Airspace Dimension Assessment (AiDA) by inhaled nanoparticles: benchmarking with hyperpolarised 129Xe diffusion-weighted lung MRI

Enlargements of distal airspaces can indicate pathological changes in the lung, but accessible and precise techniques able to measure these regions are lacking. Airspace Dimension Assessment with inhaled nanoparticles (AiDA) is a new method developed for in vivo measurement of distal airspace dimensions. The aim of this study was to benchmark the AiDA method against quantitative measurements of distal airspaces from hyperpolarised 129Xe diffusion-weighted (DW)-lung magnetic resonance imaging (MRI). AiDA and 129Xe DW-MRI measurements were performed in 23 healthy volunteers who spanned an age range of 23–70 years. The relationship between the 129Xe DW-MRI and AiDA metrics was tested using Spearman’s rank correlation coefficient. Significant correlations were observed between AiDA distal airspace radius (rAiDA) and mean 129Xe apparent diffusion coefficient (ADC) (p < 0.005), distributed diffusivity coefficient (DDC) (p < 0.001) and distal airspace dimension (LmD) (p < 0.001). A mean bias of − 1.2 µm towards rAiDA was observed between 129Xe LmD and rAiDA, indicating that rAiDA is a measure of distal airspace dimension. The AiDA R0 intercept correlated with MRI 129Xe α (p = 0.02), a marker of distal airspace heterogeneity. This study demonstrates that AiDA has potential to characterize the distal airspace microstructures and may serve as an alternative method for clinical examination of the lungs.

Detection of microstructural changes in the distal airspaces can be crucial for clinical evaluation of early stage lung disease and longitudinal monitoring of pulmonary diseases. The standard procedures for detection of disease in the distal airspaces include spirometry, test of diffusing capacity of the lung for carbon monoxide (D LCO ), and lung density analysis from computed tomography (CT). However, all the mentioned techniques can fail to reveal early indications of pathological changes 1,2 . Therefore, considerable disease can be present with minimal effect on standard pulmonary function tests (PFTs).
Diffusion-weighted (DW) magnetic resonance imaging (MRI) with inhaled hyperpolarised noble gases helium-3 ( 3 He) or xenon-129 ( 129 Xe) is an in vivo imaging method that is sensitive to changes in the distal airspaces [3][4][5] . The method is based on measurement of the Brownian diffusional restriction of the inhaled hyperpolarised gas atoms within the distal airspace walls. This property is used to derive the apparent diffusion coefficient (ADC), which provides 3D in vivo information on the distal airspace microstructure. In addition to ADC, theoretical models of hyperpolarised gas diffusion within the lungs, such as the stretched exponential model (SEM) 6,7 and the cylinder airway model (CM) 8,9 , can be used to derive distal airspace dimensions, analogous to those obtained through histological analysis. These in vivo distal airspace measurements from hyperpolarised gas DW-MRI have shown good agreement with direct morphometric measurements in validation studies with lung specimens 8,10,11 . Numerous studies have used hyperpolarised gas DW-MRI to elucidate changes in distal airspaces related to smoking [12][13][14] , ageing 15 , lung inflation 16 , and diseases such as COPD 17,18 , asthma 19 , and idiopathic pulmonary fibrosis (IPF) 20  www.nature.com/scientificreports/ hyperpolarised gas MRI, the technique is not typically employed in standard pulmonary function testing or screening for pulmonary disease. Airspace Dimension Assessment (AiDA) with inhaled nanoparticles, could potentially be more cost efficient and more widely accessible than MRI with hyperpolarised gases. The method is based on measurement of the exhaled recovery of inhaled nanoparticles which deposit in the distal airspaces due to Brownian diffusion 21 . The fraction of deposited particles is directly related to the size of the airspaces, and analysis of the nanoparticle recovery yields two metrics. The first metric is an effective airspace radius (r AiDA ), which is a root mean square measure of a collection of airspaces, and the second metric is the recovery at an imaginary zero-seconds breathhold (R 0 ). r AiDA has been found to correlate with the extent of emphysema 22 and proton lung tissue density as quantified by standard pulmonary structural MRI 23 . R 0 significantly correlates with the carbon monoxide transfer coefficient (K CO ) and age 24 . Further benchmarking of the technique with established methods of in vivo distal airspace assessment is required to evaluate the clinical potential of the AiDA technique.
The aim of this study was to benchmark inhaled nanoparticle measurements with AiDA against 129 Xe DW-MRI derived ADC and distal airspace dimensions. Since both AiDA 21 and 129 Xe DW-MRI 6 use diffusion of nanoparticles and gas molecules, we hypothesise that measures from the two different techniques will correlate.

Results
Volunteer demographics, pulmonary function test data, DW-MRI metrics and AiDA variables. Benchmarking of AiDA using the DW-MRI measurements. Statistically significant linear correlations were observed between r AiDA and 129 Xe DW-MRI metrics ADC (p < 0.005), DDC (p < 0.001) and Lm D (p < 0.001) (Fig. 1a-c). The R 0 intercept measurement from AiDA was significantly correlated with 129 Xe α heterogeneity index (p = 0.02) (Fig. 1d), but not with any of the other 129 Xe DW-MRI metrics. Bland-Altman analysis of r AiDA and 129 Xe Lm D showed a mean bias of − 1.2 µm (95% agreement limits − 36.0 to 33.6 µm) towards r AiDA (Fig. 2). r AiDA and Lm D deviated less than 0.05% for the whole group and on average 8% between subjects.
For the comparison in the Bland-Altman plot, a linear relationship of strength ρ = 0.48 was observed (p = 0.02) suggesting r AiDA increased more than Lm D with increasing airspace size. No difference was seen between men and women for the AiDA and 129 Xe DW-MRI metric distributions.

Discussion
Measurements of distal airspace dimensions were acquired with AiDA and 129 Xe DW-MRI in 23 healthy volunteers. AiDA metrics, r AiDA and R 0 , corresponded to previously reported values from healthy volunteers 22-24 . 129 Xe DW-MRI metrics of ADC, DDC and distal airspace dimensions (Lm D ) were larger than the values previously reported for younger healthy volunteers (29 ± 4 years), but smaller than values in ex-smoker volunteers 6 . Distal airspace radius r AiDA , significantly correlated with 129 Xe DW-MRI metrics ADC, DDC and Lm D , which confirm to us that r AiDA is a measure of distal airspace dimensions. Agreement between r AiDA and 129 Xe DW-MRI Lm D was confirmed with Bland-Altman analysis where a mean bias of − 1.2 µm towards r AiDA , corresponding to 0.43% deviation from the mean measured r AiDA , was observed. Both AiDA and 129 Xe DW-MRI measure the Brownian motion, presumably within the same distal airspace. The breath-hold times and diffusion times were optimised specifically for diffusion across small distances but assume that the particles stay within the same airway duct. The statistically significant correlation between R 0 and α heterogeneity index (p = 0.02) could be indicating that R 0 represents heterogeneity in the acini, which both metrics are hypothesized to measure 6,24 .
The ADC, DDC and Lm D dependency on age suggests a prevalence of age-related distal airspace changes in the older healthy volunteers 25 . The significant correlations between volunteer age and 129 Xe ADC and Lm D further demonstrate the sensitivity of DW-MRI to age-related airspace changes, and matches trends previously observed with 3 He DW-MRI 15,26 . A trend towards increasing r AiDA with age was observed. α heterogeneity correlated with volunteer height (p < 0.001). AiDA and 129 Xe DW-MRI metrics did not correlate with any other volunteer demographic or PFT data.
The bias and relatively wide 95% agreement limits (− 36.0 to 33.6 µm) observed in the Bland-Altman plot can be explained by the difference in acquisition method of the two techniques. AiDA is acquired over multiple breath-holds with diffusion times ranging between 5 and 15 s. In contrast, 129 Xe DW-MRI is a single breathhold acquisition with an 8.5 ms diffusion time. The linear trend in bias visible in the Bland Altman plot (Fig. 2) suggests that the agreement between r AiDA and Lm D changes with increasing mean airspace dimension size. This implies that the measurements from the two methods might diverge for increasing distal airspaces with an increasing bias between r AiDA and Lm D in larger airspaces. The increasing bias may be attributed to the difference between the diffusion coefficients for xenon (0.14 cm 2 /s when mixed with air in the lungs 27   www.nature.com/scientificreports/ nanoparticles (2 × 10 -5 cm 2 /s 28 ). Due to the much smaller mass and size of 129 Xe-atoms they diffuse much faster than the nanoparticles. Hence, the exponential decay, from which the diffusion distance in the lung is calculated, is different for the two techniques. For 129 Xe DW-MRI the theoretical 1D free diffusion length ( L 1D ) is approximately 500 µm ( L 1D = √ 2 D 0 ; = 8.5 ms, D 0 = 0.14 cm 2 /s) while for 50 nm nanoparticles the corresponding displacement in one direction is approximately 200 µm ( = 10 s, D 0 = 2 × 10 −5 cm 2 /s).
While the 129 Xe DW-MRI method provides voxel-wise regional information about the lung structure the AiDA method provides one measure of the distal airspace dimensions. Therefore, the 129 Xe DW-MRI method can give more detailed and regional information about the structure of the distal airspaces compared to AiDA. However, as indicated by this study, the AiDA method has the potential to give a faster and more accessible, but still precise, measurement of the distal airspace dimensions, which can be of great importance when hyperpolarised lung MRI is not available. To further compare the relative sensitivity between the two methods more measurements, in particular including a large variation in distal space sizes, are needed.
Although the two methods obtained approximately similar airspace dimensions (average r AiDA and Lm D deviated less than 0.5%), there were also several significant differences in the measurement procedures for the subjects. 129 Xe DW-MRI measurements were made at FRC + 1 L while AiDA measurements were made at TLC. Hence, the lung inflation was larger for all AiDA measurements as compared to 129 Xe DW-MRI measurements. Previous studies have shown that measured ADC in hyperpolarised 3 He DW-MRI increases with increasing lung inflation volumes 29 . Which means that, AiDA could be expected to have measured larger dimensions r AiDA when compared to Lm D .
Distal airspace size also depends on posture and ADC decreases from the non-dependent region of the lung down to the dependent region. For hyperpolarised 3 He DW-MRI measurements, ADC has been found to vary significantly depending on posture, and this was attributed to the compression of parenchyma, due to the lungs own weight, and the mass of the heart 30 . In this study AiDA measurements were performed in upright sitting position while the 129 Xe DW-MRI measurements were performed in supine position. Therefore, a postural variation between whole lungs AiDA measurement and the regionally averaged 129 Xe DW-MRI metrics is expected. For a more elaborate comparison of the two methods the breath-hold volumes could be set to be equal and breath-hold times set to correspond to the same diffusion distances. In addition, AiDA could potentially be measured with subjects in a supine posture enabling an even more efficient comparison with minimized systematic errors.
AiDA measurements have been shown to be repeatable to approximately < 2.4% when measured at different times over a period of 18 months 31 . Previous studies have shown that 3 He and 129 Xe ADC is highly repeatable in COPD patients with a coefficient of variation of 2.98% and 2.77% respectively, over 5 visits 32 . Similarly it has been showed that 3 He Lm D in patients with idiopathic pulmonary fibrosis is highly repeatable with a 0.6% difference between same-day visits 20 . This is the first study that compares AiDA with an independent and validated non-invasive method for assessment of distal airspaces. The study includes a limited number of healthy subjects, which mainly was due to the logistics related to travel between Sweden and the UK. The subject group was thus homogenous with small variations in PFT results. However, even with the small group of volunteers, significant correlations were found between AiDA and 129 Xe DW-MRI metrics, and interestingly these increased with age indicating age dependent changes in alveolar size or 'aging emphysema' 25 . Ideally, a future extension of this study would include subjects with a greater range of distal airspace sizes, such as patients with emphysematous lung disease.
In conclusion, this work has compared estimates of airspace radii from inhaled nanoparticles by the AiDA method with 129 Xe DW-MRI in a healthy volunteer cohort. The significant correlations show that the distal airspace radius of the lungs measured by AiDA (r AiDA ) can be related to distal airspace microstructure dimensions as quantified by 129 Xe DW-MRI. Quantitavely the two methods are in close agreement, with mean r AiDA and Lm D deviating < 0.5% for the whole group and on average 8% on an individual level. Further benchmarking in selected groups of patients could be used to evaluate the relative sensitivity of AiDA and 129 Xe DW-MRI in detecting early emphysematous changes to the distal airspace microstructure.

Methods
Study subjects and study design. The study enrolled 23 healthy adult volunteers (14M, 9F) in the age range 23-70 years with no history of pulmonary disease, in Sweden in the spring 2019. The study was approved by the Regional Ethical Review Board in Lund, Sweden (application number 2018/659), and performed in accordance with the Declaration of Helsinki, including obtaining informed written consent from all volunteers. All volunteers underwent spirometry, body plethysmography, carbon monoxide gas transfer, and AiDA measurements at Skåne University Hospital in Malmö, Sweden. Pulmonary function tests were performed according to the European Respiratory Society guidelines 33 . All 129 Xe DW-MRI measurements were performed at University of Sheffield, Sheffield, UK.
AiDA measurements. The AiDA method and instrumentation has been described in detail elsewhere 31 . Figure 4 displays a schematic illustration of the AiDA instrument 23 and example data from one subject.
In short, each subject initially inhaled particle-free air to remove background particles from their lungs. The subject was then instructed to exhale to residual volume (RV) prior to inhaling 50 nm polystyrene latex (PSL) aerosol particles to total lung capacity (TLC), hold their breath for a predefined time and finally exhale. For each measurement, the subject sat upright and a nose clip prevented them from breathing through the nose. Inhaled and exhaled nanoparticle number concentrations were registered for 8 consecutive measurements, with breath-hold times between 5 and 15 s. The inhaled nanoparticle concentration employed for AiDA was less than 10,000 cm −3 , which is lower than the concentration of ambient nanoparticles in an urban environment 34  www.nature.com/scientificreports/ The probability of particle deposition in the distal airspaces depends on residence time and airspace size. Enlarged airspaces yield a lower deposited fraction, corresponding to a higher particle recovery ( R ) in exhaled gas. The estimation of airspace dimensions using AiDA is based on the solution of the diffusion equation in circular tubes randomly distributed with axisymmetric boundary conditions 21 . The solution shows that the recovery R decays exponentially with residence time (t) in the lung 35 according to: where R 0 is the recovery at zero-second breath-hold and and t 1/2 is the deposition half-life time. Residence time t was established and linear least-squares regression was fitted to the data. From the fit, R 0 and t 1/2 were determined. R 0 is presumably related to the dynamic phase of breathing and small conducting airways (generation 10-15 36 ), but remains to be evaluated further 21,24 . The airspace radius r AiDA , was calculated from t 1/2 and the diffusion coefficient ( D) for 50 nm particles according to: 129 Xe DW-MRI measurements. Hyperpolarised 129 Xe DW-MRI was performed on a GE HDx 1.5T scanner with a flexible transmit/receive quadrature vest coil (Clinical MR Solutions, Brookfield, WI, USA) after the inhalation of a 1 L gas mixture containing 550 mL 129 Xe (> 25% polarization 37 ) and 450 mL N 2 from the level of function residual capacity. A 3D multiple b-value spoiled gradient echo (SPGR) sequence with compressed sensing was used with a 16 s breath-hold 6 . Specific DW-MRI acquisition parameters were: TE/TR = 14.0/17.3 ms, 129 Xe diffusion time = 8.5 ms, b = [0, 12, 20, 30 s/cm 2 ]. 129 Xe ADC was calculated using a mono-exponential fit between the signal of the b = 0 (S 0 ) and 12 (S b=12 ) s/cm 2 interleaves: The mean diffusive length scale (Lm D ), a measure of mean distal airspace dimension from the SEM was derived by fitting the 129 Xe diffusion signal from all four b-values to a stretched exponential function 6 : where p(D) is the probability distribution of different apparent diffusivities within each image voxel, DDC is the distributed diffusivity coefficient, and α is the heterogeneity index that describes the deviation from a monoexponential decay (α = 1). α is thought to be a measure of the underlying complexity or heterogeneity of the geometry of the restricting distal boundaries 38 . From α and DDC a numerical expression for p(D) is estimated 39   www.nature.com/scientificreports/ Both ADC and SEM metrics were calculated on a voxel-by-voxel basis, and averaged across the entire lung volume to derive global means. Figure 5 displays examples of ADC and SEM-derived Lm D maps from the hyperpolarised 129 Xe DW-MRI.
Benchmarking of AiDA versus DW-MRI measurements. All of the AiDA analysis, 129 Xe DW-MRI lung morphometry calculations and comparisons of metrics were implemented using MATLAB Version 2020a (The MathWorks, Inc., Natick, Massachusetts, United States). Correlations between AiDA variables, 129 Xe DW-MRI metrics and standard PFT measurements were assessed using Spearman's rank correlation test. The significance threshold was set at 0.05. Bland-Altman analysis was used to assess the agreement between the r AiDA and Lm D .

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.