COPD: P = 0.04; COPD with emphysema: P = 0.006), but not for 129 Xe ADC b30 (P = 0.06).
The Pearson correlation coefficients between CT lung density thresholds (RA 950 , RA 910 , RA 856 , HU 15% ) and visual CT emphysema scores are shown in Table 3. Following Holm-Bonferroni correction, 3 He ADC was significantly correlated with RA 950 (r = 0.90, P = 0.008) and HU 15% (r = À0.92, P = 0.004); 129 Xe ADC b12 was significantly correlated with RA 950 (r = 0.85, P = 0.03) and HU 15% (r = À0.90, P = 0.008); 129 Xe ADC b20 was significantly correlated with HU 15% (r = À0.86, P = 0.02); 129 Xe ADC b30 was not significantly correlated with any CT emphysema measurements. Table 4 shows the results of multiple linear regression models with 3 He ADC, 129 Xe ADC b12 , 129 Xe ADC b20 , and 129 Xe ADC b30 as the predictor variables. In the multivariate regression model for DL CO with 3 He and all 129 Xe ADC values included, 129 Xe ADC b12 (P = 0.049) was the only significant predictor. In the multivariate regression model for RA 950 with 3 He and all 129 Xe ADC values  129 Xe magnetic resonance imaging apparent diffusion coefficient (ADC) measurements for older never-smokers, all chronic obstructive pulmonary disease (COPD) ex-smokers, and COPD ex-smokers with emphysema. Box-and-Whisker plot where box extends from the 25th to 75th percentiles of ADC values, the line within the box is the median ADC value and error bars are the minimum and maximum ADC values. Significance of difference (P < 0.05) determined using a Mann-Whitney unpaired two-tailed t-test. included, the only significant predictor was 3 He ADC (P = 0.03). We also evaluated the multivariate regression model for HU 15% , and although the overall model was statistically significant (P < 0.05), 3 He ADC, 129 Xe ADC b12 , or 129 Xe ADC b20 were not significant predictors.

Discussion
The low and variable polarization levels and low diffusivity of 129 Xe gas results in technical challenges for extracting physiologically meaningful pulmonary microstructural information based on 129 Xe MRI. Moreover, although previous studies (Kaushik et al. 2011;Kirby et al. , 2012d reported 129 Xe ADC based on a b value of 12 sec/ cm 2 , it is still unclear how to provide the optimal balance between sensitivity to the lung microstructural abnormalities in emphysematous patients and feasibility with respect to image quality. Moreover, 129 Xe MRI ADC generated using different b values may access different regions of the lung and likely probe different lung microstructure length scales thereby providing somewhat different spatial information relative to 3 He ADC. For a better understanding of these measurements, we investigated the different 129 Xe MRI ADC b values and their associations with anatomical and physiological measurements of COPD in a small but well-characterized group of older never-smokers and COPD ex-smokers. We observed (i) significantly greater ADC values for COPD ex-smokers compared to older never-smokers for 3 He ADC, 129 Xe ADC b12 , and ADC b20 , but not for 129 Xe ADC b30 , (ii) significantly greater anterior-posterior ADC differences in COPD ex-smokers with emphysema compared to older never-smokers for 3 He ADC, 129 Xe ADC b12 , and ADC b20 , but not for 129 Xe ADC b30 ,(iii) significantly different mean ADC in the superior as compared to inferior lung regions between older neversmokers and COPD ex-smokers for 3 He ADC, but not for 129 Xe ADC, and finally, (iv) strong and significant correlations for 3 He and 129 Xe ADC with DL CO and CT measurements of emphysema, but in a multivariate model, 129 Xe ADC b12 was the only significant predictor of DL CO and 3 He ADC was the only significant predictor of CT measured RA 950 .
As expected, we observed significantly greater (worse) ADC for COPD ex-smokers than for older never-smokers for 3 He, and as well for 129 Xe ADC b12 and ADC b20 , but  129 Xe magnetic resonance imaging apparent diffusion coefficient (ADC) anterior-posterior gradients (AP G ) for older neversmokers, all chronic obstructive pulmonary disease (COPD) ex-smokers, and COPD ex-smokers with emphysema. Box-and-Whisker plot where box extends from the 25th to 75th percentiles of ADC values, the line within the box is the median ADC AP G value and error bars are the minimum and maximum ADC AP G values. Significance of difference (P < 0.05) determined using a Mann-Whitney unpaired two-tailed t-test. not for 129 Xe ADC b30 . There was a large difference in mean ADC between subject subgroups for both 3 He and 129 Xe MRI, and this important finding suggested that both 129 Xe MRI generated using b values of 12 and 20 sec/cm 2 provide adequate sensitivity to lung microstructural abnormalities. These data are also in agreement with previous findings (Kaushik et al. 2011) that 129 Xe ADC b12 can be used to distinguish between older neversmokers and COPD subjects with emphysema. However, we must point out that 129 Xe ADC b30 was not significantly different between subject subgroups and this may have been due to the low sample size as well as low SNR.
For ADC b30 , we demonstrated that the diffusion attenuation curves deviate from single exponential behavior after b = 20 sec/cm 2 , suggesting greater diffusion anisotropy (Ouriadov et al. 2013) compared to ADC b12 and ADC b20 , and this greatly influences image quality, which was poor for ADC b30 images here. It is important to note that while SNR was not significantly different for 129 Xe MRI acquired at all three b values, there was a significant correlation between image SNR and ADC for 129 Xe ADC b30 only, suggesting that there may have been inadequate image quality to derive reliable ADC measurements. Clearly these are important considerations when optimizing and selecting pulse sequence parameters since image signal is expected to be reduced when the gradient amplitude is increased. Second, there were statistically significantly greater AP G for the COPD ex-smokers with emphysema than the older never-smokers for 3 He ADC, 129 Xe ADC b12 , and ADC b20 , but not for 129 Xe ADC b30 . DW MRI is sensitive to changes in the lung microstructure (Saam et al. 2000;Salerno et al. 2002) and is therefore an indirect measure of airspace size. Since the DW images were acquired in breath-hold condition in the supine position, the measurement of the ADC on a slice-by-slice basis can be exploited to measure compression of the dependent lung due to gravity. Several studies have reported lower ADC values in the dependent (or posterior) lung slices relative to the nondependent anterior lung slices for 3 He MRI  129 Xe magnetic resonance imaging (MRI) apparent diffusion coefficient (ADC) for superior and inferior regions of interest in older never-smokers, all chronic obstructive pulmonary disease (COPD) ex-smokers, and COPD ex-smokers with emphysema. There was a significant difference in mean ADC in the superior to inferior lung regions between the older never-smokers and the COPD ex-smokers with emphysema for 3 He MRI ADC (P = 0.04), but there were no other significant differences. Significance of difference (P < 0.05) determined using a Mann-Whitney unpaired two-tailed t-test.   Fichele et al. 2004;Evans et al. 2007;Diaz et al. 2008) and 129 Xe MRI with b = 12 sec/cm 2 (Kaushik et al. 2011), which is expected due to compression of the lung parenchyma in healthy never-smokers. However, in subjects with emphysema, the gradient in alveolar expansion in the gravity-dependent direction is reduced, in part, because of gas trapping Diaz et al. 2008). The finding that 129 Xe ADC b12 anterior-posterior gradients were also reduced in the COPD ex-smokers with emphysema in comparison to healthy never-smokers is in agreement with these previous studies Diaz et al. 2008;Kaushik et al. 2011). Furthermore, Figure 5. Correlations for 3 He and 129 Xe apparent diffusion coefficient (ADC) with DL CO in all subjects. D LCO was significantly correlated with 3 He ADC (r = À0.95, r 2 = 0.90, P < 0.0001, Holm-Bonferroni corrected P = 0.003, y = À0.0044x + 0.69), 129 Xe ADC b = 12 (r = À0.95, r 2 = 0.90, P < 0.0001, Holm-Bonferroni corrected P = 0.002, y = À0.00056x + 0.11), 129 Xe ADC b = 20 (r = À0.87, r 2 = 0.76, P = 0.0005, Holm-Bonferroni corrected P = 0.01, y = À0.00042x + 0.088), and 129 Xe ADC b = 30 (r = À0.89, r 2 = 0.78, P = 0.008, Holm-Bonferroni corrected P = 0.10, y = À0.00029x + 0.069). Dotted lines indicate the 95% limits of agreement. Table 3. Correlations for 3 He and 129 Xe apparent diffusion coefficient (ADC) with computed tomography emphysema measurements for chronic obstructive pulmonary disease ex-smokers.
Pearson correlation coefficients r (p/p 1 ) 3 He ADC (cm 2 /sec) 129 Xe ADC b12 (cm 2 /sec) 129 Xe ADC b20 (cm 2 /sec) 129 Xe ADC b30 (cm 2 /sec) the finding that 129 Xe ADC b20 anterior-posterior gradients were also reduced in the COPD ex-smokers with emphysema in comparison to healthy never-smokers suggests that 129 Xe MRI with b = 20 sec/cm 2 also provides a way to evaluate microstructural abnormalities. Third, there was a significant difference in mean ADC in the superior as compared to inferior lung regions between the older never-smokers and the COPD exsmokers with emphysema for 3 He ADC, but not for any 129 Xe ADC. It is important to note that in contrast with the results reported here, previous studies evaluating the 129 Xe ADC SI gradients in older never-smokers demonstrated elevated ADC values in the inferior lung regions (Sindile et al. 2007) or elevated ADC values the superior lung regions (Kaushik et al. 2011). It is clear that because of the lower diffusion coefficient of pure 129 Xe gas, more heterogeneous gas mixing within the lung is likely to occur during image acquisition and this may, in part, explain the discrepancies in these previous reports. However, the 129 Xe/ 4 He gas mixture used in this study may have allowed more homogeneous mixing within the lung resulting in the lower SI gradient seen in the older neversmokers, which is to be expected in healthy volunteers without emphysema. Although it is unclear why elevated ADC was demonstrated in the superior lung regions in comparison to the inferior lung regions for 3 He MRI, but not for 129 Xe MRI in the COPD ex-smokers with emphysema, these differences may be explained by the different lung regions accessed by the gases, as shown in Figure 1, due to the different densities and viscosities of the gas mixtures. In this regard, we previously demonstrated significantly greater ventilation abnormalities for 129 Xe MRI than 3 He MRI in the same COPD ex-smokers (Kirby et al. 2012d). In a regional evaluation of 3 He and 129 Xe ADC, we showed that mean ADC in lung regions accessed by 3 He gas only were significantly greater than in the remaining lung , suggesting that the more emphysematous lung regions filled more readily with 3 He as compared to 129 Xe gas. By extension therefore, if here the 3 He gas more readily filled the centrilobular emphysematous regions within the upper lobe, this may explain the greater SI differences for 3 He MRI. However, it is important to note that based on the estimated self-diffusion coefficients for the 3 He and 129 Xe gas mixtures (0.211 and 0.826 cm 2 /sec for 129 Xe/ 4 He and 3 He/N 2 , respectively; Kirby et al. 2012d), 129 Xe gas is a better estimate of room air, which has an estimated self-diffusion coefficient of 0.218 cm 2 /sec, as compared to the 3 He gas mixture. Taken together this suggests that due to the different physical properties of 129 Xe, the distribution of the 129 Xe gas mixture that is imaged during a breath-hold condition may more accurately reflect the regional distribution of oxygen gas mixed in room air as compared to 3 He gas mixtures used during imaging.
Finally, we demonstrated strong and significant correlations for 3 He ADC, 129 Xe ADC b12 , and ADC b20 with DL CO and CT measurements of emphysema. DW gradients and diffusion time influence the measured ADC, and more importantly, they influence the spatial scale that the gas is probing (Woods et al. 2005). Therefore, 3 He ADC and 129 Xe ADC with different b-values may provide different sensitivities to different lung microstructural environments and abnormalities. We investigated this by evaluating the relationship for DL CO and CT lung density thresholds with 3 He and 129 Xe ADC. We observed strong correlations for 3 He ADC, 129 Xe ADC b12 , and ADC b20 , further suggesting that 129 Xe DW imaging is sensitive to the same lung microstructural abnormalities as 3 He ADC with b value 1.6 cm 2 /sec. However, we must also note that although strong and significant univariate correlations were observed for 3 He and 129 Xe ADC with DL CO , in the multivariate regression model 129 Xe ADC b12 was the only significant predictor of DL CO . In contrast, 3 He ADC was the only significant predictor of RA 950 . Although CT density-threshold measurements correlate with pathology measurements of emphysema on resected lung specimens (Avni et al. 1996), it has been suggested that CT is not sensitive to subclinical or more mild emphysema (Muller et al. 1988;Miller et al. 1989). DL CO , on the other hand, despite low reproducibility, has been shown to be a very sensitive measure of early or mild subclinical emphysematous abnormalities. A recent study demonstrated elevated levels of endothelial microparticles in smokers with abnormally low DL CO but normal spirometry (Gordon et al. 2011). Furthermore, in ex-smokers without COPD, elevated 3 He ADC, worse symptoms and exercise capacity were reported in subjects with abnormal as compared to normal DL CO ). Since 129 Xe gas has a lower diffusivity than 3 He, we hypothesized that 129 Xe MRI with greater diffusion weighting may be more sensitive to early or mild emphysema. The finding that 129 Xe MRI was the only significant predictor of DL CO lends support to this notion. The diffusion time of 5 msec was selected for 129 Xe MRI (Sukstanskii and Yablonskiy 2012;Boudreau et al. 2013) in order to probe a similar spatial length scale as 3 He, and using this diffusion time, the characteristic diffusion lengths were approximately 0.49 and 0.46 mm for 3 He and 129 Xe, respectively. For these diffusion lengths, it is likely that both 3 He and 129 Xe gas atoms diffused beyond a single alveolus (diameter~0.20 mm; Ochs et al. 2004) to a distal respiratory bronchiole (diameter 0.40 mm; Horsfield and Cumming 1968). Although the majority of the COPD subjects evaluated in this study had moderate-to-severe emphysema, future studies, perhaps with improved polarization levels, should investigate the sensitivity of different 129 Xe MRI imaging parameters, including different diffusion weighting and diffusion times, for evaluating emphysematous changes in subjects with more mild disease. Importantly, in our previous investigations we have demonstrated 129 Xe MRI gas distribution measurements were more sensitive to airway wall abnormalities in asthmatics  and to emphysematous bullae in COPD . These findings suggest that due to the different physical properties of 129 Xe gas compared to 3 He gas, 129 Xe may be more sensitive for the detection of airway obstruction and may therefore also be more sensitive for the detection of airway obstruction in patients with early or subclinical disease. However, this will need to be confirmed in future 129 Xe MRI investigations. We also believe that 129 Xe MRI DW imaging may provide increased sensitivity for evaluating emphysematous changes in subjects with more mild disease. In this regard, we must acknowledge that 129 Xe ADC b30 may not provide an appropriate estimate of emphysema in subjects with moderate-to-severe disease. In regions of the lung with significant emphysema, the 129 Xe gas diffusion coefficient is likely strong enough to destroy the MRI signal in that region due to the greater 129 Xe diffusion weighting, and the only signal remaining in the lung will be from the nonemphysematous lung microstructure. Therefore, the measured ADC values will reflect the healthier lung parenchyma, and not the emphysematous regions. However, in very early or mild subclinical disease, this may in fact be advantageous, as any small increase in airspace size will result in a reduction in signal intensity in the DW images. 129 Xe ADC b30 , due to the greater diffusion weighting may therefore provide greater sensitivity to mild microstructural changes than 3 He ADC or 129 Xe ADC b12 , although this cannot be ascertained by the current study.
We recognize that this work was limited by several factors. First, the small number of subjects evaluated and in particular the small number of subjects in the 129 Xe ADC b30 group, certainly limits the generalizability of these results. Another important limitation for both 3 He and 129 Xe ADC measurements is that these cannot be reported for the regions of the lung that are poorly or not ventilated. Such unventilated or poorly ventilated lung regions may be due to small airway occlusion, mucous plugs, airway wall thickening and inflammation, severe emphysema, or bullous disease. Previous studies (Kaushik et al. 2011;Kirby et al. 2012a) comparing 3 He and 129 Xe in COPD suggest that 129 Xe gas distribution is sensitive to emphysematous bullae, and this is another consideration when performing 129 Xe DW imaging in COPD subjects.
In summary, we evaluated 129 Xe ADC b values of 12, 20, and 30 sec/cm 2 in the same older never-smokers and COPD ex-smokers and compared these with 3 He ADC, DL CO , and CT measurements of emphysema. We showed that 3 He ADC and 129 Xe ADC b12 or ADC b20 distinguished subjects with and without COPD. Although strong correlations for both 3 He and 129 Xe ADC were observed with DL CO and CT measurements of emphysema, in a multivariate analysis 129 Xe ADC b12 was the only significant predictor of DL CO . However, 129 Xe ADC b30 may be appropriate for evaluating subclinical or mild emphysema, in this small group of older never-smokers and ex-smokers with moderate-to-severe emphysema, 129 Xe ADC b12 provided a physiologically appropriate estimate of gas exchange abnormalities and alveolar microstructure, providing good evidence to support future 129 Xe MRI studies of emphysema.