Advertisement

A two-center analysis of hyperpolarized 129Xe lung MRI in stable pediatric cystic fibrosis: Potential as a biomarker for multi-site trials

Open AccessPublished:March 25, 2019DOI:https://doi.org/10.1016/j.jcf.2019.03.005

      Highlights

      • Hyperpolarized 129Xe images were retrospectively analyzed for 26 pediatric subjects.
      • 18 CF and 8 healthy pediatric subjects from two institutions were studied.
      • CF subjects from both institutions had a similar LCI and ventilation defect percent.
      • In a combined dataset, MRI results were correlated with LCI and FEV1.

      Abstract

      Background

      The ventilation defect percent (VDP), measured from hyperpolarized (HP) 129Xe magnetic resonance imaging (MRI), is sensitive to functional changes in cystic fibrosis (CF) lung disease. The purpose of this study was to measure and compare VDP from HP 129Xe MRI acquired at two institutions in stable pediatric CF subjects with preserved lung function.

      Methods

      This retrospective analysis included 26 participants from two institutions (18 CF, 8 healthy, age range 10–17). Pulmonary function tests, N2 multiple breath washout (to measure lung clearance index, LCI), and HP 129Xe MRI were performed. VDP measurements were compared between two trained analysts using mean-anchored linear binning. Correlations were investigated for VDP compared to the forced expiratory volume in one second (FEV1) and LCI.

      Results

      VDP measurements agreed for the two analysts with an intraclass correlation coefficient of 0.99. In the combined dataset, VDP measured by Analyst 1 was 5.96 ± 1.82% and 15.96 ± 6.76% for the healthy and CF groups, respectively (p = .0004). Analyst 2 showed similar differences between healthy and CF (p = .0003). VDP measured by either analyst was shown to correlate with FEV1 (R2 = 0.33, p = .003; and R2 = 0.26, p = .009 for Analysts 1 and 2, respectively) and LCI (R2 = 0.76, p < .0001; and R2 = 0.77, p < .0001 for Analysts 1 and 2, respectively).

      Conclusion

      HP 129Xe MRI provides a robust measurement of ventilation heterogeneity in stable pediatric CF subjects at two sites. Since measurements performed at two sites yielded similar VDP values with near-identical values between different analysts, implementation of the technique in multi-center trials in CF appears feasible.

      Keywords

      1. Introduction

      Hyperpolarized (HP) 3He or 129Xe magnetic resonance imaging (MRI) provides a sensitive measure of ventilation heterogeneity in pediatric cystic fibrosis (CF) patients with preserved lung function [
      • Santyr G.
      • Kanhere K.
      • Morgado F.
      • Rayment J.H.
      • Ratjen F.
      • Couch M.J.
      Hyperpolarized gas magnetic resonance imaging of pediatric cystic fibrosis lung disease.
      ,
      • Thomen R.P.
      • Walkup L.L.
      • Roach D.J.
      • Cleveland Z.I.
      • Clancy J.P.
      • Woods J.C.
      Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
      ]. HP gas MRI is typically performed during a breath-hold following inhalation of a dose of HP 3He or 129Xe, and the images reflect the distribution of the inhaled gas inside the lungs. In obstructive lung diseases such as CF, the HP gas does not fully equilibrate with the whole lung in a single breath, resulting in a heterogenous image with signal voids that represent unventilated or poorly ventilated regions of the lung. Early HP gas MRI studies in CF have assessed the ventilation distribution using reader scoring systems [
      • Donnelly L.F.
      • MacFall J.R.
      • McAdams H.P.
      • Majure J.M.
      • Smith J.
      • Frush D.P.
      • et al.
      Cystic fibrosis: combined hyperpolarized 3He-enhanced and conventional proton MR imaging in the lung -preliminary observations.
      ] and counting the ventilation defects [
      • Mentore K.
      • Froh D.K.
      • de Lange E.E.
      • Brookeman J.R.
      • Paget-Brown A.O.
      • Altes T.A.
      Hyperpolarized HHe 3 MRI of the lung in cystic fibrosis: assessment at baseline and after bronchodilator and airway clearance treatment.
      ]. In recent years, HP gas MRI studies in CF have objectively quantified the spatial distribution of ventilation using a ventilation defect percent, or VDP [
      • Thomen R.P.
      • Walkup L.L.
      • Roach D.J.
      • Cleveland Z.I.
      • Clancy J.P.
      • Woods J.C.
      Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
      ,
      • Kanhere N.
      • Couch M.J.
      • Kowalik K.
      • Zanette B.
      • Rayment J.H.
      • Manson D.
      • et al.
      Correlation of lung clearance index with hyperpolarized 129Xe magnetic resonance imaging in pediatric subjects with cystic fibrosis.
      ], which is defined as the fraction of unventilated lung (i.e., the ventilation defect volume divided by the total lung volume, represented as a percentage).
      While previous HP gas MRI studies in CF have used both 3He [
      • Altes T.A.
      • Johnson M.
      • Fidler M.
      • Botfield M.
      • Tustison N.J.
      • Leiva-Salinas C.
      • et al.
      Use of hyperpolarized helium-3 MRI to assess response to ivacaftor treatment in patients with cystic fibrosis.
      ,
      • Kirby M.
      • Svenningsen S.
      • Ahmed H.
      • Wheatley A.
      • Etemad-Rezai R.
      • Paterson N.A.
      • et al.
      Quantitative evaluation of hyperpolarized helium-3 magnetic resonance imaging of lung function variability in cystic fibrosis.
      ,
      • O'Sullivan B.
      • Couch M.
      • Roche J.P.
      • Walvick R.
      • Zheng S.
      • Baker D.
      • et al.
      Assessment of repeatability of hyperpolarized gas MR ventilation functional imaging in cystic fibrosis.
      ] and 129Xe [
      • Thomen R.P.
      • Walkup L.L.
      • Roach D.J.
      • Cleveland Z.I.
      • Clancy J.P.
      • Woods J.C.
      Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
      ,
      • Kanhere N.
      • Couch M.J.
      • Kowalik K.
      • Zanette B.
      • Rayment J.H.
      • Manson D.
      • et al.
      Correlation of lung clearance index with hyperpolarized 129Xe magnetic resonance imaging in pediatric subjects with cystic fibrosis.
      ,
      • Walkup L.L.
      • Thomen R.P.
      • Akinyi T.G.
      • Watters E.
      • Ruppert K.
      • Clancy J.P.
      • et al.
      Feasibility, tolerability and safety of pediatric hyperpolarized 129Xe magnetic resonance imaging in healthy volunteers and children with cystic fibrosis.
      ], future studies will likely focus on using HP 129Xe due to the lower cost and greater availability of the gas. HP 3He and 129Xe VDP measurements have been successfully validated in other obstructive lung diseases, such as chronic obstructive pulmonary disease, with a slightly greater VDP expected from HP 129Xe MRI due to the greater density and lower diffusivity of the inhaled gas compared to 3He [
      • Kirby M.
      • Svenningsen S.
      • Owrangi A.
      • Wheatley A.
      • Farag A.
      • Ouriadov A.
      • et al.
      Hyperpolarized 3He and 129Xe MR imaging in healthy volunteers and patients with chronic obstructive pulmonary disease.
      ]. VDP is potentially more sensitive for detecting functional changes in the lungs compared to pulmonary function test (PFT) indices, such as the forced expiratory volume in one second (FEV1) [
      • Thomen R.P.
      • Walkup L.L.
      • Roach D.J.
      • Cleveland Z.I.
      • Clancy J.P.
      • Woods J.C.
      Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
      ]. VDP has been shown to be reproducible in stable CF lung disease [
      • O'Sullivan B.
      • Couch M.
      • Roche J.P.
      • Walvick R.
      • Zheng S.
      • Baker D.
      • et al.
      Assessment of repeatability of hyperpolarized gas MR ventilation functional imaging in cystic fibrosis.
      ], and sensitive to changes in ventilation distribution that may not be detectable using FEV1 alone [
      • Kirby M.
      • Svenningsen S.
      • Ahmed H.
      • Wheatley A.
      • Etemad-Rezai R.
      • Paterson N.A.
      • et al.
      Quantitative evaluation of hyperpolarized helium-3 magnetic resonance imaging of lung function variability in cystic fibrosis.
      ]. One study in pediatric CF reported an increased VDP in CF patients compared to healthy controls, whereas FEV1 did not differentiate between both groups [
      • Thomen R.P.
      • Walkup L.L.
      • Roach D.J.
      • Cleveland Z.I.
      • Clancy J.P.
      • Woods J.C.
      Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
      ]. In addition, VDP is strongly correlated with multiple breath washout (MBW) measurements of the lung clearance index (LCI), which is a measure of ventilation heterogeneity [
      • Kanhere N.
      • Couch M.J.
      • Kowalik K.
      • Zanette B.
      • Rayment J.H.
      • Manson D.
      • et al.
      Correlation of lung clearance index with hyperpolarized 129Xe magnetic resonance imaging in pediatric subjects with cystic fibrosis.
      ]. Therefore, VDP has the potential to be used as an outcome measure for the management of CF lung disease and testing novel therapeutics, such as CF transmembrane conductance regulator (CFTR) modulators [
      • Altes T.A.
      • Johnson M.
      • Fidler M.
      • Botfield M.
      • Tustison N.J.
      • Leiva-Salinas C.
      • et al.
      Use of hyperpolarized helium-3 MRI to assess response to ivacaftor treatment in patients with cystic fibrosis.
      ].
      Before HP gas measures can be adopted clinically, multi-center prospective clinical trials will need to be performed, which will require a harmonized VDP analysis approach. Therefore, standardized procedures will be required for consistent acquisition and analysis between centers, which will be especially important for sites that may be using different MRI platforms, different RF coil hardware, and different HP 129Xe dosing strategies. As a preliminary step in the pathway to clinical translation, we performed a retrospective analysis of HP 129Xe images from similar stable pediatric CF subjects obtained at two different institutions using slightly different MRI systems to assess the following: (i) the agreement between VDP measurements performed by two trained analysts using the same software, and (ii) the correlation between VDP and PFT measurements (i.e., FEV1 and LCI).

      2. Methods

      2.1 Subjects

      This study was a retrospective analysis that included previously reported image data acquired at the Hospital for Sick Children (Site #1) in Toronto, Ontario, Canada [
      • Kanhere N.
      • Couch M.J.
      • Kowalik K.
      • Zanette B.
      • Rayment J.H.
      • Manson D.
      • et al.
      Correlation of lung clearance index with hyperpolarized 129Xe magnetic resonance imaging in pediatric subjects with cystic fibrosis.
      ] and Cincinnati Children's Hospital Medical Center (Site #2) in Cincinnati, OH, USA [
      • Thomen R.P.
      • Walkup L.L.
      • Roach D.J.
      • Cleveland Z.I.
      • Clancy J.P.
      • Woods J.C.
      Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
      ]. The study at Site #1 was regulated by Health Canada and approved by the Hospital for Sick Children Research Ethics Board (clinicaltrials.gov registry number NCT02740868). The study at Site #2 was regulated by an FDA investigational new drug approval (IND 123,577) and approved by the Cincinnati Children's Institutional Review Board (clinicaltrials.gov registry number NCT02272049). Table 1 summarizes the subject demographics for 16 pediatric participants at Site #1 and 10 pediatric participants at Site #2, all between the ages of 10 and 17. The combined dataset included 18 well-controlled stable CF participants and 8 age-matched healthy controls (p = .73). Well-controlled stable CF was defined as having no changes in respiratory symptoms or medication for four weeks (Site #1) or one week (Site #2) prior to participation in the study.
      Table 1Subject demographics and pulmonary function tests for all participants included in the retrospective study.
      SubjectAgeSexHeight (cm)Weight (kg)FEV1 (% pred.)LCIGenotype
      Site #1 Healthy110F14640.5846.17
      213F15241.3967.17
      311M14532.6946.74
      417M16860.8956.06
      511F13242.61377.11
      614M16352.9129
      Site #1 CF717M17055.37416.87DeltaF508/DeltaF508
      810F14330.61096.83DeltaF508/DeltaF508
      910F14031.2869.10DeltaF508/DeltaF508
      1011F14130.48111.71DeltaF508, Duplications of exons 7–11
      1114M17557.07111.98G480S/A559T
      1211F14431.46115.35C.1653_1655delCTT/c.1837 T > G
      1316F15237.47815.98DeltaF508/DeltaF508
      1416F14843.58310.88W1282X/1282X
      1514F14538.5969.08W1282X/1282X
      1611M15638.2999.56DeltaF508/DeltaF508
      Site #2 Healthy1714M17764.8106
      1812M14134.8103
      Site #2 CF1914M16754.49711.28DeltaF508/DeltaF508
      2012F16243.67715.28DeltaF508/DeltaF508
      2113F16148.9968.49G551D, F508
      2214F15651.21066.73F508, L206 W
      2316M17965.51209.81DeltaF508/DeltaF508
      2411M14234.5102F508, G178R
      2515F15943.772DeltaF508/DeltaF508
      2611F14734.38611.01DeltaF508/DeltaF508

      2.2 Site #1 data acquisition

      The study at Site #1 performed pulmonary function tests (PFTs), nitrogen multiple breath washout (N2 MBW) and HP 129Xe MRI in a single study visit. Spirometry and plethysmography were performed (Vmax, VIASYS CareFusion, San Diego, CA, USA) by qualified research team members according to ATS guidelines [
      • Miller M.R.
      • Hankinson J.
      • Brusasco V.
      • Burgos F.
      • Casaburi R.
      • Coates A.
      • et al.
      Standardisation of spirometry.
      ]. N2 MBW was performed to determine LCI (Exhalyzer D, EcoMedics, Duernten, Switzerland) by qualified research team members, according to ERS/ATS standards [
      • Robinson P.D.
      • Latzin P.
      • Verbanck S.
      • Hall G.L.
      • Horsley A.
      • Gappa M.
      • et al.
      Consensus statement for inert gas washout measurement using multiple- and single- breath tests.
      ]. LCI was defined as the cumulative expired volume required to achieve 1/40th of the initial N2 concentration divided by the functional residual capacity (FRC).
      HP 129Xe MRI was performed in the coronal plane at 1.5 T (HDx, GE Healthcare, Waukesha, WI) using a flexible wrap-around quadrature transmit/receive RF coil tuned to the 129Xe resonance frequency (Clinical MR Solutions, Brookfield, WI, USA). Isotopically enriched 129Xe (83%) was polarized to approximately 15% (Model 9800, Polarean, Durham, NC, USA) and dispensed into a 1 L Tedlar bag (Jensen Inert Products, Coral Springs, FL, USA). The HP 129Xe dose was set to 10% of total lung capacity (TLC), as measured using plethysmography, and balanced to 1 L with medical grade N2. Throughout the imaging session, the subject's oxygen saturation and heart rate were monitored.
      Following inhalation of the dose bag from FRC, HP 129Xe images were acquired in a 16 s breath-hold using a fast spoiled gradient recalled echo (GRE) sequence (see the online supplement for a summary of all imaging parameters used in both studies). Most participants had two repeated HP 129Xe scans within a few minutes of each other. Following HP 129Xe imaging, the 129Xe vest coil was exchanged for an 8-channel torso array coil (GEHC) and conventional 1H lung images were acquired. For 1H imaging, subjects breathed in a 1 L dose bag of N2 from FRC and GRE images were acquired in a 13 s breath-hold.

      2.3 Site #2 data acquisition

      The study at Site #2 used an approach that was similar to the Site #1 study, with some important differences. PFTs, N2 MBW and HP 129Xe MRI were not necessarily performed in a single study visit. Same-day spirometry was performed (Koko, nSpire, Longmont, CO) if clinical PFTs had not been performed in the last six months. Plethysmography was not performed, but instead the participant's height was used to estimate TLC for HP 129Xe dosing. Unlike the Site #1 study, where dose bags used a fixed 1 L volume, the Site #2 study used dose bags that were filled to 1/6th of the participant's TLC. Dose bags were either 100% HP 129Xe, or a mixture of 50% HP 129Xe/ 50% N2, depending on the gas polarization (described below). N2 MBW was performed on the same day if possible, or up to 34 weeks after the MRI study visit (Exhalyzer D, EcoMedics, Duernten, Switzerland).
      HP 129Xe MRI was performed in the axial plane at 3 T (Philips Achieva, Best, Netherlands) using a home-built 129Xe saddle RF coil. 129Xe was polarized using either a commercially-built system (Polarean, Durham, NC, USA) or a home-built polarizer. Dose bags from the commercially-built polarizer were filled with 100% 129Xe, while dose bags from the home-built system were diluted with 50% N2. Either case resulted in a similar signal-to-noise ratio (SNR) since the home-built system provided up to double the polarization of the commercially-built system (~30%). Similar to the Site #1 study, HP 129Xe and conventional 1H acquisitions used coached inhalations from FRC and breath-holds of up to 16 s. Conventional 1H imaging used the integrated 1H body coil without moving the participant or 129Xe coil. The 1H and HP 129Xe image acquisitions at Site #2 both used a GRE pulse sequence, and the parameters are shown in the online supplement.

      2.4 Data analysis

      VDP was measured by two trained individuals using the mean-anchored linear binning technique described by Thomen et al. [
      • Thomen R.P.
      • Walkup L.L.
      • Roach D.J.
      • Cleveland Z.I.
      • Clancy J.P.
      • Woods J.C.
      Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
      ]. Both analysts independently processed the images from all 26 participants. Analyst 1 had one year of experience in lung image analysis, while Analyst 2 had seven years of experience. The 1H images were segmented to create a lung mask using seeded region-growing in 3D Slicer (https://www.slicer.org), and then registered to HP 129Xe images using a landmark-based approach with affine transformations [
      • Santyr G.
      • Kanhere K.
      • Morgado F.
      • Rayment J.H.
      • Ratjen F.
      • Couch M.J.
      Hyperpolarized gas magnetic resonance imaging of pediatric cystic fibrosis lung disease.
      ]. The region-growing algorithm excluded large blood vessels from the thoracic cavity mask, but an explicit vesselness filter was not implemented to remove smaller blood vessels [
      • Tustison N.J.
      • Avants B.B.
      • Flors L.
      • Altes T.A.
      • de Lange E.E.
      • Mugler 3rd, J.P.
      • et al.
      Ventilation-based segmentation of the lungs using hyperpolarized (3)He MRI.
      ,
      • He M.
      • Kaushik S.S.
      • Robertson S.H.
      • Freeman M.S.
      • Virgincar R.S.
      • McAdams H.P.
      • et al.
      Extending semiautomatic ventilation defect analysis for hyperpolarized (129)Xe ventilation MRI.
      ]. If there were any registration errors, due to differences in lung inflation between the 1H and HP 129Xe acquisitions, the analyst performed manual corrections to the thoracic cavity mask.
      After the preliminary segmentation and registration steps, VDP was measured in Matlab (Mathworks, Natick, MA, USA). A bias field correction was used to account for RF coil inhomogeneities, where each HP 129Xe voxel was divided by the mean lung signal value in each dimension (anterior-posterior, left-right, apex-base). The HP 129Xe image volume was divided by the mean signal value inside the lung mask, and voxels with a signal <60% of the mean value were considered part of the defect region. The VDP threshold was set based on a study that included a similar age-matched population of healthy children and pediatric CF patients, where a 60% signal threshold was shown to provide the maximum difference between VDP values measured in healthy and diseased lungs [
      • Thomen R.P.
      • Walkup L.L.
      • Roach D.J.
      • Cleveland Z.I.
      • Clancy J.P.
      • Woods J.C.
      Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
      ]. Before calculating the final VDP, a median filter (3 × 3 kernel) was applied to the defect mask obtained from linear binning [
      • Thomen R.P.
      • Sheshadri A.
      • Quirk J.D.
      • Kozlowski J.
      • Ellison H.D.
      • Szczesniak R.D.
      • et al.
      Regional ventilation changes in severe asthma after bronchial thermoplasty with (3)He MR imaging and CT.
      ].
      The final VDP was calculated as the total volume of unventilated lung obtained from the 129Xe images divided by the total lung volume obtained from the 1H image masks. VDP was compared between analysts using a Bland-Altman analysis in GraphPad Prism (Graphpad Software, La Jolla, CA, USA). The inter-analyst reliability of VDP measurements was assessed by calculating the intraclass correlation coefficient (ICC) in Matlab. HP 129Xe images with a center slice SNR below 8.5 were excluded from the Bland-Altman and correlation analyses. The SNR threshold of 8.5 was determined by the trained analysts, based on the ability to reliably co-register the 1H and 129Xe images in a given subject. VDP was compared between groups using a Mann-Whitney test, with p < .05 being considered significant. VDP was also correlated with FEV1 and LCI using a linear regression analysis in GraphPad Prism. Where possible, LCI was retrospectively analyzed by the MBW over-read center at Site #1. For eight subjects at Site #2, the MBW raw data was lost and the LCI could not be re-processed. In those cases, the MBW raw data was not over-read and the LCI value provided by Site #2 was used.

      3. Results

      Fig. 1 shows a comparison of segmented ventilation and defect maps obtained from HP 129Xe MRI performed by both analysts in two representative CF patients at Site #1 and Site #2, respectively. Both analysts identified qualitatively similar thoracic cavity masks, as well as similar segmented masks and VDP results. The HP 129Xe images included in this study (SNR > 8.5) had a mean SNR of 15.1 ± 3.8 and 18.2 ± 4.7 for Sites 1 and 2, respectively (p = .12). Table 2 summarizes the VDP results for the healthy and CF patient cohorts, separated by institution and analyst. Fig. 2(a) and (b) show the VDP results as box and whisker plots separated by institution and also as a combined dataset. VDP was significantly different between the healthy and CF groups for the Site #1 data and also for the combined dataset (p < .05 for either analyst). There were not enough healthy participants at Site #2 to reach a significant difference for that subset. The CF cohorts from the two sites had VDP values that were not significantly different from one another (p = .44 and p = .46 for Analysts 1 and 2, respectively).
      Fig. 1
      Fig. 1Comparison of segmented ventilation and defect maps obtained from two representative CF patients at Site #1 (top) and Site #2 (bottom). The leftmost images show an overlay of the center slice HP 129Xe image in green, with the 1H image in greyscale. VDP results and segmented maps are shown for both trained analysts, where ventilated areas are shown in green and defect areas are shown in purple.
      Table 2Comparison of subject demographics, pulmonary function tests, and VDP results for all participants included in the retrospective study. Results are shown separately for each institution and also as a combined dataset.
      Site #1Site #2Combined
      HealthyCFp ValueHealthyCFp ValueHealthyCFp Value
      n61028818
      Sex (males)332356
      Age (mean ± SD)12.7 ± 2.613.0 ± 2.70.894513.0 ± 1.413.3 ± 1.8112.7 ± 2.313.1 ± 2.30.7299
      FEV1 (% pred. ± SD)105.8 ± 21.683.8 ± 14.30.0593104.5 ± 2.194.5 ± 15.70.2889105.5 ± 18.385.6 ± 15.50.0636
      LCI (± SD)6.68 ± 0.4711.52 ± 3.120.0017
      LCI was not available for the healthy participants from Site #2.
      10.39 ± 2.77
      LCI was not available for the healthy participants from Site #2.
      7.05 ± 0.9911.10 ± 2.950.0012
      Analyst 1 VDP (% ± SD)6.55 ± 0.7416.38 ± 6.150.00121.80
      Only one healthy individual at Site #2 had HP 129Xe images that were acceptable for analysis (SNR > 8.5).
      13.72 ± 7.25
      Only one healthy individual at Site #2 had HP 129Xe images that were acceptable for analysis (SNR > 8.5).
      5.96 ± 1.8215.96 ± 6.760.0004
      Analyst 2 VDP (% ± SD)5.76 ± 2.0516.56 ± 6.900.00122.81
      Only one healthy individual at Site #2 had HP 129Xe images that were acceptable for analysis (SNR > 8.5).
      13.47 ± 6.58
      Only one healthy individual at Site #2 had HP 129Xe images that were acceptable for analysis (SNR > 8.5).
      5.39 ± 2.1715.48 ± 6.790.0003
      low asterisk LCI was not available for the healthy participants from Site #2.
      low asterisklow asterisk Only one healthy individual at Site #2 had HP 129Xe images that were acceptable for analysis (SNR > 8.5).
      Fig. 2
      Fig. 2Box and whisker plots showing the VDP measured by (a) Analyst 1 and (b) Analyst 2 for healthy and CF cohorts. VDP results are shown separately for each institution and also as a combined cohort. (c) Bland-Altman analysis showing the inter-analyst agreement for all VDP measurements in the combined dataset.
      Fig. 2(c)shows a Bland-Altman analysis for VDP measurements performed by the two analysts. The mean bias (± standard deviation) in VDP between analysts was insignificant (0.14 ± 1.46), and the inter-analyst ICC was 0.99 (p = 0). The average absolute difference in VDP between Site #1 and Site #2 was 3.7 and 3.0 percentage points for Analysts 1 and 2, respectively. VDP measurements obtained by the two analysts were strongly correlated with one another (R2 = 0.96, p < .0001). Fig. 3(a) and (b) show a correlation analysis for FEV1 compared to VDP measured by Analyst 1 and Analyst 2, respectively. For both analysts, there was a weak to moderate negative correlation between VDP and FEV1, when either the Site #1 data or the combined dataset were considered. Fig. 3(c) and (d) show the relationship between VDP and LCI for both analysts, confirming the expected strong positive correlations; however, there was a significant correlation between VDP and LCI only when either the Site #1 data or the combined dataset were considered, underlining the value added by combining data from the two sites.
      Fig. 3
      Fig. 3(a,b) Correlation between VDP and FEV1 for both analysts. (c,d) Correlation between VDP and LCI for both analysts. For each plot (a-d), the solid line represents the correlation for the combined dataset, while the plot title includes the separate correlations for each site.

      4. Discussion

      To our knowledge, this study is the first to compare VDP measurements obtained from HP 129Xe MRI performed in stable pediatric CF subjects at two institutions. In a combined 129Xe image dataset consisting of 28 pediatric participants (8 healthy, 18 CF), good inter-observer agreement was demonstrated for VDP values measured by two trained analysts, and the VDP results were similar for the two populations of mild pediatric CF lung disease (Table 2). VDP was significantly different between the healthy and CF groups in the combined dataset, while data from the individual sites may lack statistical power. Therefore, there is strong potential for the use of HP 129Xe MRI in future multi-center trials.
      This study showed the expected correlations with physiological parameters; VDP showed a weak to moderate correlation with FEV1 and a strong correlation with LCI. Although stronger correlations have been previously reported between VDP and LCI in stable pediatric CF subjects [
      • Kanhere N.
      • Couch M.J.
      • Kowalik K.
      • Zanette B.
      • Rayment J.H.
      • Manson D.
      • et al.
      Correlation of lung clearance index with hyperpolarized 129Xe magnetic resonance imaging in pediatric subjects with cystic fibrosis.
      ], the slightly reduced correlations in the present study likely reflect the variability introduced by the inclusion of multi-center data combined with differences in the data acquisition. Although there was good agreement in VDP between analysts, some small differences in VDP results can likely be attributed to the 1H/129Xe registration and thoracic cavity segmentation. While the VDP calculation was largely automated, the initial segmentation and registration steps required some manual input and adjustments. As the VDP increased, the analysts noted that it became more difficult to find matching landmarks between 1H and HP 129Xe images, especially when the defects were large and encompassed an entire lung lobe. To avoid potential mis-registration issues, future work will explore the simultaneous acquisition of 1H and HP 129Xe lung images in the same breath-hold [
      • Altes T.A.
      • Johnson M.
      • Fidler M.
      • Botfield M.
      • Tustison N.J.
      • Leiva-Salinas C.
      • et al.
      Use of hyperpolarized helium-3 MRI to assess response to ivacaftor treatment in patients with cystic fibrosis.
      ,
      • Wild J.M.
      • Ajraoui S.
      • Deppe M.H.
      • Parnell S.R.
      • Marshall H.
      • Parra-Robles J.
      • et al.
      Synchronous acquisition of hyperpolarised 3He and 1H MR images of the lungs - maximising mutual anatomical and functional information.
      ]. Without the need to register the 1H and HP 129Xe images, the VDP analysis could potentially be fully automated.
      There are a few limitations in this study, which stem from the retrospective nature of the analysis. That is, images were acquired at two institutions that used different hardware, magnetic field strengths, scanner vendors, and RF coil designs. In addition, the two sites used similar pulse sequences, but with slightly different acquisition parameters. Even though the two sites had different configurations, both were able to acquire good quality HP 129Xe images, and the VDP appears to be a robust measure of ventilation heterogeneity in these healthy children and stable pediatric CF subjects. The SNR was expected to be slightly higher at 3 T compared to 1.5 T [
      • Xu X.
      • Norquay G.
      • Parnell S.R.
      • Deppe M.H.
      • Ajraoui S.
      • Hashoian R.
      • et al.
      Hyperpolarized 129Xe gas lung MRI-SNR and T2* comparisons at 1.5 T and 3 T.
      ]; however, the SNR difference in this retrospective study was too small to be statistically significant. Therefore, SNR was not expected to significantly influence the VDP. The original studies at both institutions only acquired low resolution 1H images to facilitate VDP calculation; however, future studies will also acquire high resolution 1H images for morphological scoring. Previous studies have shown that 1H MRI morphological scores are able to detect structural abnormalities in CF [
      • Wielputz M.O.
      • Puderbach M.
      • Kopp-Schneider A.
      • Stahl M.
      • Fritzsching E.
      • Sommerburg O.
      • et al.
      Magnetic resonance imaging detects changes in structure and perfusion, and response to therapy in early cystic fibrosis lung disease.
      ,
      • Roach D.J.
      • Cremillieux Y.
      • Fleck R.J.
      • Brody A.S.
      • Serai S.D.
      • Szczesniak R.D.
      • et al.
      Ultrashort Echo-time magnetic resonance imaging is a sensitive method for the evaluation of early cystic fibrosis lung disease.
      ]. In addition, 1H morphological scores are correlated with LCI and are sensitive to CF treatment response [
      • Stahl M.
      • Wielputz M.O.
      • Graeber S.Y.
      • Joachim C.
      • Sommerburg O.
      • Kauczor H.U.
      • et al.
      Comparison of lung clearance index and magnetic resonance imaging for assessment of lung disease in children with cystic fibrosis.
      ].
      Another limitation of this study was the difference in dosing strategies used by the two institutions. Since the Site #1 cohorts used a 1 L bag volume, the total lung volume for imaging breath-holds (i.e., FRC + 1 L) was likely close to TLC in some of the younger pediatric subjects. On the other hand, the Site #2 cohorts used a smaller bag volume (i.e., 1/6th of TLC), which would have been closer to a comfortable inspiratory tidal volume. It has been shown that the inhaled dose volume affects VDP results, with larger inspiratory volumes leading to a reduced VDP [
      • Smith L.
      • Hughes P.J.C.
      • Marshall H.
      • Collier G.
      • West N.
      • Horsley A.
      • et al.
      Hyperpolarised gas MRI shows a decrease in lung ventilation defects at increased inspiratory lung volumes in cystic fibrosis.
      ]. A larger inhaled volume can fill areas of the lung that would otherwise be obstructed at smaller volumes; however, it has also been shown that both dosing approaches lead to similar correlations between VDP and LCI [
      • Smith L.
      • Hughes P.J.C.
      • Marshall H.
      • Collier G.
      • West N.
      • Horsley A.
      • et al.
      Hyperpolarised gas MRI shows a decrease in lung ventilation defects at increased inspiratory lung volumes in cystic fibrosis.
      ]. Despite the different dosing strategies used at the two sites, VDP appears to provide a robust measure of ventilation heterogeneity. This result might be explained by the use of mean-anchored linear binning to calculate VDP, which uses a fixed signal threshold that includes some partially ventilated areas of the lung. Since Site #1 used a larger dose volume, it is possible that some partially ventilated areas of the lung might have appeared as defect had the dosing used a smaller volume. Nevertheless, the inclusion of partially ventilated regions in mean-anchored linear binning results in a similar VDP for the two CF populations. Future work will continue to investigate the sensitivity of other VDP calculation methods in multi-center trials, such as k-means [
      • Zha W.
      • Niles D.J.
      • Kruger S.J.
      • Dardzinski B.J.
      • Cadman R.V.
      • Mummy D.G.
      • et al.
      Semiautomated ventilation defect quantification in exercise-induced bronchoconstriction using hyperpolarized Helium-3 magnetic resonance imaging: a repeatability study.
      ,
      • Kirby M.
      • Heydarian M.
      • Svenningsen S.
      • Wheatley A.
      • McCormack D.G.
      • Etemad-Rezai R.
      • et al.
      Hyperpolarized 3He magnetic resonance functional imaging semiautomated segmentation.
      ], spatial fuzzy c-means clustering [
      • Hughes P.J.C.
      • Horn F.C.
      • Collier G.J.
      • Biancardi A.
      • Marshall H.
      • Wild J.M.
      Spatial fuzzy c-means thresholding for semiautomated calculation of percentage lung ventilated volume from hyperpolarized gas and 1 H MRI.
      ] or variations of the linear binning approach [
      • He M.
      • Driehuys B.
      • Que L.G.
      • Huang Y.T.
      Using hyperpolarized 129Xe MRI to quantify the pulmonary ventilation distribution.
      ,
      • Collier G.J.
      • Acunzo L.
      • Smith L.J.
      • Hughes P.J.
      • Norquay G.
      • Chan H.
      • et al.
      Linear binning maps for image analysis of pulmonary ventilation with hyperpolarized gas MRI: Transferability and clinical applications.
      ].
      HP 129Xe MRI is a relatively new and sensitive technique, and although more experience is required to fully implement in multi-center clinical trials [
      • Verbanck S.
      • Vanderhelst E.
      The respective roles of lung clearance index and magnetic resonance imaging in the clinical Management of Patients with cystic fibrosis.
      ,
      • Kanhere N.
      • Couch M.J.
      • Rayment J.H.
      • Ratjen F.
      • Santyr G.
      Reply to Verbanck and Vanderhelst: the respective roles of lung clearance index and magnetic resonance imaging in the clinical Management of Patients with cystic fibrosis.
      ], this study presents a step forward in assessing multi-site variability in methods and endpoints. This comparative work demonstrates strong agreement using HP 129Xe MRI between two pediatric CF settings and we conclude that multi-center clinical trials using this approach are feasible. Future multi-center studies will focus on standardizing the image acquisition protocols and parameters. A recent multi-center study has shown that standardizing 1H-based MRI is feasible in pediatric CF [
      • Wielputz M.O.
      • von Stackelberg O.
      • Stahl M.
      • Jobst B.J.
      • Eichinger M.
      • Puderbach M.U.
      • et al.
      Multicentre standardisation of chest MRI as radiation-free outcome measure of lung disease in young children with cystic fibrosis.
      ]. Based on the somewhat limited availability of HP 129Xe MRI technology at the current time, it is likely that 1H-based imaging techniques will see more widespread clinical adoption; however, HP 129Xe MRI is anticipated to be used in specialized centers upon regulatory approval. HP 129Xe MRI will likely play a role in providing outcome measures for interventional trials involving CFTR modulators, and may also play a role in future clinical management via precision medicine. Since HP 129Xe MRI can potentially detect CF lung disease before traditional PFTs [
      • Thomen R.P.
      • Walkup L.L.
      • Roach D.J.
      • Cleveland Z.I.
      • Clancy J.P.
      • Woods J.C.
      Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
      ], its sensitivity may allow for interventional trials with fewer subject numbers than previous studies using FEV1 or LCI as outcome measures [
      • Subbarao P.
      • Stanojevic S.
      • Brown M.
      • Jensen R.
      • Rosenfeld M.
      • Davis S.
      • et al.
      Lung clearance index as an outcome measure for clinical trials in young children with cystic fibrosis. A pilot study using inhaled hypertonic saline.
      ].

      Acknowledgements

      The authors gratefully acknowledge helpful discussions with members of the 129Xe MRI Clinical Trials Consortium. The authors would like to thank the following individuals at SickKids for their help with data collection: Yonni Friedlander, Michelle Klingel, Krzysztof Kowalik, Andras Lindenmaier, Tammy Rayner, Laura Seed, Elaine Stirrat, Ruth Weiss, David Wilson, and Brandon Zanette; in addition to the following at Cincinnati Children's: Laura Walkup, Zackary Cleveland, Erin Watters, and Kelly Thornton. We would also like to thank the following sources of funding: The Hospital for Sick Children (Catalyst Grant from the Cystic Fibrosis Centre), Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery grant (RGPIN 217015-2013), Canadian Institutes of Health Research (CIHR) operating and project grants (MOP 123431, PJT 153099), the Cincinnati Children's Research Foundation, and the National Institutes of Health (T32 HL007752, R01 HL131012). MJC was funded by a Research Training Competition (Restracomp) Fellowship from the Hospital for Sick Children and a Mitacs Elevate Postdoctoral Fellowship.

      Appendix A. Supplementary data

      References

        • Santyr G.
        • Kanhere K.
        • Morgado F.
        • Rayment J.H.
        • Ratjen F.
        • Couch M.J.
        Hyperpolarized gas magnetic resonance imaging of pediatric cystic fibrosis lung disease.
        Acad Radiol. 2019; 26: 344-354https://doi.org/10.1016/j.acra.2018.04.024
        • Thomen R.P.
        • Walkup L.L.
        • Roach D.J.
        • Cleveland Z.I.
        • Clancy J.P.
        • Woods J.C.
        Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients.
        J Cyst Fibros. 2017; 16: 275-282
        • Donnelly L.F.
        • MacFall J.R.
        • McAdams H.P.
        • Majure J.M.
        • Smith J.
        • Frush D.P.
        • et al.
        Cystic fibrosis: combined hyperpolarized 3He-enhanced and conventional proton MR imaging in the lung -preliminary observations.
        Radiology. 1999; 212: 885-889
        • Mentore K.
        • Froh D.K.
        • de Lange E.E.
        • Brookeman J.R.
        • Paget-Brown A.O.
        • Altes T.A.
        Hyperpolarized HHe 3 MRI of the lung in cystic fibrosis: assessment at baseline and after bronchodilator and airway clearance treatment.
        Acad Radiol. 2005; 12: 1423-1429
        • Kanhere N.
        • Couch M.J.
        • Kowalik K.
        • Zanette B.
        • Rayment J.H.
        • Manson D.
        • et al.
        Correlation of lung clearance index with hyperpolarized 129Xe magnetic resonance imaging in pediatric subjects with cystic fibrosis.
        Am J Respir Crit Care Med. 2017; 196: 1073-1075
        • Altes T.A.
        • Johnson M.
        • Fidler M.
        • Botfield M.
        • Tustison N.J.
        • Leiva-Salinas C.
        • et al.
        Use of hyperpolarized helium-3 MRI to assess response to ivacaftor treatment in patients with cystic fibrosis.
        J Cyst Fibros. 2017; 16: 267-274
        • Kirby M.
        • Svenningsen S.
        • Ahmed H.
        • Wheatley A.
        • Etemad-Rezai R.
        • Paterson N.A.
        • et al.
        Quantitative evaluation of hyperpolarized helium-3 magnetic resonance imaging of lung function variability in cystic fibrosis.
        Acad Radiol. 2011; 18: 1006-1013
        • O'Sullivan B.
        • Couch M.
        • Roche J.P.
        • Walvick R.
        • Zheng S.
        • Baker D.
        • et al.
        Assessment of repeatability of hyperpolarized gas MR ventilation functional imaging in cystic fibrosis.
        Acad Radiol. 2014; 21: 1524-1529
        • Walkup L.L.
        • Thomen R.P.
        • Akinyi T.G.
        • Watters E.
        • Ruppert K.
        • Clancy J.P.
        • et al.
        Feasibility, tolerability and safety of pediatric hyperpolarized 129Xe magnetic resonance imaging in healthy volunteers and children with cystic fibrosis.
        Pediatr Radiol. 2016; 46: 1651-1662
        • Kirby M.
        • Svenningsen S.
        • Owrangi A.
        • Wheatley A.
        • Farag A.
        • Ouriadov A.
        • et al.
        Hyperpolarized 3He and 129Xe MR imaging in healthy volunteers and patients with chronic obstructive pulmonary disease.
        Radiology. 2012; 265: 600-610
        • Miller M.R.
        • Hankinson J.
        • Brusasco V.
        • Burgos F.
        • Casaburi R.
        • Coates A.
        • et al.
        Standardisation of spirometry.
        Eur Respir J. 2005; 26: 319-338
        • Robinson P.D.
        • Latzin P.
        • Verbanck S.
        • Hall G.L.
        • Horsley A.
        • Gappa M.
        • et al.
        Consensus statement for inert gas washout measurement using multiple- and single- breath tests.
        Eur Respir J. 2013; 41: 507-522
        • Tustison N.J.
        • Avants B.B.
        • Flors L.
        • Altes T.A.
        • de Lange E.E.
        • Mugler 3rd, J.P.
        • et al.
        Ventilation-based segmentation of the lungs using hyperpolarized (3)He MRI.
        J Magn Reson Imaging. 2011; 34: 831-841
        • He M.
        • Kaushik S.S.
        • Robertson S.H.
        • Freeman M.S.
        • Virgincar R.S.
        • McAdams H.P.
        • et al.
        Extending semiautomatic ventilation defect analysis for hyperpolarized (129)Xe ventilation MRI.
        Acad Radiol. 2014; 21: 1530-1541
        • Thomen R.P.
        • Sheshadri A.
        • Quirk J.D.
        • Kozlowski J.
        • Ellison H.D.
        • Szczesniak R.D.
        • et al.
        Regional ventilation changes in severe asthma after bronchial thermoplasty with (3)He MR imaging and CT.
        Radiology. 2015; 274: 250-259
        • Wild J.M.
        • Ajraoui S.
        • Deppe M.H.
        • Parnell S.R.
        • Marshall H.
        • Parra-Robles J.
        • et al.
        Synchronous acquisition of hyperpolarised 3He and 1H MR images of the lungs - maximising mutual anatomical and functional information.
        NMR Biomed. 2011; 24: 130-134
        • Xu X.
        • Norquay G.
        • Parnell S.R.
        • Deppe M.H.
        • Ajraoui S.
        • Hashoian R.
        • et al.
        Hyperpolarized 129Xe gas lung MRI-SNR and T2* comparisons at 1.5 T and 3 T.
        Magn Reson Med. 2012; 68: 1900-1904
        • Wielputz M.O.
        • Puderbach M.
        • Kopp-Schneider A.
        • Stahl M.
        • Fritzsching E.
        • Sommerburg O.
        • et al.
        Magnetic resonance imaging detects changes in structure and perfusion, and response to therapy in early cystic fibrosis lung disease.
        Am J Respir Crit Care Med. 2014; 189: 956-965
        • Roach D.J.
        • Cremillieux Y.
        • Fleck R.J.
        • Brody A.S.
        • Serai S.D.
        • Szczesniak R.D.
        • et al.
        Ultrashort Echo-time magnetic resonance imaging is a sensitive method for the evaluation of early cystic fibrosis lung disease.
        Ann Am Thorac Soc. 2016; 13: 1923-1931
        • Stahl M.
        • Wielputz M.O.
        • Graeber S.Y.
        • Joachim C.
        • Sommerburg O.
        • Kauczor H.U.
        • et al.
        Comparison of lung clearance index and magnetic resonance imaging for assessment of lung disease in children with cystic fibrosis.
        Am J Respir Crit Care Med. 2017; 195: 349-359
        • Smith L.
        • Hughes P.J.C.
        • Marshall H.
        • Collier G.
        • West N.
        • Horsley A.
        • et al.
        Hyperpolarised gas MRI shows a decrease in lung ventilation defects at increased inspiratory lung volumes in cystic fibrosis.
        in: Proceedings of the 26th annual meeting of ISMRM. 2018 (1083)
        • Zha W.
        • Niles D.J.
        • Kruger S.J.
        • Dardzinski B.J.
        • Cadman R.V.
        • Mummy D.G.
        • et al.
        Semiautomated ventilation defect quantification in exercise-induced bronchoconstriction using hyperpolarized Helium-3 magnetic resonance imaging: a repeatability study.
        Acad Radiol. 2016; 23: 1104-1114
        • Kirby M.
        • Heydarian M.
        • Svenningsen S.
        • Wheatley A.
        • McCormack D.G.
        • Etemad-Rezai R.
        • et al.
        Hyperpolarized 3He magnetic resonance functional imaging semiautomated segmentation.
        Acad Radiol. 2012; 19: 141-152
        • Hughes P.J.C.
        • Horn F.C.
        • Collier G.J.
        • Biancardi A.
        • Marshall H.
        • Wild J.M.
        Spatial fuzzy c-means thresholding for semiautomated calculation of percentage lung ventilated volume from hyperpolarized gas and 1 H MRI.
        J Magn Reson Imaging. 2018; 47: 640-646
        • He M.
        • Driehuys B.
        • Que L.G.
        • Huang Y.T.
        Using hyperpolarized 129Xe MRI to quantify the pulmonary ventilation distribution.
        Acad Radiol. 2016; 23: 1521-1531
        • Collier G.J.
        • Acunzo L.
        • Smith L.J.
        • Hughes P.J.
        • Norquay G.
        • Chan H.
        • et al.
        Linear binning maps for image analysis of pulmonary ventilation with hyperpolarized gas MRI: Transferability and clinical applications.
        in: Proceedings of the 26th annual meeting of ISMRM. 2018 (4482)
        • Verbanck S.
        • Vanderhelst E.
        The respective roles of lung clearance index and magnetic resonance imaging in the clinical Management of Patients with cystic fibrosis.
        Am J Respir Crit Care Med. 2018; 197: 409
        • Kanhere N.
        • Couch M.J.
        • Rayment J.H.
        • Ratjen F.
        • Santyr G.
        Reply to Verbanck and Vanderhelst: the respective roles of lung clearance index and magnetic resonance imaging in the clinical Management of Patients with cystic fibrosis.
        Am J Respir Crit Care Med. 2018; 197: 411-412
        • Wielputz M.O.
        • von Stackelberg O.
        • Stahl M.
        • Jobst B.J.
        • Eichinger M.
        • Puderbach M.U.
        • et al.
        Multicentre standardisation of chest MRI as radiation-free outcome measure of lung disease in young children with cystic fibrosis.
        J Cyst Fibros. 2018; 17: 518-527
        • Subbarao P.
        • Stanojevic S.
        • Brown M.
        • Jensen R.
        • Rosenfeld M.
        • Davis S.
        • et al.
        Lung clearance index as an outcome measure for clinical trials in young children with cystic fibrosis. A pilot study using inhaled hypertonic saline.
        Am J Respir Crit Care Med. 2013; 188: 456-460