Dynamic changes in lung water density and volume following supine body positioning

Measure the changes in relative lung water density (rLWD), lung volume, and total lung water content as a function of time after supine body positioning.


INTRODUCTION
Pulmonary edema, or fluid accumulation in the lungs, is a primary feature of heart failure that is associated with clinical symptoms such as dyspnea and poor exercise capacity [1][2][3] and poor health outcomes. 4ime-efficient 3D ultrashort-TE (UTE) imaging methods designed for imaging of the lung parenchyma in combination with postprocessing methods for signal normalization and segmentation now enable calculation of water density-weighted lung images with breath-hold or short-duration free-breathing acquisitions. 5,6RI studies of lung water typically take place with supine body positioning.The distribution of abdominal organs and thus lung volume is dependent on body position due to the effects of gravity, 7,8 as is the distribution of blood and free water in the body. 9,10There is shift of 300-500 mL of blood to the thoracic cavity with supine positioning [11][12][13] and differences in the chest to back gradient in lung water density for prone versus supine positioning, using MRI 6,14 and other modalities. 15,16However, the time course of changes in regional lung water density, lung volumes, and total lung water content following supine positioning remain unknown. 6,17he goals of the current study were to measure the changes in global and regional relative lung water density (rLWD) as a function of time after supine body positioning and their relationship with changes in lung volumes and total lung water content.We expect to determine the time needed to achieve steady-state lung water distributions following supine positioning and whether regional or total lung water content vary significantly over the duration of a typical MRI scan session.

Study procedure and subjects
Studies were completed in 9 healthy volunteers (5 male, 4 female, 33.2 ± 10.2 years [21-53 years], 1.72 ± 0.05 m, 69.8 ± 12.4 kg; Table 1).A health questionnaire was used to confirm that all subjects had no known history of pulmonary or cardiovascular disease and with no current medications.This study was approved by the University of Alberta Health Research Ethics Board, and written informed consent was given by all study participants.Serial 3D water-density images of the lungs were acquired following supine body positioning.A timer was started with subject positioning on the scanner bed, followed by placement of receiver coils and insertion into the scanner.A localizer scan was performed immediately, after which the lung water imaging acquisition was prescribed and started.The lung water acquisitions were repeated a total of 18 times over 25 min, with the time after supine positioning recorded at the onset of each repeated scan.Three subjects repeated the procedure on a separate day with the addition of a total lung capacity breath-hold (10 s) at the half-way time point.Finally, a nitrogen washout study for calculation of functional residual capacity (FRC) (minimum lung volume during tidal breathing) was performed in 3 subjects, following supine body positioning.Repeatability studies were used to determine the time needed for complete nitrogen washout (i.e., the minimum time between studies without contamination from the previous study).Complete washout was measured to be less than 5 min in the supine body position in all test cases; thus, a time interval of 5 min was selected for serial studies to evaluate changes in lung volume following supine body positioning.Nitrogen measurements were completed upright and at six separate time points for 30 min following supine positioning at 5-min intervals.For all studies, volunteers were instructed to breath normally over the entire scan duration with no other instructions.

Image acquisition
All imaging data were acquired on a Siemens Prisma 3T scanner (Erlangen, Germany) using spine and body arrays (36 total coils) for signal reception to provide full torso coverage.Three-dimensional spin density-weighted lung images were acquired in a free-breathing acquisition using an optimized yarnball k-space trajectory. 5cquisition parameters were FOV = 350 mm in all directions, TR = 2.72 ms, TE = 100 μs, flip angle = 1 , readout time = 1.3 ms, voxel size = 3.5 × 3.5 × 3.5 mm (2-mm interpolated reconstruction with zero padding), and 2738 trajectories.Each full k-space acquisition was 7.4 s with 10 serial repetitions performed during free breathing for a total acquisition time of 74 s per time point.The k-space trajectories were incremented in a golden-ratio manner to ensure that any subset of trajectories used for respiratory-resolved reconstructions (described subsequently) included an even and pseudo-random distribution of trajectories throughout k-space.Short TEs (100 μs) and low flip angles (1 • ) ensured minimal T 2 * and T 1 weighting for proton-density weighting. 5The complete 74-s scan is referred to as a single time step and was repeated 18 times in each subject with a 2-s delay between sequential acquisitions.All raw data were transferred off the scanner for processing with custom MATLAB software (The MathWorks Inc., Natick, MA, USA).On a separate day, the study procedure was repeated in 3 subjects, with 11 data sets acquired immediately following supine body positioning (15-min acquisition time), followed by a total lung capacity inspiration breath-hold of 10 s and the acquisition of an addition 11 data sets following the breath-hold.

Image reconstruction and image processing
The image reconstruction and postprocessing steps for each complete free-breathing acquisition (each time step) have been previously reported. 5Briefly, a respiratory navigator signal was calculated using the center of k-space from each acquisition (every TR), which was used to assign each acquired k-space arm a respiratory phase with values of from 0% to 100% of a breathing cycle.Images at FRC (minimum free-breathing lung volume) were reconstructed using one of the 10 repeated k-space arms for which the respiratory phase was closest to FRC (0% of breathing cycle), independently for each k-space arm.Similarly, images were reconstructed at 50% of the breathing cycle to approximate FRC + tidal breathing volume (TV) (maximum free-breathing lung volume) using one of the 10 repeated k-space arms closest to the targeted 50% respiratory phase.Acquisitions included 10 repetitions of each k-space arm to provide respiratory phases near FRC (0% respiratory phase) and FRC + TV (50% respiratory phase) for each arm.Following data selection for image reconstruction at FRC and FRC + TV, raw data were sampling density-compensated 18 to a  = 2 Kaiser k-space filtering shape with zero padding to the final 2-mm pixel size in each dimension.Images were then constructed using standard gridding methodology, 19 and the coil elements combined with the SUPER approach. 20econstructed images were normalized to units of rLWD using a previously described and validated approach that corrects for signal intensity variations over space from B 1 transmit and receive coil inhomogeneity. 5rLWD is reported in units of percentage, relative to the reference tissues.
The lung parenchyma was segmented using an in-house nnU-Net machine learning algorithm 21,22 with separate masks for the lung parenchyma and vascular pool, with unique identification of right and left lungs (Figure 1A).Total lung volume was calculated as the total number of lung pixels (parenchyma plus vascular) multiplied by the voxel volume.Water volume in each region was calculated as the sum of rLWD values in the masked region (percentage values were divided by 100 to change units to a fraction) multiplied by the voxel volume, separately for the parenchyma, vascular, and total lung regions.All parameters (rLWD, lung volume, and lung water volume) were calculated at each of the 18 time points at FRC and FRC + TV respiratory phases.Tidal volume was calculated as the difference between minimum and maximum lung volume.The same image-processing pipeline (FRC analysis only) was used for the additional 3 subjects with 11 time steps (15 min) after supine positioning and 11 additional time steps following the total lung capacity breath hold.Continuous variables were expressed as mean ± SD.Paired student t-tests were used to analyze differences in average rLWD, lung volume, and lung water volume over time (first scan vs. 18th scan) and differences between all values at the two respiratory phases.p-Values less than 0.05 were considered statistically significant.Repeatability of rLWD, lung volume, and lung water volume was calculated using images from Time-steps 17 and 18 in each subject, calculated as intraclass correlation coefficients and displayed using Bland and Altman plots.
The nitrogen wash-out technique was performed as per guidelines. 23Briefly, participants were situated on the mouthpiece and instructed to breath at normal tidal volume.Importantly, before initiating each maneuver, participant tidal breathing nitrogen concentration was confirmed to be at baseline level (∼78%) to ensure no "spillover" from repeat maneuvers.Once stability of tidal breathing was reached, participants were given 100% oxygen while maintaining tidal breathing.The test was terminated when the expired gas contained less than 1.5% nitrogen for three consecutive breathes.

Regional lung analysis
The predominant spatial variation in lung water content is in the chest-to-back direction, 5,6,24,25 with uniform distributions expected in the head-to-foot and right-to-left directions. 5To calculate chest-to-back gradients and their trends over time, the 3D images were grouped into 10 equally sized bins with coronal slice orientations (Figure 1B).For each 3D data set, all bins contained the same number of coronal slices, with Bin 1 representing the slices closest to the chest and Bin 10 closest to the back.The average rLWD was calculated from all lung pixels (parenchyma plus vascular) within each bin of slices as well as for just the parenchyma (lung segmentation illustrated in Figure 1A).The slices included within each bin were recalculated for each time point to account for changes in total lung volume over time.

RESULTS
The average time from the supine body positioning to the completion of the first imaging time point was 3:54 ± 0:26 min, with an increment of 1:16 min between each subsequent time point, for an average total study time of 25:24 ± 0:25 min.All grouped results are reported using the mean time for the first time point, to define a common time axis for all subjects.Representative images at FRC show typical normalized images and rLWD maps within the lung parenchyma and vascular spaces (Figure 1C,D).Sample FRC images for all subjects are provided in Figure S1.Additionally, images for multiple respiratory phases and all 18 study time points for a single subject are provided in Video S1.

Lung water density and volume changes at FRC
At FRC, there was a statistically significant increase in global rLWD (parenchyma plus vascular) over time (31.8 ± 5.5% at t = 3:54 min vs. 34.8± 6.8% at t = 25:24 min; p = 0.001).There was a statistically significant decrease in total lung volume over time (2.39 ± 0.62 L at t = 3:54 min vs. 2.13 ± 0.63 L at t = 25:24 min; p < 0.001) and a small but significant decrease in total lung water volume (730 ± 125 mL at t = 3:54 min vs. 706 ± 126 mL at t = 25:24 min; p = 0.028).Vascular water volumes did not change significantly over time (112 ± 28 mL at t = 3:54 min vs. 121 ± 27 mL at t = 25:24; p = 0.077).Starting and final values for rLWD, lung volume, lung water volume, and vascular water volume at the FRC respiratory phase for all 9 subjects are given in Table 1.rLWD values in the parenchyma region (i.e., excluding the vascular region) are summarized in Table S1.The time course of all parameters for the FRC respiratory phase showed a logarithmic growth pattern for rLWD with a total increase of 9.2 ± 4.4%, whereas total lung volume decreased by 11.1 ± 5.3% with a similar pattern over time, with a smaller reduction in water volume of 3.3 ± 4.0% (Figure 2A-C).All time-course data showed the average values for all participants with error bars for the SD of the mean of each subject.
The average regional rLWD in all 10 chest-to-back bins showed an increase in lung water density over time with an expected chest-to-back gradient clearly visible at the first time point (baseline), with 20.7 ± 4.6% rLWD for the chest and 39.9 ± 6.1% rLWD at the back (Figure 2D).The absolute increase in rLWD over time was uniform from chest to mid-lung (Bins 1 to 6), all with an increase of 1.8 ± 1.2%, with larger absolute increases of 5.4 ± 1.9% toward the back, increasing from 39.6 ± 6.1% to 45.3 ± 6.4% by 25 min after supine positioning (Figure 2E).Volume changes in the 10 chest-to-back bins showed a similar pattern as rLWD, with larger changes toward the back (Figure 2F).Similar results were found at the FRC + TV respiratory phase (not shown).
Fractional change over time for whole-lung (global) average relative lung water density (rLWD) (A), total lung volume (B), and total lung water volume at functional residual capacity (FRC; minimum lung volume) (C), following supine body positioning.All values are normalized to the first time point with a total of 18 time points over 25 min after supine positioning.Average values for all participants are shown with error bars for the SD of the mean.Evaluation in 10 chest-to-back regions illustrate the evolution of the absolute rLWD values in each region (D) and the absolute changes in rLWD over time (E) following supine body positioning.F, The corresponding absolute regional changes in lung volumes.
(A) (B) (C) Fractional change over time for whole-lung average relative lung water density (rLWD) (A), total lung volume (B), and total lung water volume at functional residual capacity (FRC; minimum lung volume) (C), following supine body positioning and subsequently following a total lung capacity (end-inspiration) breath hold (subgroup of 3 participants).All values are normalized to the first time point with a total of 11 time points over 15 min after supine positioning and after the breath-hold maneuver.Average values for all participants are shown with error bars for SD of the mean.

Lung water density and volume changes at FRC + TV
At FRC + TV, there was a statistically significant increase in average rLWD over time (26.6 ± 3.2% at t = 3:54 min vs. 28.9 ± 2.8% at t = 25:24 min; p = 0.005).There was a statistically significant decrease in total lung volume over time (2.94 ± 0.63 L at t = 3:54 min vs. 2.61 ± 0.64 L at t = 25:24 min; p < 0.001).Total lung water volume did not change significantly over time (767 ± 122 mL at t = 3:54 min vs. 736 ± 137 mL at t = 25:24; p = 0.088) nor did vascular water volumes (107 ± 30 mL at t = 3:54 vs. 113 ± 29 mL at t = 25:24 min, p = 0.216).Starting and final values for rLWD, lung volume, lung water volume, and vascular water volume at the FRC + TV respiratory phase are given in Table S2, with values in the parenchyma region in Table S3. Figure S2 shows the time course of rLWD, lung volume, and water volume up to the 25-min time point for the FRC + TV respiratory phase.Lung volumes were significantly larger, and rLWD density values were significantly lower at FRC + TV as compared with FRC for all time points following supine positioning (p < 0.001 for all).
Tidal volume, or the change in lung volume over a respiratory cycle, decreased significantly over time after supine positioning (474 ± 89 mL at t = 3:54 vs. 382 ± 91 mL at t = 25:24; p = 0.018) (Figure S3).There were no significant changes in breathing rate over time, with average values of 13.3 ± 3.3 breaths per minute over the full study duration and less than 1 breath per minute change over time.
Bland and Altman plots display the percentage difference in rLWD, lung volume, and lung water volumes between Time-steps 17 and 18 for all subjects, for FRC and FRC + TV respiratory phases (intraclass correlation coefficient > 0.99 for comparisons) (Figure S4).
In additional studies, changes in rLWD, lung volume, and total water volume following a total lung capacity inspiration breath hold followed the same pattern as observed following supine body positioning (Figure 3).In particular, a 9.2% increase in rLWD was observed over 15 min following the breath hold, with a corresponding decrease in lung volume over the same interval.
Nitrogen washout studies further confirmed the MRI-derived changes in lung volume following supine body position with similar time course and magnitude of volume changes (Figure 4), with 12.5 ± 0.9% lung volume reduction measured with nitrogen washout versus 11.1 ± 5.3% for MRI studies and slightly faster volume changes for nitrogen washout.Lung volumes immediately decreased 27 ± 6% from upright to the first supine time point (not shown in Figure 4).

DISCUSSION
The current study illustrated a dynamic decrease in lung volume (11%) following supine body positioning that

F I G U R E 4
Relative changes in lung volume at functional residual capacity measured with the nitrogen washout method in 3 participants as a function of time following supine body positioning.There was an immediate 27 ± 6% drop in lung volumes from upright to supine positioning, before these time points (not shown).MRI-derived lung volumes from Figure 2B, in 10 separate participants, are overlaid for comparison with nitrogen washout data.
approached steady state after 25 min with a corresponding increase in global lung water density (rLWD), during restful tidal breathing.The increase in water density by 9% was primarily the result of an underlying increase in parenchyma tissue density but not a net movement of water into the lungs.In particular, despite the increased rLWD, the global water volume was shown to decrease slightly over the 25-min window following supine positioning.The expected increase in total thoracic blood volume with supine positioning [9][10][11][12][13] was not observed within the lungs, nor may it have occurred before the first imaging time point at 4 min after the change in body position.
The decrease in lung volume following supine body positioning was validated using a nitrogen washout study, confirming the time course and magnitude of the volume changes at FRC (12.5% change over 30 min).7][28][29] Additional serial MRI experiments following a large inspiratory breath hold at total lung capacity further confirmed the relatively slow dynamic reduction in lung volume and the associate increase in rLWD following large changes in lung volume (in this case, an end-inspiration breath hold), suggesting of the respiratory muscles as a potential mechanism.
The temporal evolution of each parameter (rLWD, lung volume, lung water volume, vascular water volume) was similar at minimum lung volume (FRC) and maximum lung volume (FRC + TV) during tidal breathing and thus do not depend on the respiratory phase of the acquired data.Breathing rates did not vary, but tidal volumes dropped significantly over time following supine body positioning.
Most of the gravity effects on the formation of the chest-to-back water density gradient were completed before the earliest time point at 4 min.In particular, all time points following supine body positioning showed a similar characteristic chest-to-back gradient in water density (Figure 2D) or the so-called slinky effect, whereby the dependent lung regions (lowest in relation to gravity) are relatively compressed. 14Regional analyses showed that the absolute change in rLWD over time was larger at the back as compared with chest and mid-lung positions, which is most likely due to the higher water content at the back as opposed to a shift of water to the back or in relation to large volume changes at the back.
A recent study of rLWD imaging study at 0.55 T showed no significant changes in water density over time with changes between supine and prone body positioning, with four image volumes acquired over 10 min (2.5 min per image set), starting 6 min after positioning. 6However, the timing from the upright to supine positioning was not directly reported; thus, the larger volume changes that occurred at the earlier time points following this particular positional change may not have been captured.
The observed reduction of FRC by 11% over 25 min, with a corresponding increase in rLWD, indicates that the timing of the lung water density measurements should be considered for free-breathing acquisitions.An end-inspiration breath hold was shown to have a similar effect on lung volumes with a relatively long recovery of FRC volumes with tidal breathing following a single breath hold, again highlighting the variability of lung volumes even during resting breathing.
At the final study time point, the FRC + TV volumes were 18% larger than FRC (2.61 L vs. 2.13 L), with a correspondingly lower rLWD of 28.9% at FRC + TV versus 34.8% at FRC (i.e., a 17% reduction in rLWD).These findings agree with the expected linear relationship between lung volume and water density 30 and highlight the extreme sensitivity of rLWD to the highly variable physiologic parameter of lung inflation.Breath-hold duration measurement of rLWD is also possible 5 but may be prone to larger variations in rLWD due to the subject-controlled lung volumes based on breath-hold positioning.Image registration has been shown to enable calculation of the Jacobian, which has the potential to correct the changes in rLWD with changes in lung inflation 6 and thus detect changes in rLWD independent of lung inflation in an individual.

Limitations
The current study has several limitations.First, the use of all solid tissues surrounding the lungs as a signal reference for lung water estimation is prone to systematic errors, particularly in cases of fluid accumulation in the reference tissues (e.g., heart failure, kidney failure) or fat accumulation, which could lead to underestimation of rLWD.The rLWD and water volumes reported in the current study are calculated relative to the reference tissues (which are themselves 70%-80% water content [31][32][33][34] ) and therefore overestimate the true densities and volumes.All measured values within the lungs (rLWD, volumes, and water volumes) depend on the machine-learning masks used to identify parenchyma and vascular spaces and could therefore be a source of systematic errors.Corrections for coil shading effects (normalization) have not been validated in the chest-to-back direction and may be prone to systematic errors in the calculation of the chest-to-back rLWD gradient.The maximum lung volume over the respiratory cycle, FRC + TV, was approximated to occur at a respiratory phase of 50% which may underestimate the true maximum value if the peak volumes occur at an earlier or later phase.Finally, the study design included imaging to 25 min following supine body positioning, which unfortunately was not sufficient to achieve steady state in lung volumes or rLWD in all cases.

CONCLUSIONS
Lung volumes during tidal breathing decrease significantly over tens of minutes following supine body positioning with corresponding increases in lung water density.The total volume of lung water is slightly reduced over this interval.Evaluation of rLWD should take into consideration the time after supine positioning and more generally all sources of lung volume changes, avoid significant bias of water density.
U R E 1 A, One sample coronal slice (position shown as yellow dashed line in [B]) and the corresponding segmentation mask show the five segmented regions, including the reference tissue outside of the lungs, parenchyma in the right and left lungs, and the vascular pool in the right and left lungs.B, Location of the 10 binned lung regions containing coronal slices in the chest-to-back direction for regional analysis of the lung parenchyma with an equal number of slices in each bin.C, Sample chest-to-back coronal slices from a free-breathing yarnball acquisition (FRC) following image normalization (8 of 120 coronal slices are shown).D, Images from (C) illustrating the lung segmentation and the chest-to-back gradient in relative lung water density (rLWD).

rLWD (%) Lung volume (L) Lung water volume (mL) Vascular water volume (mL) Subject Age (years) Sex Height (m) Weight (kg) Start End 𝚫 Start End 𝚫 Start End 𝚫 Start End 𝚫
Note: Δ is the change in each parameter from the first (Start) to the last (End) time points.Abbreviation: rLWD, relative lung water density.* Values are significantly different from the first time point (Start) to the last time point (End); p < 0.05.