Assessment of the influence of lung inflation state on the quantitative parameters derived from hyperpolarized gas lung ventilation MRI in healthy volunteers

Abstract

In this study, the effect of lung volume on quantitative measures of lung ventilation was investigated using MRI with hyperpolarized ³He and ¹²⁹Xe. Six volunteers were imaged with hyperpolarized ³He at five different lung volumes [residual volume (RV), RV + 1 liter (1L), functional residual capacity (FRC), FRC + 1L, and total lung capacity (TLC)], and three were also imaged with hyperpolarized ¹²⁹Xe. Imaging at each of the lung volumes was repeated twice on the same day with corresponding ¹H lung anatomical images. Percent lung ventilated volume (%VV) and variation of signal intensity [heterogeneity score (H_score)] were evaluated. Increased ventilation heterogeneity, quantified by reduced %VV and increased H_score, was observed at lower lung volumes with the least ventilation heterogeneity observed at TLC. For ³He MRI data, the coefficient of variation of %VV was <1.5% and <5.5% for H_score at all lung volumes, while for ¹²⁹Xe data the values were 4 and 10%, respectively. Generally, %VV generated from ¹²⁹Xe images was lower than that seen from ³He images. The good repeatability of ³He %VV found here supports prior publications showing that percent lung-ventilated volume is a robust method for assessing global lung ventilation. The greater ventilation heterogeneity observed at lower lung volumes indicates that there may be partial airway closure in healthy lungs and that lung volume should be carefully considered for reliable longitudinal measurements of %VV and H_score. The results suggest that imaging patients at different lung volumes may help to elucidate obstructive disease pathophysiology and progression.

NEW & NOTEWORTHY We present repeatability data of quantitative metrics of lung function derived from hyperpolarized helium-3, xenon-129, and proton anatomical images acquired at five lung volumes in volunteers. Increased regional ventilation heterogeneity at lower lung inflation levels was observed in the lungs of healthy volunteers.

INTRODUCTION

Hyperpolarized (HP) gas ventilation-weighed magnetic resonance imaging (MRI) allows the visualization of gas distribution within the lung and has been shown to detect early lung disease in patients with cystic fibrosis and normal spirometry (4, 19) and in the lungs of smokers (33). Additionally, it has been used to assess the response to treatment in patients with asthma (13, 29) to longitudinally assess patients with chronic obstructive pulmonary disease (17) and has been shown to be clinically feasible for assessing lung function in children (1).

HP gas and proton anatomical (¹H) lung magnetic resonance imaging (MRI) can be combined to quantify lung ventilation using percent lung-ventilated volume (%VV) or its counterpart the ventilation defect percentage (33), both of which have been widely adopted as simple and robust image-derived metrics. %VV is the ratio of the ventilated lung, defined from HP gas ventilation-weighted images to the thoracic cavity volume, defined from the ¹H anatomical images (16, 33). Previous work has shown improved repeatability of %VV when the anatomical image is acquired in the same breath-hold as the HP gas ventilation-weighted image (12).

Ventilation heterogeneity may be assessed by using the H_score metric developed by Tzeng et al. (31), which calculates the variation of signal intensity in a kernel around a given voxel as the standard deviation divided by the mean (e.g., Fig. 1) (23, 31).

Fig. 1.

Example of local heterogeneity score (H_score) calculation in the residual volume + 1 liter ventilation-weighted image of volunteer 5 (top). The yellow box on the left shows an area of low H_score (enhanced image on the left, with the local area outlined and the voxel that is replaced denoted with an “x”), which is highlighted with the blue box in the H_score map (bottom). The same is shown for a region of high H_score on the right.

The clinical standard for assessing lung volumes is body plethysmography (22, 32), while changes in forced expiratory volume in 1 s (FEV₁) and forced vital capacity, measured using spirometry, are used as clinical markers for lung function decline in certain diseases (22, 32). However, patient coaching of inhalation from a bag of gas rather than spirometric gating is generally used to achieve the lung volumes for HP gas MR imaging, which may lead to variability in lung volumes as will the ability of the patient to inhale the entire contents of the bag of gas being used. The most frequently used lung volume is functional residual capacity plus 1 liter (FRC + 1L) (6–8, 16, 18, 27, 34). However, if a 1-liter bag is inhaled from FRC in smaller patients, this volume may be close to total lung capacity (TLC); thus understanding the effect of lung inflation level on these image-derived metrics is important.

Previous work by Muradyan et al. (23) analyzed the effect of inhalation of HP xenon-129 (¹²⁹Xe) from residual volume (RV) in healthy volunteers and sub-RV in elite divers by acquiring coronal projection images with an in-plane resolution of 4.7 × 9.4 mm. Muradyan et al. calculated the global H_score in the ventilated regions of the image and found that when the elite divers inhaled low volumes of gas (0.9 and 0.4 liters, respectively) compared with larger volumes of gas (1.3 and 0.9 liters, respectively) from sub-RV, increased heterogeneity was seen in the images, consistent with punctate reopening of some airways that were closed at sub-RV. Marshall et al. (20) carried out preliminary work demonstrating the effect of airway opening between FRC + 1L and TLC using HP ³He imaging showing decreased heterogeneity and increased %VV at TLC when compared with FRC + 1L. With these studies demonstrating important mechanisms at work in healthy controls and patients, it is clear that understanding the effect of lung inflation on quantitative metrics derived from HP gas and ¹H anatomical MRI is an important step in moving these techniques forward into standard clinical practice.

Historically, noble gas MRI studies have made use of HP helium-3 (³He); however, with the rising cost and scarcity of ³He, the focus of the pulmonary imaging community is switching to the use of HP ¹²⁹Xe (18, 27) where differences in metrics have been reported due to the differences in diffusivity and the achievable signal of ¹²⁹Xe MRI. Thus the aims of this study were to use both HP ³He and ¹²⁹Xe MRI to 1) assess the effect of different lung inflation levels on the HP gas image-derived metrics %VV and H_score; and 2) assess the repeatability of %VV and H_score from two same-day imaging sessions.

MATERIALS AND METHODS

Subjects.

The study was performed with national research ethics committee approval and with written informed consent from all volunteers. Six volunteers (all men) were recruited for this study with the only criterion being that subjects were suitable for MRI and had no known respiratory complications. Two volunteers were former smokers, two were occasional smokers, and two were never smokers. Table 1 shows the subject demographics.

Table 1.

Subject demographics

Subject	Age, yr	Height, cm	Weight, kg	FEV1	Pack Years
V1	32	183.0	87.0	102.0	0.15
V2	35	184.0	76.0	77.2	0.13
V3	31	182.0	83.0	105.0	0.06
V4	34	185.6	94.0	83.6	0.70
V5	27	189.5	74.0	102.9	0
V6	28	187.6	90.0	99.9	0

V, volunteer; FEV₁, forced expiratory volume in 1 s %predicted.

Study protocol.

Spirometry was performed to international standards (32) to ensure subjects were defined as spirometrically free from respiratory conditions.

All ³He imaging was carried out on a GE HDx 1.5T MRI scanner (GE Healthcare, Milwaukee, WI) using a ³He transmit-receive flexible chest coil (Clinical MR Solutions, Brookfield, WI). ³He was polarized using a commercial polarizer (GE Healthcare, Amersham, UK). HP ³He 3D-balanced steady-state free precession and ¹H spoiled gradient echo images were acquired in the same breath (12) at five different lung volumes: RV, RV + 1 liter (1L), FRC, FRC + 1L, and TLC. For ¹²⁹Xe imaging, the gas was polarized using a home-built polarizer (24) and images were acquired using a ¹²⁹Xe transmit-receive flexible vest coil (Clinical MR Solutions) and the ¹H system body coil at five different lung volumes, as with ³He imaging. ¹²⁹Xe and ¹H images were acquired in separate breath-holds as previously described (27, 28), and this was due to the longer acquisition time of the ¹²⁹Xe scan. Note that only a subset of the volunteers (V2, V3, and V6) were scanned using HP ¹²⁹Xe and separate-breath ¹H imaging as a feasibility study as some participants were no longer available to be scanned. A 1-liter mixture of hyperpolarized gas and nitrogen was used as it is the most commonly used volume in adults (6–8, 16, 18, 27, 34).

For the breathing maneuvers (Fig. 2), volunteers were coached and instructed to breathe within the scanner by a pulmonary physiologist. During imaging, breathing maneuvers started with inhalation of the contents of the 1-liter bag from FRC, except for imaging at RV + 1L where volunteers first exhaled to RV. To acquire images at TLC, volunteers inhaled room air to maximum lung capacity after the inhalation of 1-liter of gas from the bag. For imaging at FRC, volunteers inhaled the contents of the 1-liter bag from FRC and then exhaled back to FRC. For RV imaging, volunteers inhaled the contents of the 1-liter bag from FRC and then exhaled to RV. Gas doses were increased for the exhalation maneuvers and for imaging at TLC with the aim of ensuring sufficient signal for imaging, and before exhalation participants held their breath for 5 s to allow the gas to diffuse into the peripheral lung. Inhaled gas doses are given in Table 2; note that images were also acquired in the order presented in Table 2.

Fig. 2.

Breathing maneuvers and acquisition volumes used in this study. Solid gray lines indicate an inhalation from a 1-liter bag, solid black lines indicate an exhalation and dashed gray lines indicate an inhalation of room air. Solid boxes represent acquisition volumes and dashed boxes represent intermediate volumes as part of the breathing maneuver. RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity.

Table 2.

Gas doses for HP ³He and ¹²⁹Xe acquisitions reported as HP gas dose (N₂)

Acquisition	3He (N2), ml	129Xe (N2), ml
RV	200 (800)	1000 (0)
RV + 1L	150 (850)	750 (250)
FRC	200 (800)	1000 (0)
FRC + 1L	150 (850)	600 (400)
TLC	200 (800)	750 (250)

RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity; N₂, nitrogen; ³He, helium-3; ²⁹Xe, xenon-129; HP, hyperpolarized.

For ³He acquisitions, subjects were scanned twice on the same day, with a 10- to 20-min break (remaining supine within the scanner) in between imaging sessions. ³He imaging sessions lasted 20–30 min on average. For ¹²⁹Xe imaging, subjects were scanned twice on the same day, with a 20- to 40-min break between imaging sessions and were removed from the scanner during this break. ¹²⁹Xe imaging sessions lasted 35–45 min on average, due to limitations imposed by gas polarization time.

Image analysis.

Thoracic cavity volume (TCV) and ventilated volume (VV) were extracted from the ¹H anatomical and HP gas ventilation images, respectively, using the semiautomated segmentation method based on spatial Fuzzy C-means thresholding previously described (15). Percent lung-ventilated volume was calculated according to %VV = (VV/TCV × 100).

Ventilation heterogeneity was assessed using a modified version of the H_score method previously described (31). Images were subsampled from 256 × 256 voxels in-plane to 128 × 128 voxels, resulting in an apparent image resolution of ~3.2 × 3.2 × 5 mm for ³He images or ~3.2 × 3.2 × 10 mm for ¹²⁹Xe images. To avoid partial volume effects at the edge of the ventilation-weighted images, the TCV mask was eroded by one pixel, and the ventilation-weighted image was then multiplied by the VV mask and eroded TCV masks, with voxels outside of the VV and TCV masks being excluded from the local heterogeneity calculation. To generate maps of ventilation heterogeneity, a 3 × 3 voxel kernel (~9 × 9 mm) was then passed over the images, centered on every voxel in the ventilated volume, to calculate the local variation of signal intensity (H_i,j,k at voxel i,j,k). H_score in this work was then defined as the median of the non-zero values of the local heterogeneity map rather than the mean as previously reported, as the histograms of H_score were not normally distributed. For images acquired at TLC, where there was clear signal dropout due to coil sensitivity coverage, VV and TCV masks were matched, i.e., where signal dropout occurred emulating a defect it was manually excluded on both the TCV and VV masks, to ensure that this did not cause increased H_score and decreased %VV.

Additionally, the mean H_score of the most posterior slice was compared with the mean H_score of the remaining image slices for each volunteer at each inflation level for the data acquired with ³He. The mean values were grouped by volunteer and lung volume, and significant differences were assessed using either a paired t-test or Wilcoxon matched-pairs signed rank test depending on the normality of the data. This analysis was not carried out for ¹²⁹Xe data due to the reduced number of subjects.

Repeatability and statistical analysis.

To assess the repeatability of %VV and H_score between session 1 and session 2, the coefficient of variation (CoV), Bland-Altman analysis (2), paired t-tests, and the repeatability limit were used. For CoV analysis, values were grouped by inflation level and session e.g., RV session 1 for all volunteers was compared with RV session 2 for all volunteers. Additionally, to assess repeatability in the image domain, voxel-wise correlation (25) was carried out where each of the six same-inflation intersession image pairs were spatially aligned via deformable image registration (3) to facilitate computation of Spearman correlation coefficients as previously described (30). The repeatability limit was calculated as $1.96\hspace{0.17em}\times \hspace{0.17em}\sqrt{2}{s}_w$, where s_w is the within-subjects standard deviation calculated using SPSS (version 23, IBM) (21).

Spearman’s correlation was also used to assess the relationship between TCV and %VV and H_score along with the relationship between TCV and the absolute change of %VV and H_score over the two imaging sessions. Finally, a two-way repeated measures ANOVA was performed to statistically validate the effect of lung volume on H_score and %VV where within subject factors were defined as the imaging session and lung inflation level, and multiple comparisons were carried out using the Tukey correction. Voxel-wise correlation and two-way repeated measures ANOVA were not carried out for the ¹²⁹Xe data due to the reduced number of subjects scanned.

RESULTS

Comparison of HP ¹²⁹Xe and HP ³He MRI at different inflation levels.

The signal-to-noise ratio (SNR) of the ¹²⁹Xe images was lower than the SNR of the ³He images, particularly at RV, RV + 1L, and FRC, as can be seen in Fig. 3. The RV image of HV3 (¹²⁹Xe, session 2) had complete loss of signal from posterior sections of the lung due to a coil sensitivity issue at the time of the experiment and was thus excluded from analysis. ¹²⁹Xe images had consistently lower %VV (P < 0.0001) and higher H_score (P < 0.0001) when compared with those obtained with HP ³He (Tables 3 and 4).

Fig. 3.

Signal-to-noise ratio (SNR) values from the volunteers scanned with both ³He and ¹²⁹Xe only. RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity.

Table 3.

Mean %VV and median H_score values at each lung volume and S1/S2 over all volunteers derived from hyperpolarized ³He and ¹²⁹Xe

	RV S1	RV S2	RV + 1L S1	RV + 1L S2	FRC S1	FRC S2	FRC + 1L S1	FRC + 1L S2	TLC S1	TLC S2
%VV 3He	95.65	97.17	97.39	97.84	97.30	97.80	98.18	98.05	97.33	97.98
Hscore 3He	10.47	9.98	10.12	10.1	9.37	9.23	9.10	9.20	7.55	7.39
%VV 129Xe	82.94	87.43	92.36	90.86	93.53	94.99	92.99	96.55	95.83	94.98
Hscore 129Xe	15.89	14.92	11.51	12.29	10.83	10.99	11.09	10.53	8.71	8.24

RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity; ³He, helium-3; ²⁹Xe, xenon-129; HP, hyperpolarized; %VV, percent lung ventilated volume; S, session; H_score, local heterogeneity score.

Table 4.

Average %VV and median H_score values over sessions 1 and 2 generated from hyperpolarized ³He and ¹²⁹Xe images for the three volunteers scanned with both gases

	3He V2	129Xe V2	3He V3	129Xe V3	3He V6	129Xe V6
Acquisition, %VV
RV	98.68	96.26	97.03	NA	95.41	88.88
RV + 1L	97.67	90.46	99.49	91.54	98.26	92.84
FRC	98.85	97.53	98.34	94.41	97.85	90.85
FRC + 1L	98.46	96.70	98.76	94.51	98.10	93.11
TLC	99.46	95.58	99.12	93.14	97.82	97.50
Acquisition, Hscore
RV	8.96	12.53	11.63	NA	10.16	16.16
RV + 1L	10.26	12.08	8.82	11.74	9.04	11.89
FRC	8.85	11.01	9.15	10.91	8.90	10.82
FRC + 1L	9.33	10.83	8.61	10.62	8.59	10.98
TLC	6.15	8.10	7.33	8.95	6.50	8.38

RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity; V, volunteer; %VV, percent lung ventilated volume; H_score, local heterogeneity score.

The effect of lung inflation level on %VV and H_score.

The effect of lung inflation level on ³He and ¹²⁹Xe images acquired at different lung volumes is shown in Fig. 4 for volunteer 2. There was a trend toward increased ventilation homogeneity at higher lung volumes, which was seen using both gases. For ³He data, significant differences between H_score were found when TLC was compared to all other lung volumes via the two-way ANOVA (P < 0.0001 for all). No other significant differences in H_score between different inflation levels were found.

Fig. 4.

Representative slices from all acquisition volumes in volunteer 2 (V2) from both ³He and ¹²⁹Xe images. Top row: ³He images; bottom row: ¹²⁹Xe images. RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity.

%VV also varied with lung volume as can be seen from the mean values of %VV and H_score shown in Table 4, which are visualized in Fig. 5. For ³He data, %VV at RV and FRC + 1L were the only volumes that were significantly different from each other when compared using the two-way ANOVA (P = 0.0155). Lung volume had a significant effect on both %VV (P = 0.0265) and H_score (P < 0.0001).

Fig. 5.

Plots of percent lung-ventilated volume (A) from ³He data, heterogeneity score (H_score; %; B) from ³He data, percent lung-ventilated volume from ¹²⁹Xe data (C), and H_score (%; D) from ¹²⁹Xe data at each acquisition volume. Each circle represents a volunteer while the lines represent the mean of the values. RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity; %VV, percent venilated volume.

When considering the ¹H MRI acquired in the same breath as ³He MRI, TCV generated from the ¹H images correlated strongly with H_score (r = −0.75, P < 0.0001) but not with %VV (r = 0.27, P = 0.15). TCV had a weak correlation with the absolute change in %VV (r = −0.39, P = 0.03) but not H_score (r = 0.01, P = 0.53). For the ¹H MRI acquired in a separate breath to the ¹²⁹Xe MRI, TCV had a strong correlation with H_score (r = −0.90, P < 0.0001) and a moderate correlation with the absolute change of H_score over the two sessions (r = −0.66, P = 0.01). TCV had no significant correlation with %VV or the absolute change in %VV over both sessions (r = 0.44, P = 0.12 and r = −0.33, P = 0.25 respectively).

Regardless of the acquisition volume increased H_score was seen in the posterior region of the lung (Fig. 6) with the most posterior slice having a mean ± SD H_score over all volunteers and inflation levels of 15.4 ± 7.1% while all other slices combined had values of 9.8 ± 3.1% when considering ³He data. Additionally, significant differences between the most posterior slice and the remaining slices (Table 5) of the image were seen at RV + 1L and FRC + 1L (P = 0.0087 and P = 0.031, respectively) while no significant difference was seen at RV, FRC, and TLC (P = 0.1562, P = 0.3125, and P = 0.0790 respectively).

Fig. 6.

Representative posterior slices of hyperpolarized (HP) ³He ventilation images and heterogeneity maps at all acquisition volumes from volunteer 2 (V2). The arrows are pointing to areas of decreased ventilation and increased heterogeneity score (H_score). RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity.

Table 5.

Mean H_score (over sessions 1 and 2) at the most posterior slice and all remaining slices for all volunteers at each lung volume for all images acquired using hyperpolarized ³He

	RV		RV + 1L		FRC		FRC + 1L		TLC
Volunteer	Posterior slice	Remaining slices	Posterior slice	Remaining slices	Posterior slice	Remaining slices	Posterior slice	Remaining slices	Posterior slice	Remaining slices
V1	14.95	11.75	21.38	12.11	23.9	12.19	21.13	10.72	17.45	10.3
V2	12.68	9.13	23.97	11.62	15.11	9.30	23.59	10.22	9.28	6.85
V3	8.28	11.15	18.18	8.58	5.59	9.06	10.16	8.65	7.64	7.40
V4	20.23	10.91	17.83	10.24	18.15	10.65	20.46	10.74	25.71	9.39
V5	9.98	10.98	27.03	11.03	5.37	9.42	9.53	9.60	8.29	8.25
V6	15.57	11.00	9.58	9.57	11.53	9.40	13.60	9.02	27.03	7.18

RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; ³He, helium-3; V, volunteer; FRC, functional residual capacity.

Repeatability of %VV and H_score.

Table 6 shows the CoV of %VV and H_score over all six volunteers at each of the lung volumes imaged with ³He and over all three volunteers imaged with ¹²⁹Xe. For ³He data, CoV was <1.5% for %VV and <5.5% for H_score at all lung volumes. Concerning ¹²⁹Xe data, CoV was <4% for %VV and <10% for H_score at all lung volumes.

Table 6.

Coefficient of variation at each inflation level for metrics derived from hyperpolarized ³He and ¹²⁹Xe

	3He				129Xe
Acquisition	TCV	VV	%VV	Hscore	TCV	VV	%VV	Hscore
RV	3.40	3.05	1.29	5.32	3.33	1.34	3.98	9.37
RV + 1L	4.13	4.64	0.63	4.62	2.19	2.26	1.16	7.60
FRC	4.63	4.64	0.87	3.99	5.88	4.80	1.49	2.86
FRC + 1L	3.42	3.42	0.38	2.74	6.88	6.00	3.18	3.74
TLC	1.19	1.00	0.54	5.46	1.97	1.91	0.62	4.62

RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; ³He, helium-3; V, volunteer; FRC, functional residual capacity; TLC, total lung capacity; TLV, total lung volume; VV, ventilated volume; H_score, local heterogeneity score; %VV, percent lung ventilated volume.

Concerning the ³He data, strong intersession voxel-wise correlation was observed for all lung volumes (mean ± SD Spearman coefficients: 0.92 ± 0.03 for RV; 0.94 ± 0.03 for RV + 1L; 0.95 ± 0.02 for FRC; 0.95 ± 0.03 for FRC + 1L; and 0.93 ± 0.02 for TLC).

Bland-Altman bias ± limits of agreement are visualized in Fig. 7 (³He) and Fig. 8 (¹²⁹Xe) for both %VV (Figs. 7A and 8A) and H_score (Figs. 7B and 8B). For ³He data, the limits of agreement were <5% for %VV and <2.5% for H_score. For ¹²⁹Xe data, the limits of agreement were <10% for %VV and <4% for H_score. For ³He MRI, the bias for %VV was <2% at all lung volumes and H_score bias was <1% at all lung volumes, while for ¹²⁹Xe MRI %VV bias was <6% at all lung volumes and H_score bias was <2% at all lung volumes.

Fig. 7.

Bland-Altman plots of percent ventilated volume (%VV; A) and heterogeneity score (H_score; B) generated from images acquired with hyperpolarized ³He at all acquisition volumes. Black dots indicate bias, gray dots are the 95% confidence intervals and the black dashed line is 0. RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity.

Fig. 8.

Bland-Altman plots of ventilated volume (%VV; A) and heterogeneity score (H_score; B)generated from images acquired with hyperpolarized ¹²⁹Xe at all acquisition volumes. Black dots indicate bias, gray dots are the 95% confidence intervals, and the black dashed line is 0. RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity.

Table 7 details the repeatability limit for %VV and H_score from both HP ³He and ¹²⁹Xe images. When considering ³He data %VV repeatability was <3% for all volumes except RV and <2% for all volumes when considering H_score. When considering ¹²⁹Xe data %VV repeatability was <10% for all volumes except RV and <3% for all volumes when considering H_score.

Table 7.

Repeatability limit for %VV and median H_score for images acquired using hyperpolarized ³He and ¹²⁹Xe

Acquisition	3He %VV	3He Hscore	129Xe %VV	129Xe Hscore
RV	5.08	1.80	11.69	3.72
RV + 1L	2.19	1.86	3.06	2.62
FRC	2.90	1.29	4.50	1.07
FRC + 1L	1.39	0.80	9.59	1.45
TLC	2.02	1.35	1.95	1.36

RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; ³He, helium-3; V, volunteer; FRC, functional residual capacity; TLC, total lung capacity; VV, ventilated volume; H_score, local heterogeneity score.

DISCUSSION

The work carried out here has demonstrated that lung volume has a significant bearing on quantitative measurements of lung ventilation derived from both ³He and ¹²⁹Xe MRI. Additionally, from the effect of lung volume on the quantitative metrics of %VV and H_score evident in healthy volunteers, it can be concluded that the lung volume during imaging must be well controlled to ensure that these metrics can be used reliably in longitudinal studies.

Imaging over all volunteers revealed increased ventilation heterogeneity at lower lung volumes, potentially indicating partial airway closure in certain regions of the lung. Increased heterogeneity was particularly observed in the posterior section of the lung at RV + 1L, exemplified by the median H_score per slice plotted against slice number for volunteer 5 in Fig. 9. This increased heterogeneity is likely due to the breathing maneuver used to obtain the images at RV + 1L; that is the volunteers first exhaled to RV, which may have caused some airway closure. In contrast, the HP gas mixture was inhaled from FRC for all other lung volumes, and so the ventilation seen in the RV and FRC images was influenced by the gas distribution within the lungs at FRC + 1L. Note that although increased heterogeneity is seen in the anterior portion of the lung, the increased H_score in those areas is due to the reduced SNR due to decreased gas reaching those areas within the lung.

Fig. 9.

Exemplary plot of H_score from anterior to posterior for volunteer 5 (V5). RV, residual volume; RV + 1L, residual volume plus 1 liter of gas mixture; FRC, functional residual capacity; FRC + 1L, functional residual capacity volume plus 1 liter of gas mixture; TLC, total lung capacity; H_score, heterogeneity score.

The increased ventilation heterogeneity at RV + 1L in volunteers suggests the same underpinning mechanisms as reported in the work by Muradyan et al. (23), where there were distinct focal areas of lung affected by airway closure after inhalation of small gas volumes from below residual volume in elite divers. We hypothesize that the areas of decreased ventilation signal at RV + 1L were caused by airways remaining closed following inhalation of the gas mixture. We believe that this same effect was not observed at RV in the current study since the maneuver to RV required first inhaling to FRC + 1L, such that gas would remain in the areas opened by this first inhalation maneuver even if the airways were to close later on. The areas of reduced ventilation in lungs of the elite divers following inhalation from sub-RV levels observed in the work by Muradyan et al. (23) were larger than those seen here in these volunteers, while they did not see the same heterogeneity seen here in their volunteers following inhalation from RV. One possible reason for this is the improvements in the image resolution for ¹²⁹Xe when compared with their experiments that were carried out with two-dimensional projection imaging, thus providing us with better spatial sampling of regional heterogeneity.

Imaging after smaller inspirations from RV would be interesting to assess at which point the ventilation heterogeneity would return to a distribution closer to that seen at FRC or FRC + 1L. In this case, it would be expected that the smaller the volume inhaled from RV, the greater the ventilation heterogeneity would be although the feasibility of these experiments would be limited by the volume of HP gas required for sufficient image SNR if carrying the experiment out with ¹²⁹Xe. Another factor that may contribute to increased H_score at RV when compared with FRC + 1L and FRC is the increased ratio of blood vessel volume to lung volume at RV, resulting in increased H_score.

The small CoV of %VV between sessions further confirms the growing body of evidence that %VV is a robust global metric of lung ventilation (5, 12, 17), and the high interscan repeatability makes %VV (or ventilation defect percentage) a good candidate metric for longitudinal assessment of lung function in patients (17). The proportionally larger CoV of the H_score suggests that this measure of global ventilation image heterogeneity may be less repeatable.

The generally lower SNR of ¹²⁹Xe images when compared with ³He images is a well-known phenomenon and follows previous publications (14, 28), with ¹²⁹Xe acquisitions having a mean ± SD SNR of 30 ± 13 compared with the 42 ± 15 of the ³He acquisitions. Consequently, the higher H_score seen in the ¹²⁹Xe images when compared with images acquired with HP ³He is at least partially due to the lower SNR and thus increased heterogeneity of signal within ventilated regions. The lower %VV values measured from ¹²⁹Xe images compared with ³He images may be due to the lower diffusivity of ¹²⁹Xe compared with ³He and are consistent with %VV values reported previously in healthy volunteers, patients with chronic obstructive pulmonary disease, and patients with lung cancer who were imaged with both gases (18, 27). Furthermore, lower SNR in one of the ¹²⁹Xe acquisitions (V6, RV, and session 1) caused an increase in H_score showing that the maneuvers or gas doses need to be optimized for the ¹²⁹Xe imaging acquisitions if this methodology is applied to patient cohorts. The need to register the anatomical images to the ventilation images for ¹²⁹Xe %VV calculation will also contribute to the lower repeatability of ¹²⁹Xe %VV when compared with ³He %VV (12), where anatomical images were acquired in the same breath-hold. Additionally, due to imaging constraints, HP ¹²⁹Xe images were acquired with double the slice thickness (10 mm) of the HP ³He images (5 mm); thus differences would be expected due to different inherent physical properties and image acquisition considerations of the respective gas.

Imaging patients with HP gas at different lung volumes may provide a clearer picture of the nature of lung disease. For example, in patients with obstructive lung disease, following deep inhalation to TLC, the effect of increased positive pressure within the airways may result in a reduced H_score and increased %VV due to opening of obstructed airways (20). Additionally, as patients with chronic respiratory disease may have increased closing volumes, imaging at expiration may identify areas of gas trapping similar to those observed by Holmes et al. (9–11).

An increased number of healthy volunteers with a larger age range and inclusion of female subjects would extend this preliminary work into the effect of lung volume on ventilation heterogeneity in healthy volunteers. Additionally, mitigating the signal dropout seen in TLC images with larger coil coverage is an important consideration for future studies. The smoking history of four of the six volunteers (2 former smokers and 2 occasional smokers) means that these data may not represent the ventilation patterns seen in a group of healthy never smokers. However, the number of pack years reported by the volunteers scanned was low (<0.7), and in a previous ³He MRI study of pulmonary ventilation (26), three of the smokers would have been classified as never smokers (<0.5 pack yr). However, the volunteers scanned were spirometrically defined as free from respiratory disease and not unrepresentative of the general population in terms of smoking history. The fact that increased ventilation heterogeneity at lower lung inflation levels was seen in the two never smokers as well as those with a smoking history suggests this effect is not due to smoking related obstructive airways disease.

Conclusions.

Increased ventilation heterogeneity was observed in HP gas images acquired at lower lung volumes in healthy volunteers. This work has shown that although total lung volume and VV may vary considerably between repeated scans there was little effect on %VV in these healthy volunteers. This indicates it may be important to image patients over a range of lung volumes with different breathing maneuvers to fully understand disease progression and accurately characterize ventilation defects and pulmonary mechanics. Finally, the variation in lung volume must be considered when monitoring patients longitudinally with hyperpolarized gas MRI particularly in the cases of disease with a reversible nature such as asthma.

GRANTS

We thank the University of Sheffield, Health Education England and National Institute for Health Research Grant RP-R3-12-027 and Integrated Clinical Academic-Clinical Doctoral Research Fellowship 2015-01-027, Sheffield Hospitals Charity Grant 161731, and Medical Research Council Grant MR/M008894/1 for funding this research.

DISCLOSURES

P. J. Hughes was partially funded by GlaxoSmithKline (STU100037614). None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

P.J.C.H. and J.M.W. conceived and designed research; P.J.C.H., L.S., H.-F.C., G.N., and G.J.C. performed experiments; P.J.C.H. and B.A.T. analyzed data; P.J.C.H., L.S., H.M., and J.M.W. interpreted results of experiments; P.J.C.H. prepared figures; P.J.C.H. drafted manuscript; P.J.C.H., L.S., H.-F.C., B.A.T., G.N., G.J.C., A.M.B., H.M., and J.M.W. edited and revised manuscript; P.J.C.H., L.S., H.-F.C., B.A.T., G.N., G.J.C., A.M.B., H.M., and J.M.W. approved final version of manuscript.