Skip to main content
NCBI home page
Wiley Open Access Collection logoLink to Wiley Open Access Collection
. 2024 Sep 16;93(2):775–787. doi: 10.1002/mrm.30299

Comparison of 3D UTE free‐breathing lung MRI with hyperpolarized 129Xe MRI in pediatric cystic fibrosis

Samal Munidasa 1,2,✉, Brandon Zanette 1, Marie‐Pier Dumas 3, Wallace Wee 3, Sharon Braganza 1, Daniel Li 1, Felix Ratjen 1,3, Giles Santyr 1,2
PMCID: PMC11604841  PMID: 39285622

Abstract

Purpose

To compare phase‐resolved functional lung (PREFUL) regional ventilation derived from a free breathing 3D UTE radial MRI acquisition to hyperpolarized 129Xe‐MRI (Xe‐MRI), conventional 2D multi‐slice PREFUL MRI, and pulmonary function tests in pediatric cystic fibrosis (CF) lung disease.

Methods

Free‐breathing 3D UTE and 2D multi‐slice 1H MRI as well as Xe‐MRI were acquired in 12 stable pediatric CF patients. Using PREFUL, regional ventilation (RVent) maps were calculated from the free‐breathing data. Ventilation defect percentage (VDP) was determined from 3D and 2D RVent maps (2D VDPRVent and 3D VDPRVent, respectively) and Xe‐MRI ventilation (VDPXe). VDP was calculated for the whole lung and for eight regions based on left/right, anterior/posterior, and superior/inferior divisions of the lung. Global and regional VDP was compared between the three methods using Bland–Altman analysis, linear mixed model‐based correlation, and one‐way analysis of variance and multiple comparisons tests.

Results

Global 3D VDPRVent, VDPXe, and 2D VDPRVent were all strongly correlated (all R 2 > 0.62, p < 0.0001) and showed minimal, non‐significant bias (all <2%, p > 0.05). Three dimensional and 2D VDPRVent significantly correlated to VDPXe in most of the separate lung regions (R 2 = 0.18–0.74, p < 0.04), but showed lower inter‐agreement. The superior/anterior lung regions showed the least agreement between all three methods (all p > 0.12).

Conclusion

Absolute VDP assessed by 3D UTE PREFUL MRI showed good global agreement with Xe‐MRI and 2D multi‐slice PREFUL MRI in pediatric CF lung disease. Therefore, 3D UTE PREFUL MRI offers a sensitive and potentially more accessible alternative to Xe‐MRI for regional volumetric evaluation of ventilation.

Keywords: 3D free‐breathing 1‐H imaging, cystic fibrosis, hyperpolarized 129Xenon imaging, pulmonary ventilation, ultra‐short echo time

1. INTRODUCTION

Cystic fibrosis (CF), caused by the dysfunction of the CF transmembrane regulator (CFTR) gene, is characterized by the build‐up of mucous in the airways leading to chronic lung infection and inflammation. 1 Pulmonary MRI provides a non‐invasive, radiation‐free imaging modality for evaluating structural and functional lung abnormalities in pediatric CF 2 , 3 and is useful for the longitudinal monitoring of disease progression and response to CFTR modulatory therapy. 4 , 5 Hyperpolarized 129Xenon gas MRI (Xe‐MRI) provides regional pulmonary ventilation defect percent (VDP) measures and has been shown to detect early CF lung disease, 6 capture response to antibiotic treatment for a pulmonary exacerbation, 7 and track improvements in ventilation following therapy with the triple combination of CFTR modulators: elexacaftor, tezacaftor, and ivacaftor (ETI). 8 However, the coached breathing maneuvers required for Xe‐MRI can be difficult to perform for young and/or very sick patients. Furthermore, the additional equipment required for Xe‐MRI (e.g., polarizer, multi‐nuclear coil, gas) and trained personnel may hinder clinical translation.

Free breathing 1H MRI techniques offer a contrast‐agent free methodology of assessing pulmonary ventilation and perfusion by measuring local changes in proton density following cardiorespiratory motions. These techniques include Fourier decomposition (FD) MRI, 9 self‐gated non‐contrast‐enhanced functional lung (SENCEFUL) MRI, 10 matrix pencil decomposition (MP) MRI, 11 and phase‐resolved functional lung (PREFUL) MRI. 12 Ventilation‐weighted PREFUL MRI, in particular, is performed by sorting lung MR images according to the motion of the diaphragm to interpolate high temporally resolved respiratory phases, which can be used to assess lung ventilation dynamics (e.g., flow‐volume loops). 12 , 13 VDP derived from 2D single‐slice PREFUL MRI has been shown to correlate well with Xe‐MRI, 14 be a responsive measure of pulmonary exacerbation treatment, 15 and demonstrates good repeatability in pediatric CF lung disease. 16 Two‐dimensional multi‐slice PREFUL MRI, performed by acquiring four to five separate coronal slices of the lung, has also shown good correlation with Xe‐MRI in healthy volunteers and chronic obstructive pulmonary disease (COPD) and CF patients. 13 However, because of the incomplete sampling of the lung with limited number of slices and/or slice gaps, ventilation defects may be missed using the 2D approach and the limited through‐plane resolution can potentially reduce detection accuracy of CF‐related pathologies perpendicular to the plane of interest, motivating the need for 3D PREFUL approaches.

More recently, a pseudo‐3D stack‐of‐stars PREFUL MRI method was developed to interrogate whole‐lung ventilation with high spatial resolution (2 mm3 × 2 mm3 × 2 mm3). 17 The 3D PREFUL MRI method has been shown to correlate well with spirometry, generate repeatable PREFUL MRI ventilation maps in COPD and healthy adults, 18 and was shown to track ventilatory improvement in CF patients at baseline and 8–16 weeks after ETI therapy. 19 3D approaches may benefit from the inclusion of UTE acquisitions to improve signal‐to‐noise in the lung tissue, where T2* is short. UTE MRI can provide high‐resolution 3D images with isotropic spatial resolution and has been used to assess pathology in CF lung disease. 20 PREFUL with a 2D UTE sequence has been shown to be feasible in healthy adults, 21 whereas 3D UTE‐SENCEFUL MRI has been developed using a center‐out radial trajectory, which has been shown to produce 3D ventilation maps with higher SNR as compared to the 2D approach. 21 However, a 3D‐UTE free breathing MRI approach has not been explored in pediatric CF patients, particularly those undergoing CFTR modulator treatment, and has not been directly compared to Xe‐MRI. Although conventional PREFUL MRI and Xe‐MRI have been compared in pediatric CF lung disease, previous PREFUL MRI analysis was limited to a single slice. We hypothesize that the proposed 3D UTE PREFUL MRI technique will provide a whole lung ventilation assessment, which more closely agrees with multi‐slice Xe‐MRI.

The objective of this study was to determine the feasibility of obtaining regional ventilation maps and corresponding VDP using 3D UTE PREFUL in pediatric CF lung disease and compare the method to multi‐slice Xe‐MRI and PREFUL MRI, as well as pulmonary function tests (PFTs).

2. METHODS

2.1. Subjects and pulmonary function tests

Twelve stable pediatric CF patients performed pulmonary function tests (spirometry and nitrogen gas multiple‐breath washout [MBW]) and underwent same‐day imaging using a clinical 3 T scanner (MAGNETOM Prismafit, Siemens Healthcare). Five participants were imaged 3 to 12 months after their first visit. The participants were recruited through a protocol approved by The Hospital for Sick Children (REB 1000063021; Clinical‐Trials.gov NCT04391322) and Health Canada as part of an ongoing, multi‐site study assessing the response of Xe‐MRI and pulmonary function changes to CFTR‐modulator treatment in pediatric CF patients (HyPOINT‐Part 2: REB 1000077493; Clinical‐Trials.gov NCT04259970). Written informed assent/consent was obtained from the legal guardian or parent of each participant before examination.

Spirometry 22 (Vmax, VIASYS CareFusion) as well as N2 MBW 23 (Exhalyzer D, EcoMedics AG) were performed by trained professionals according to American Thoracic Society guidelines. The percent predicted forced expiratory volume in 1 s (FEV1% predicted) and the lung clearance index (LCI) were determined from spirometry and N2 MBW, respectively.

2.2. MRI

2.2.1. Xe‐MRI

HP 129Xe images were acquired with a flexible 129Xe vest coil (Clinical MR Solutions), using isotopically enriched 129Xe (85%) polarized to ˜25% (Model 9820, Polarean). The HP 129Xe dose was set to 1/10th of the participant's total lung capacity (TLC) (estimated using patient height and sex) and balanced with medical‐grade N2 to a total volume of 1/6th of the participant's TLC. The participant was coached to inhale the dose bag from functional residual capacity (FRC) and, using a fast spoiled gradient recalled echo (GRE) sequence, 10 to 14 coronal slices were acquired during a 6‐ to 8‐s breath‐hold. As previously described, 16 the following scan parameters were used: TR/TE = 7.8 ms/2.78 ms, flip angle = 10°, FOV = 360 × 384 mm2, matrix = 180 × 192, slice thickness = 15 mm, and bandwidth (BW) = 170 Hz/pixel, 62.5% with partial echo and centrically ordered phase ordering. Volume‐matched anatomical 1H images were acquired with a 3D‐FLASH sequence using a combination of a torso array and spine coil (Siemens Healthcare) as previously reported. 16

Following Xe‐MRI acquisition, a N4ITK bias‐field correction was applied to the images in 3D Slicer (http://www.slicer.org) 24 and was registered to the corresponding 1H images. Using MATLAB, a thoracic cavity mask was generated from anatomical 1H MR images using a 2D U‐net based deep learning segmentation approach. 25 A signal intensity threshold of 60% of the mean signal was used to determine the defect region within the masked 129Xe images and VDP was determined as the total defect volume as a percentage of the total lung volume. 26

2.2.2. Free breathing MRI

The 3D free breathing acquisition was performed using a 3D UTE sequence with a 20 μs non‐selective hard pulse and a center‐out 3D golden‐means radial “koosh‐ball” trajectory 27 with ramped sampling. As described above, 1H MRI images were acquired using a torso array and spine coil for a total of 30 receiving coil elements. The acquisition was performed with the following scan parameters: TR/TE = 1.92 ms/ 0.05 ms, FOV = 500 × 500 × 500 mm3, BW = 830 Hz/pixel, and flip angle = 3°. A total of 280 000 radial spokes were acquired for a total acquisition time of 8 min and 58 s. Figure 1 summarizes the post‐processing pipeline for the 3D UTE PREFUL MRI analysis. For each participant, the average signal magnitude of the first five readout points (i.e., direct current [DC], signal) from a single coil element closest to the diaphragm was used to determine a respiratory waveform. The respiratory waveform was temporally filtered using a Savitzky–Golay function. 21 To address the drift in the waveform's amplitude, the DC signal was rescaled from 0 to 1 using the minimum and maximum values within consecutive 5000‐spoke increments. A binning window was run across the DC signal to reconstruct 30 respiratory phases from expiration to inspiration using ˜21 000 spokes per state. Approximately 30% of the spokes were shared between neighboring states to increase apparent SNR by reducing under sampling artifacts. The weighting of these shared spokes at the periphery of the binning window was determined using a decaying exponential function as described by Jiang et al. 28

FIGURE 1.

FIGURE 1

Open in a new tab

MATLAB‐based post processing pipeline for the 3D UTE phase‐resolved functional lung (PREFUL) MRI analysis. In brief, (A) the direct current (DC) signal of a single coil element closest to the diaphragm was used to sort the radial spokes into 30 respiratory phases. (B) Each respiratory phase was reconstructed with ˜21 000 spokes. (C) Each image was registered to the mid‐respiration state, spatially fitlering was applied, and the thoracic cavity was segmented . (D) Following the PREFUL algorithm, 15 equidistant phases were interpolated using non‐parametric regression. The regional ventilation (RVent) of each voxel within the thoracic cavity mask was determined and (E) a corresponding ventilation defect percentage (VDP) map was calculated using k‐means clustering.

Each respiratory phase was reconstructed to a resolution of 2 × 2 × 2 mm3 by applying iterative SENSE with the gpuNUFTT toolbox in MATLAB. 29 Here, coil sensitivity maps were estimated from non‐SENSE reconstructed 3D images by applying adaptive coil combination. Regularization was applied along the spatial dimension using 3D total variation (λ1 = 0.01) and 3D Symlet 4 wavelet (λ2 = 0.01) regularization terms. Each image was registered to the mid‐respiratory phase using a polynomial non‐rigid registration using the Forsberg Registration Toolbox. 30 The thoracic cavity was segmented using a 3D seeded region‐growing algorithm and a 3D image‐guided filter was applied. Following the PREFUL algorithm, 12 , 17 15 equidistant phases were interpolated using non‐parametric regression and the regional ventilation (RVent) of each voxel within the thoracic cavity mask was determined. As a comparator, 2D PREFUL analysis was performed on five separately acquired 2D coronal slices. The first slice was positioned slightly posterior to the heart, one slice was posterior to the first slice, whereas the remaining three slices were spaced equidistantly toward the anterior direction. The following acquisition parameters were used: TR/TE = 0.99 ms/0.73 ms, flip angle = 5°, FOV = 500 × 500 mm2, matrix = 128 × 128 interpolated to 256 × 256, slice thickness = 15 mm, BW = 1500 Hz/pixel, phase partial Fourier = 6/8. A total of 512 repetitions of the slice were acquired in ˜61 s using a TurboFLASH sequence for a total average duration of ˜5 min and PREFUL analysis was performed in MATLAB (R2021a, The MathWorks) as previously described. 16 For both the 2D and 3D UTE PREFUL MRI analysis, RVent was calculated within the thoracic cavity mask pixel‐wise according to Klimes et al. 17 Two dimensional and 3D k‐means clustering approaches 16 , 31 were used to determined 2D and 3D VDPRVent, respectively.

2.3. Statistical analysis

All statistical analyses were performed using MATLAB. Patient demographics, 3D VDPRVent, 2D VDPRVent, VDPXe, FEV1 (% pred.), forced vital capacity (FVC) (% pred.), FEV1/FVC (% pred.), and LCI data were reported as a median and range. Bland–Altman analysis was used to assess agreement between VDP metrics determined by Xe‐MRI and 2D or 3D UTE PREFUL MRI and significant differences, between and within subjects, were compared using a one‐way analysis of variance with a multiple comparison test using a Bonferroni correction. A linear mixed‐effects model was used to correlate the VDP between three MRI methods using the R statistical software 32 to separate the fixed effects from the random effects (i.e., the within‐subject correlation). A marginal coefficient of determination, R 2 , was determined to measure the correlation strength solely because of the fixed effects and corresponding p‐values were determined using Satterthwaite approximation. 33 , 34 Spatial overlap of VDPXe with 2D or 3D UTE PREFUL MRI‐derived VDPRVent was evaluated using a Dice similarity coefficient (DSC) to determine spatial agreement of the healthy ventilatory and defect regions detected by the two imaging modalities. The 3D UTE PREFUL MRI defect maps were averaged in the anterior–posterior direction to match the out‐of‐plane resolution of the Xe‐MRI defect maps and a landmark‐based method was used to register slice‐matched defect maps before DSC calculation. For comparisons between 2D PREFUL MRI and Xe‐MRI, slices were visually matched and registered to compute DSC values.

In addition to the calculation of DSC, a regional comparison of ventilation defects was investigated by determining VDP separately at eight different regions of the lung. First, the left and right lungs were separated. To separate the superior (upper) and inferior (lower) region of the lung, the axial position approximating the position of the carina was used as the separation point. Similarly, the sagittal position best approximating the position of the carina was used to separate the anterior and posterior lung. Although the position of the carina was discerned directly from the corresponding 1H MR images used to generate the 3D UTE and 2D multi‐slice PREFUL MRI regional ventilation maps, the anatomically matched 1H MR images obtained for Xe‐MRI analysis were used to divide the Xe‐MRI ventilation maps. Correlation and significant differences between regional VDP of the three methods were determined similar to global VDP as described above.

A p‐value <0.05 was considered statistically significant for all tests.

3. RESULTS

MRI and pulmonary function tests were feasible for all CF participants (n = 12; 7 male: 5 female). Table 1 summarizes the patient demographics, pulmonary function measures (FEV1 and LCI), and MRI outcomes for the pediatric CF patients. None of the VDP metrics within the cohort of CF patients were significantly different from each other (p > 0.05). Figure 2 shows the ventilation defect maps derived from 3D UTE PREFUL MRI from two representative CF patients. The CF patient with a 3D VDPRVent of 2.34% showed correspondingly low VDPXe (4.79%). The CF patient with a 3D VDPRVent of 14.8% (where ventilation defects are situated in the right lung) showed high VDPXe (16.3%). Figure 3 shows the ventilation defect maps derived from 3D UTE PREFUL MRI, Xe‐MRI, and 2D PREFUL MRI from a representative CF patient in which similar disease severity and regional ventilation defects can be observed using all three methods. Bland–Altman analysis of the VDPXe, 2D VDPRVent, and 3D VDPRVent are shown in Figure 4. Correlation plots between VDPXe and 2D and 3D VDPRVent across all three time points are shown in Figure 5. All correlations were statistically significant (all R 2 > 0.62, p < 0.0005). LCI significantly correlated with 3D VDPRVent (R 2 = 0.14, p = 0.03) and VDPXe (R 2 = 0.30, p = 0.02), but not 2D VDPRVent (p = 0.20). 3D VDPRVent, 2D VDPRVent, and VDPXe did not significantly correlate with FEV1, FVC, or FEV1/FVC (all R 2 < 0.06 and p > 0.08).

TABLE 1.

Summary of participant demographics, pulmonary function measures, and Xe‐MRI and PREFUL MRI‐derived parameters in all pediatric cystic fibrosis patients.

Characteristic Value
No. of participants 12
Total no. of visits 17
Age 16 (12–19)
FEV1 (% predicted) 93 (66–120)
FVC (% predicted) 100 (74–121)
FEV1/FVC (% predicted) 94 (77–100)
LCI 6.4 (4.6–9.2)
VDPXe (%) 4.8 (2.6–19)
3D VDPRVent (%) 5 (1.3–23)
2D VDPRVent (%) 4.2 (1.3–19)
Open in a new tab

Note: Values are reported as median (range).

Abbreviations: FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; LCI, lung clearance index; PREFUL, phase‐resolved functional lung; RVent, regional ventilation; VDP, ventilation defect percentage; Xe, hyperpolarized 129Xe‐MRI.

FIGURE 2.

FIGURE 2

Open in a new tab

Ventilation defect maps derived from 3D UTE phase‐resolved functional lung (PREFUL) MRI regional ventilation (RVent) distributions for two representative pediatric CF patients with (A) low 3D RVent VDP (3D VDPRVent) and (B) high 3D VDPRVent. Hy 129Xe‐MRI derived VDP (VDPXe) and 2D PREFUL MRI RVent VDP (2D VDPRVent) are shown for each CF patient. CF, cystic fibrosis; VDP, ventilation defect percentage.

FIGURE 3.

FIGURE 3

Open in a new tab

Normalized and segmented, gray‐scaled ventilation maps and corresponding defect maps derived from (A) 3D UTE phase‐resolved functional lung (PREFUL) MRI (B) 129Xe‐MRI (Xe‐MRI) (C) and 2D PREFUL MRI for a representative pediatric CF patient. Similar ventilation defects can be found at the inferior region of both lungs and at the superior region of the right lung toward the posterior slices. Dice coefficients for healthy and defect regions, respectively, are: 0.75 and 0.24 for 3D VDPRVent and VDPXe, 0.77 and 0.16 for 3D VDPRVent and 2D VDPRVent, and 0.75 and 0.21 for 2D VDPRVent and VDPXe. CF, cystic fibrosis; VDP, ventilation defect percentage.

FIGURE 4.

FIGURE 4

Open in a new tab

Bland–Altman analysis of (A) 3D UTE phase‐resolved functional lung (PREFUL) MRI regional ventilation (RVent) VDP (3D VDPRVent) and 129Xe‐MRI (Xe‐MRI) VDP (VDPXe), (B) 2D PREFUL MRI RVent VDP (2D VDPRVent) and VDPXe, and (C) 3D VDPRVent and 2D VDPRvent in pediatric CF patients. Each subject is represented by a unique color. The solid black line and dashed black lines represent the mean bias and the 95% confidence interval, respectively. A mean bias of 1.14% (−5.47, 7.76) was found between 3D VDPRVent and VDPXe, and − 0.84% (−5.77, 4.07) between 2D VDPRVent and VDPXe. The mean bias between 3D VDPRVent and 2D VDPRVent was 1.99% (−4.86, 8.84). CF, cystic fibrosis; VDP, ventilation defect percentage.

FIGURE 5.

FIGURE 5

Open in a new tab

Correlation plots between (A) 3D UTE phase‐resolved functional lung (PREFUL) MRI regional ventilation (RVent) VDP (3D VDPRVent) and 129Xe‐MRI (Xe‐MRI) VDP (VDPXe), (B) 2D PREFUL MRI RVent VDP (2D VDPRVent) and VDPXe, and (C) 3D VDPRVent and 2D VDPRVent in pediatric CF patients. Each subject is represented by a unique color. Significant correlations are considered to be p < 0.05. CF, cystic fibrosis; VDP, ventilation defect percentage.

Figure 6 shows the overlapping ventilation defect maps derived from both Xe‐MRI and 3D UTE PREFUL MRI in a representative subject. Moderate spatial correspondence of the VDPXe and 3D VDPRVent is shown in this patient. Overall, the median DSC values of the healthy ventilatory regions of all three methods were within 0.66 to 0.80, whereas the median DSC values of the defect regions were low, ranging from 0 to 0.05, where the highest DSC was 0.29. No pair of methods showed significantly better spatial overlap (p > 0.05). The correlations between 3D VDPRVent, VDPXe, and VDPRVent determined at specific regions of the lung are shown in Table 2, and the median VDP at these regions are shown in Table 3. Overall, a majority of the regional VDP values were strongly correlated between the three methods, where the least correspondence was found at the superior, anterior region of both the left and right lungs. Notably, 3D VDPRVent was significantly lower at the superior, anterior region of the right lung, whereas 2D VDPRVent was significantly lower in the same regions of both the right and left lung.

FIGURE 6.

FIGURE 6

Open in a new tab

The overlap of the 129Xe‐MRI (Xe‐MRI) and 3D phase‐resolved functional lung (PREFUL) MRI derived ventilation defects are shown for all lung slices (anterior/posterior) in a representative pediatric CF patient, with 3D VDPRVent = 22.9% and VDPXe = 19.2% (purple regions = both 3D VDPRVent and VDPXe red regions = only 3D VDPRVent blue regions = only VDPXe and green regions = neither VDPXe nor 3D VDPRVent). The Dice similarity coefficient (DSC) values of the healthy and defect regions are 0.75 and 0.24, respectively. CF, cystic fibrosis; VDP, ventilation defect percentage.

TABLE 2.

Correlation of 3D VDPRVent, 2D VDPRVent, and VDPXe per region of the lung.

Lung region 3D VDPRVent vs. VDPXe 2D VDPRVent vs. VDPXe 3D VDPRVent vs. 2D VDPRVent
Right/superior/anterior R 2  = 0.05, p = 0.32 R 2  = 0.14, p = 0.12 R 2  = 0.02, p = 0.44
Left/superior/anterior R 2  = 0, p = 0.94 R 2  = 0, p = 0.88 R 2  = 0.14, p = 0.08
Right/inferior/anterior R 2  = 0.66, p < 0.0001 R 2  = 0.42, p  = 0.004 R 2  = 0.01, p = 0.43
Left/inferior/anterior R 2  = 0.23, p  = 0.04 R 2  = 0.74, p < 0.0001 R 2  = 0.02, p = 0.40
Right/superior/posterior R 2  = 0.47, p  = 0.003 R 2  = 0.40, p  = 0.007 R 2  = 0.43, p  = 0.006
Left/superior/posterior R 2  = 0.34, p  = 0.01 R 2  = 0.28, p  = 0.02 R 2  = 0.27, p  = 0.03
Right/inferior/posterior R 2  = 0.18, p = 0.02 R 2  = 0.57, p  = 0.0009 R 2  = 0.58, p  = 0.0004
Left/inferior/posterior R 2  = 0.12, p = 0.16 R 2  = 0.65, p  = 0.0001 R 2  = 0.01, p = 0.78
Open in a new tab

Note: Significant correlations (p < 0.05) are bold.

Abbreviations: RVent, regional ventilation; VDP, ventilation defect percentage; Xe, 129Xe‐MRI.

TABLE 3.

3D VDPRVent, 2D VDPRVent, and VDPXe per region of the lung.

Lung region 3D VDPRVent (%) VDPXe (%) 2D VDPRVent (%) 3D VDPRVent vs. VDPXe (p‐value) 2D VDPRVent vs. VDPXe (p‐value) 3D VDPRVent vs. 2D VDPRVent (p‐value)
Right/superior/anterior 2.3 (0–22) 10 (2.9–28) 0.98 (0–15) 0.017 0.0003 0.52
Left/superior/anterior 2.4 (0–27) 10 (0.9–29) 1 (0–6.4) 0.63 0.0027 0.083
Right/inferior/anterior 7.5 (0.84–30) 3.8 (1.3–12) 2.8 (0–38) 0.23 0.97 1
Left/inferior/anterior 3.1 (0.42–39) 2.7 (0.39–14) 2.1 (0–30) 1 1 1
Right/superior/posterior 5.7 (0.11–54) 8.8 (2–57) 10 (1.8–31) 1 1 1
Left/superior/posterior 4.2 (0.2–32) 5.7 (1.3–31) 6.2 (1.4–17) 1 1 1
Right/inferior/posterior 4.8 (1.2–32) 3 (0.33–23) 2.3 (0–26) 0.37 1 0.24
Left/inferior/posterior 3.5 (0.74–23) 3.7 (0.55–13) 4.4 (0.56–19) 1 1 1
Open in a new tab

Note: Values are reported as median (range). Significant differences (p < 0.05) are indicated in bold.

Abbreviation: RVent, regional ventilation; VDP, ventilation defect percentage; Xe, 129Xe‐MRI.

4. DISCUSSION

This study demonstrates the feasibility of applying the PREFUL method with a 3D golden‐means UTE MR acquisition in pediatric CF patients. Three dimensional PREFUL MRI captures regional ventilation with a higher spatial resolution (2 × 2 × 2 mm3) and lung coverage as compared to 2D PREFUL MRI (2 × 2 × 15 mm3) for a marginal increase in scan time. The main finding of the study was global VDP derived from the proposed 3D UTE PREFUL MRI method (3D VDPRVent) correlated strongly with multi‐slice Xe‐MRI derived VDP (VDPXe) similar to the conventional 2D multi‐slice PREFUL method, whereas regional VDP showed some instances of discordance.

In the pediatric CF participants, 3D VDPRVent showed a comparable absolute bias and correlation strength to VDPXe as compared to 2D multi‐slice VDPRVent. Here, the strong correlation between 2D multi‐slice VDPRVent and VDPXe was comparable to a previous study in which 2D single‐slice VDPRVent correlated to VDPXe in a different cohort of pediatric CF patients. 16 In contrast to the work of Couch et al., 14 which showed a mean bias of 9.0% ± 11.5% between single‐slice PREFUL VDP and Xe‐MRI VDP in a cohort of stable pediatric CF patients and CF patients undergoing a pulmonary exacerbation (PEx), the bias between 3D and 2D multi‐slice PREFUL VDP were 0.84% and 1.14%, respectively. Although the reduction in bias between the two techniques may be attributed to the inclusion of multiple slices when performing PREFUL MRI, the differences in VDP quantification may also be a contributing factor to this discrepancy. Specifically, Couch et al. 14 used k‐means clustering to determine VDP for both Xe‐MRI and PREFUL MRI ventilation distributions. Whereas, in the present study, a threshold of 60% of the mean signal intensity was used to determine VDP from Xe‐MRI, although k‐means clustering was used to quantify ventilation defects in PREFUL MRI. In a previous study comparing 2D multi‐slice PREFUL MRI to Xe‐MRI in adult COPD and CF patients, and healthy volunteers, the VDP of the two methods was found to significantly correlate and moderate spatial overlap was found. 13 In that study, a linear binning method (which uses a healthy cohort's average ventilation distribution as reference) as well as a hard threshold of 0.4 × 90th percentile of the ventilation signal were used to quantify ventilation defects for both Xe‐MRI and PREFUL MRI and, in both cases, no significant biases were found. Although the absolute quantification of VDP derived from the three MRI methods correlated strongly, the median regional overlap of ventilation defects, as measured by the DSC, was low. The low spatial overlap of defect regions can likely be attributed to a combination of the fundamental differences of the MRI methods and the misregistration of small defect volumes. The findings are consistent with the low DSC values (<0.05) of defect regions found from PREFUL MRI and Xe‐MRI previously reported in healthy and stable CF patients, and moderate overlap (˜0.20) in CF patients undergoing a PEx. 14 In comparison, Kaireit et al. 13 reported median DSC values between 2D multi‐slice PREFUL MRI derived VDPRVent and VDPXe of ˜0.34 in a cohort of CF and mostly COPD patients, where disease severity was higher. Nevertheless, spatial agreement of healthy regions was much higher in agreement with both these past studies. 13 , 14

To circumvent the difficulty in aligning ventilation defects between the three methods, a regional comparison of VDP was performed by anatomically separating the lung into eight regions based on the right/left, superior/inferior, and anterior/posterior positions. Using this analysis method, 3D VDPRVent, and 2D VDPRVent showed strong correlation to VDPXe in a majority of the separate lung regions. However, there were less regions showing significant correlation between 3D and 2D VDPRVent, which could be attributed to the larger discrepancy in the acquired spatial resolution and total lung volume of the two methods as well as differences in their estimation of ventilation. For instance, the 3D PREFUL method captures changes in lung density in both the in‐plane and through‐plane direction whereas the 2D method only captures the former. Additionally, the determination of the respiratory phases using the 2D PREFUL is easily captured by tracking the motion of the diaphragm across coronal slices, whereas reconstruction of respiratory phases from the DC signal is less objective and more sensitive to noise.

No correlations were found between the three MRI methods for the superior, anterior position of both lungs. A possible reason for this finding is that this region of the lung, as defined in the methods, is the lowest percentage of the total lung volume and, therefore, the appearance or lack of ventilation defects could disproportionally increase or decrease the estimated VDP values, respectively. Although, in this region of the lung, a higher median VDPXe was found as compared 3D and 2D VDPRVent. Therefore, the inherent differences of the three methods are likely a large contributing factor to the lack of correlation in certain regions, and some spatial discordance between PREFUL and Xe‐MRI has been previously reported, although overall defect burdens are largely comparable. Future studies investigating these different imaging methods will help to understand this variability. Additionally, incorporation of structural imaging may help to determine underlying sources of any potential discordance.

In this study, Xe‐MRI was used as a reference for assessing regional pulmonary ventilation in pediatric CF lung disease to compare to VDP derived from 3D UTE PREFUL MRI. Although Xe‐MRI and PREFUL MRI moderately correlate in pediatric CF lung disease 14 , 15 , 16 and other pulmonary diseases such as COPD, 12 the techniques differ in their ability to measure lung function. First, Xe‐MRI is acquired during a breath‐hold at a fixed lung volume of approximately FRC + 1/6th of TLC, whereas PREFUL MRI is conducted during free breathing during tidal respiration. Furthermore, Xe‐MRI quantifies the level of gas mixing within the lungs, regardless of the deformation of the lung parenchymal tissue, and detects ventilation abnormalities associated with slow filling or blocked airways. In contrast, PREFUL MRI assesses the overall alteration in lung tissue signal (i.e., density) throughout the respiratory cycle, potentially capturing both ventilation‐related changes and structural and mechanical changes (e.g., compliance) in the lungs. Therefore, the effect of tissue compression because of gravity at the dependent lung (i.e., the posterior lung in a supine subject) can have a larger impact on PREFUL MRI, which is performed during tidal breathing, as compared to Xe‐MRI acquired during a single‐breath. In a previous study, these effects were posited as a possible source of higher inter‐visit variability with PREFUL MRI. 16 Future studies of 3D PREFUL MRI will include evaluating the repeatability and longitudinal changes of the proposed method in pediatric CF lung disease compared to of 2D PREFUL MRI.

In this study, neither the VDP from Xe‐MRI or PREFUL MRI significantly correlated with FEV1/FVC and FEV1. This finding contrasts with a previous study on a separate cohort of stable pediatric CF participants in which single slice VDPRVent correlated weakly with FEV1 (R 2 = 0.12, p = 0.02), whereas 2D multi‐slice VDPXe correlated moderately with FEV1 (R 2 = 0.19, p = 0.004). 16 Possible reasons for the lack of correlation include the low sample size and the low variability of disease severity within the CF cohort under investigation, compared to recent studies in similar populations. Particularly, of the 17 patient visits, only three showed a VDPXe >10% and four showed a 3D VDPRVent >10%, which is attributable to the effect of ETI treatment on these CF patients. Notably, one CF patient reported an FEV1 of 120% predicted, whereas 3D VDPRVent, 2D VDPRVent, and VDPXe were measured to be 18.5%, 14.7%, and 10.6%, respectively. These findings are in agreement with past work, which showed that MRI may be more sensitive than spirometry to CF lung disease. 6 However, 3D VDPRVent and VDPXe, but not 2D VDPRVent, significantly correlated with LCI, where VDPXe showed the strongest correlation to LCI as compared to both PREFUL derived VDP measures. In previous studies of pediatric CF lung disease, VDPXe also showed a stronger correlation to LCI than single‐slice PREFUL MRI. 14 , 15 Therefore, the lowered correlation strength of PREFUL‐derived VDP measures with LCI may not be attributed to the incomplete coverage of the lung by PREFUL MRI, but rather by the differences in which the method assesses pulmonary ventilation as described above.

Overall, the 3D UTE PREFUL MRI method feasibly provided a whole lung assessment of pulmonary ventilation. In comparison to Xe‐MRI, the 3D PREFUL MRI does not require any special equipment (e.g., a polarizer or 129Xe coil) and the ability to implement these advanced imaging sequences and reconstruction techniques is easier and more affordable at all sites with 1H MR scanning capabilities. Additionally, the PREFUL algorithm is performed on MR images acquired during quiescent tidal breathing and does not require a prolonged breath‐hold. Breath‐holds may be difficult with CF children who have severe lung disease, particularly for those undergoing a pulmonary exacerbation where instances of gas exhalation during an acquisition have been shown. 7 Image quality can suffer from the aliasing of the exhaled 129Xe gas outside of the area of the lung and may influence the quantification of ventilation defects at different sections of the lung. Although PREFUL MRI does require several minutes of relaxed tidal breathing, instances of coughing may be detected using the DC signal and the corresponding data can be rejected from reconstruction in order maintain image quality, thereby making this technique less susceptible to motion corruption. In particular, PREFUL MRI may allow for functional lung imaging in fragile populations such as neonates. Conventional PREFUL MRI has shown feasibility in neonates, imaged within their first week of birth, 35 including in premature neonates with bronchopulmonary dysplasia. 36 Although the 3D PREFUL MRI method requires more complex sequence design and reconstruction as compared to the 2D method, the improvement in spatial resolution outweighs these disadvantages, most of which, can be easily automated. Additionally, to provide full coverage of the lung, 2D PREFUL would likely require 10 to 15 min depending on the size of the patient. Therefore, the 3D PREFUL technique described in this work may allow for higher‐resolution and faster functional lung imaging than previously described.

Previous studies have shown the feasibility of 3D and UTE PREFUL MRI techniques, but not the combination of the two. A pseudo‐3D, stack‐of‐stars based PREFUL MRI method was proposed by Klimes et al. 17 to provide whole lung ventilation assessment. In comparison to the proposed 3D UTE method, the stack‐of‐stars method could be performed by reconstructing 20 to 60 respiratory states. The 3D PREFUL MRI method strongly correlated with the 2D multi‐slice method, but showed higher sensitivity to ventilation heterogeneity in adult patients with pulmonary diseases. 17 Ventilation parameters derived from 3D PREFUL MRI was also shown to be repeatable in healthy volunteers and COPD patients, 18 correlate well with spirometry, as well as track improvements in lung function in CF patients beginning CFTR‐modulator treatment. 19 In another work, PREFUL MRI was also shown to be feasible with a 2D radial UTE (0.07 ms) approach, using an in‐plane resolution of 2.8 mm2 and slice thickness of 12.5 mm, in a healthy volunteer. 21 The authors found that the areas of low ventilation were greater in the UTE acquisition as compared to that of the standard GRE sequence, but were in similar spatial locations.

Although the PREFUL MRI technique showed promising findings in this study, there were limitations. First, the spatial resolution of the 2D multi‐slice 129Xe images (2 mm × 2 mm × 15 mm) were different from the 3D UTE images (2 × 2 × 2 mm3) used for the PREFUL analysis. Xe‐MRI methods that offer isotropic resolution by using non‐Cartesian trajectories (e.g., spiral/radial) 37 , 38 could serve as a better comparator for the 3D PREFUL regional ventilation maps. Additionally, a landmark‐based registration was used to match 3D PREFUL MRI and Xe‐MRI defect maps for spatial correlation, which may lead to inaccurate matching of defect regions. Furthermore, image distortion corrections caused by non‐linearity gradient were not implemented, which could potentially improve the UTE MR image quality. Last, perfusion‐weighted 3D PREFUL MRI was not explored in this study because of the elimination of the blood in‐flow signal of the non‐selective hard RF pulse. Three dimensional lung perfusion has shown some feasibility in healthy volunteers using a volume‐selective RF pulse, 39 but has not yet been implemented with a PREFUL‐based analysis method. Improvements to the 3D PREFUL MRI approach may include increased undersampling of k‐space to reduce scan time, 17 thereby reducing acquisitions to ˜4 min, which is of particular importance in children or infants who are prone to spontaneous movement. Additionally, to reduce processing time and reader bias, deep learning‐based segmentation method 40 could be implemented in place of the semi‐automated seeded region growing method used in this study.

5. CONCLUSION

Whole lung ventilation assessment using a 3D free breathing UTE MRI acquisition with the PREFUL algorithm is feasible and shows good global agreement with multi‐slice hyperpolarized 129Xe MRI in pediatric CF lung disease and conventional 2D PREFUL MRI on a slice‐matched basis. Regional discordance of ventilation defects were noted and may be expected from the inherent differences of the techniques. Overall, PREFUL MRI provides a more accessible approach compared to Xe‐MRI, which requires an inhaled contrast agent and special equipment, for assessing regional lung function in CF lung disease and may be more advantageous for imaging very sick and young children (e.g., neonates).

Munidasa S, Zanette B, Dumas M‐P, et al. Comparison of 3D UTE free‐breathing lung MRI with hyperpolarized 129Xe MRI in pediatric cystic fibrosis. Magn Reson Med. 2025;93:775‐787. doi: 10.1002/mrm.30299

REFERENCES

  • 1. Cutting GR. Cystic fibrosis genetics: from molecular understanding to clinical application. Nat Rev Genet. 2015;16:45‐56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Dournes G, Walkup LL, Benlala I, et al. The clinical use of lung MRI in cystic fibrosis: what, now, how? Chest. 2021;159:2205‐2217. doi: 10.1016/j.chest.2020.12.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Zanette B, Greer M‐LC, Moraes TJ, Ratjen F, Santyr G. The argument for utilising magnetic resonance imaging as a tool for monitoring lung structure and function in pediatric patients. Expert Rev Respir Med. 2023;17:527‐538. [DOI] [PubMed] [Google Scholar]
  • 4. Graeber SY, Renz DM, Stahl M, et al. Effects of Elexacaftor/Tezacaftor/Ivacaftor therapy on lung clearance index and magnetic resonance imaging in patients with cystic fibrosis and one or two F508del alleles. Am J Respir Crit Care Med. 2022;206:311‐320. doi: 10.1164/rccm.202201-0219OC [DOI] [PubMed] [Google Scholar]
  • 5. Streibel C, Willers CC, Pusterla O, et al. Effects of elexacaftor/tezacaftor/ivacaftor therapy in children with cystic fibrosis ‐ a comprehensive assessment using lung clearance index, spirometry, and functional and structural lung MRI. J Cyst Fibros. 2023;22:615‐622. doi: 10.1016/j.jcf.2022.12.012 [DOI] [PubMed] [Google Scholar]
  • 6. Santyr G, Kanhere N, Morgado F, Rayment JH, Ratjen F, Couch MJ. Hyperpolarized gas magnetic resonance imaging of pediatric cystic fibrosis lung disease. Acad Radiol. 2019;26:344‐354. doi: 10.1016/j.acra.2018.04.024 [DOI] [PubMed] [Google Scholar]
  • 7. Rayment JH, Couch MJ, McDonald N, et al. Hyperpolarised 129Xe magnetic resonance imaging to monitor treatment response in children with cystic fibrosis. Eur Respir J. 2019;53:1802188. doi: 10.1183/13993003.02188-2018 [DOI] [PubMed] [Google Scholar]
  • 8. Walkup L, Roach D, Zanette B, et al. 165 hyperpolarized xenon gas ventilation magnetic resonance imaging detects lung ventilation improvement in children with cystic fibrosis treated with elexacaftor/tezacaftor/ivacaftor in a multisite study. J Cyst Fibros. 2022;21S2:S1‐S378. [Google Scholar]
  • 9. Bauman G, Puderbach M, Deimling M, et al. Non‐contrast‐enhanced perfusion and ventilation assessment of the human lung by means of fourier decomposition in proton MRI. Magn Reson Med. 2009;62:656‐664. [DOI] [PubMed] [Google Scholar]
  • 10. Fischer A, Weick S, Ritter CO, et al. SElf‐gated non‐contrast‐enhanced FUnctional lung imaging (SENCEFUL) using a quasi‐random fast low‐angle shot (FLASH) sequence and proton MRI. NMR Biomed. 2014;27:907‐917. doi: 10.1002/nbm.3134 [DOI] [PubMed] [Google Scholar]
  • 11. Bauman G, Bieri O. Matrix pencil decomposition of time‐resolved proton MRI for robust and improved assessment of pulmonary ventilation and perfusion. Magn Reson Med. 2017;77:336‐342. [DOI] [PubMed] [Google Scholar]
  • 12. Voskrebenzev A, Gutberlet M, Klimeš F, et al. Feasibility of quantitative regional ventilation and perfusion mapping with phase‐resolved functional lung (PREFUL) MRI in healthy volunteers and COPD, CTEPH, and CF patients. Magn Reson Med. 2018;79:2306‐2314. doi: 10.1002/mrm.26893 [DOI] [PubMed] [Google Scholar]
  • 13. Kaireit TF, Kern A, Voskrebenzev A, et al. Flow volume loop and regional ventilation assessment using phase‐resolved functional lung (PREFUL) MRI: comparison with 129 xenon ventilation MRI and lung function testing. J Magn Reson Imaging. 2021;53:1092‐1105. [DOI] [PubMed] [Google Scholar]
  • 14. Couch MJ, Munidasa S, Rayment JH, et al. Comparison of functional free breathing pulmonary 1H and hyperpolarized 129Xe magnetic resonance imaging in pediatric cystic fibrosis. Acad Radiol. 2021;28:e209‐e218. doi: 10.1016/j.acra.2020.05.008 [DOI] [PubMed] [Google Scholar]
  • 15. Munidasa S, Couch MJ, Rayment JH, et al. Free breathing MRI for monitoring ventilation changes following antibiotic treatment of pulmonary exacerbations in paediatric cystic fibrosis. Eur Respir J. 2021;57:2003104. doi: 10.1183/13993003.03104-2020 [DOI] [PubMed] [Google Scholar]
  • 16. Munidasa S, Zanette B, Couch M, et al. Inter‐ and intravisit repeatability of free breathing MRI in pediatric cystic fibrosis lung disease. Magn Reson Med. 2023;89:2048‐2061. doi: 10.1002/mrm.29566 [DOI] [PubMed] [Google Scholar]
  • 17. Klimeš F, Voskrebenzev A, Gutberlet M, et al. 3D phase‐resolved functional lung ventilation MR imaging in healthy volunteers and patients with chronic pulmonary disease. Magn Reson Med. 2021;85:912‐925. doi: 10.1002/mrm.28482 [DOI] [PubMed] [Google Scholar]
  • 18. Klimeš F, Voskrebenzev A, Gutberlet M, et al. Repeatability of dynamic 3D phase‐resolved functional lung (PREFUL) ventilation MR imaging in patients with chronic obstructive pulmonary disease and healthy volunteers. J Magn Reson Imaging. 2021;54:618‐629. doi: 10.1002/jmri.27543 [DOI] [PubMed] [Google Scholar]
  • 19. Klimeš F, Voskrebenzev A, Gutberlet M, et al. Effect of CFTR modulator therapy with elexacaftor/tezacaftor/ivacaftor on pulmonary ventilation derived by 3D phase‐resolved functional lung MRI in cystic fibrosis patients. Eur Radiol. 2023;34:80‐89. doi: 10.1007/s00330-023-09912-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Roach DJ, Crémillieux Y, Fleck RJ, 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. doi: 10.1513/AnnalsATS.201603-203OC [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Behrendt L, Gutberlet M, Voskrebenzev A, et al. Feasibility of phase resolved functional lung imaging (PREFUL) with ultrashort echo time sequence (PUTE). Proc. Intl. Soc. Mag. Reson. Med. 2022;30:1402. [Google Scholar]
  • 22. Miller MR, Hankinson J, Brusasco V, et al. Standardisation of spirometry. Eur Respir J. 2005;26:319‐338. [DOI] [PubMed] [Google Scholar]
  • 23. Robinson PD, Latzin P, Verbanck S, et al. Consensus statement for inert gas washout measurement using multiple‐ and single‐ breath tests. Eur Respir J. 2013;41:507‐522. [DOI] [PubMed] [Google Scholar]
  • 24. Fedorov A, Beichel R, Kalpathy‐Cramer J, et al. 3D slicer as an image computing platform for the quantitative imaging network. Magn Reson Imaging. 2012;30:1323‐1341. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. Ronneberger O, Fischer P, Brox T. U‐net: convolutional networks for biomedical image segmentation. In: Navab N, Hornegger J, Wells W, Frangi A, eds. Medical Image Computing and Computer‐Assisted Intervention – MICCAI 2015. MICCAI 2015. Lecture Notes in Computer Science(). Vol 9351. Springer; 2015. [Google Scholar]
  • 26. Thomen RP, Walkup LL, Roach DJ, Cleveland ZI, Clancy JP, Woods JC. Hyperpolarized 129Xe for investigation of mild cystic fibrosis lung disease in pediatric patients. J Cyst Fibros. 2017;16:275‐282. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Feng L. Golden‐angle radial MRI: basics, advances, and applications. J Magn Reson Imaging. 2022;56:45‐62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28. Jiang W, Ong F, Johnson KM, et al. Motion robust high resolution 3D free‐breathing pulmonary MRI using dynamic 3D image self‐navigator. Magn Reson Med. 2018;79:2954‐2967. doi: 10.1002/mrm.26958 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Pruessmann KP, Weiger M, Börnert P, Boesiger P. Advances in sensitivity encoding with arbitrary k‐space trajectories. Magn Reson Med. 2001;46:638‐651. doi: 10.1002/mrm.1241 [DOI] [PubMed] [Google Scholar]
  • 30. Daniel Forsberg . Fordanic/image‐registration. 2022. https:// github.com/fordanic/image‐registration. GitHub. Retrieved October 26, 2022.
  • 31. Kirby M, Heydarian M, Svenningsen S, et al. 3He magnetic resonance functional imaging semiautomated segmentation. Acad Radiol. 2012;19:141‐152. [DOI] [PubMed] [Google Scholar]
  • 32. Bates D, Mächler M, Bolker B, Walker S. lme4: Linear Mixed‐Effects Models using ‘Eigen’ and S4 (R package version 1.1‐31). 2023. https://CRAN.R‐project.org/package=lme4.
  • 33. Satterthwaite FE. An approximate distribution of estimates of variance components. Biometrics. 1946;2:110‐114. [PubMed] [Google Scholar]
  • 34. Nakagawa S, Schielzeth H. A general and simple method for obtaining R2 from generalized linear mixed‐effects models. Methods in Ecology and Evolution. 2013;4:133‐142. [Google Scholar]
  • 35. Zanette B, Schrauben EM, Munidasa S, et al. Clinical feasibility of structural and functional MRI in free breathing neonates and infants. J Magn Reson Imaging. 2022;55:1696‐1707. doi: 10.1002/jmri.28165 [DOI] [PubMed] [Google Scholar]
  • 36. Dyke JP, Voskrebenzev A, Blatt LK, et al. Assessment of lung ventilation of premature infants with bronchopulmonary dysplasia at 1.5 tesla using phase‐resolved functional lung magnetic resonance imaging. Pediatr Radiol. 2023;53:1076‐1084. doi: 10.1007/s00247-023-05598-6 [DOI] [PubMed] [Google Scholar]
  • 37. Zanette B, Munidasa S, Friedlander Y, Ratjen F, Santyr G. A 3D stack‐of‐spirals approach for rapid hyperpolarized 129Xe ventilation mapping in pediatric cystic fibrosis lung disease. Magn Reson Med. 2023;89:1083‐1091. doi: 10.1002/mrm.29505 [DOI] [PubMed] [Google Scholar]
  • 38. Willmering MM, Niedbalski PJ, Wang H, et al. Improved pulmonary 129Xe ventilation imaging via 3D‐spiral UTE MRI. Magn Reson Med. 2020;84:312‐320. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Kim H, Lee S, Park J, et al. Evaluation of a patient with interstitial lung disease using 3D perfusion maps based on 3D ultrashort Echo‐time imaging. Proc Intl Soc Mag Reson Med. 2023;31:4814. [Google Scholar]
  • 40. Simonyan K, Zisserman A. Very Deep Convolutional Networks for Large‐Scale Image Recognition arXiv:1409.1556 [cs.CV]. 2014.

Articles from Magnetic Resonance in Medicine are provided here courtesy of Wiley

RESOURCES