Characterization of Pulmonary Functional Abnormalities in Systemic Sclerosis Using Xenon MRI

ABSTRACT

Background

Xenon MRI is increasingly used to evaluate patients with interstitial lung disease (ILD) and pulmonary hypertension (PH), both of which are common manifestations of systemic sclerosis (SSc). As such, Xe‐MRI may be suited to interrogate lung function impairment in SSc.

Purpose

To characterize xenon MRI signatures in SSc, toward evaluating the utility of xenon MRI as a method to elucidate mechanisms of regional lung function impairment in this population.

Study Type

Prospective.

Population

Of 25 participants initially imaged, 21 participants (18 females) with SSc were included. Sixteen healthy volunteers (13 females) were enrolled.

Fieldstrength/Sequence

3 T, 1‐point Dixon imaging using xenon MRI.

Assessment

Xenon MRI measures including ventilation defect percent, membrane uptake, red blood cell (RBC) transfer, RBC defect percent, and RBC oscillation amplitude were generated. Measures were compared across healthy and SSc groups and correlated with standard clinical measures, including demographics, pulmonary function tests, CT lung density measures, and pulmonary artery pressure.

Statistical Tests

Due to a small number of male participants, statistical analysis was limited to female participants. Wilcoxon, t‐tests, or Fisher's exact tests were used to compare between healthy and SSc groups. Pearson's correlation was used to correlate xenon MRI with clinical measures. p < 0.05 was considered significant.

Results

Despite a relatively mild burden of pulmonary disease, SSc participants exhibited significantly lower RBC/Membrane ratio (0.25 [IQR, 0.10] vs. 0.34 [0.05]) and RBC transfer (0.22 ± 0.07 vs. 0.28 ± 0.07), and significantly greater RBC defect percent (22.8 [IQR, 16.9] vs. 11.9 [13.7]) compared to healthy volunteers.

Data Conclusion

Xenon MRI measures, including RBC transfer, RBC/Membrane, and RBC defect percent were markedly different in female SSc patients compared to age‐matched healthy volunteers, suggesting that xenon MRI may be an effective method for examining regional impairments to pulmonary gas exchange in SSc.

Evidence Level

2.

Technical Efficacy

Stage 2.

Plain Language Summary

Systemic sclerosis is an autoimmune disease that is highly female predominant and commonly impacts the lungs, causing interstitial lung disease (ILD) and pulmonary hypertension (PH). Xenon MRI can be used to image these lung diseases, but it has not been applied in systemic sclerosis. The results in women with systemic sclerosis show that xenon MRI is sensitive to lung function impairment, particularly impairment in gas exchange. These findings suggest that xenon MRI may be a tool that can be used to come to an improved understanding of systemic sclerosis lung disease.

1. Introduction

Systemic sclerosis (SSc) is a rare, chronic autoimmune disease marked by progressive fibrosis and microvascular dysfunction of the skin and internal organs, affecting ~19 per 100,000 persons globally [1]. Pulmonary involvement, including interstitial lung disease (ILD) and pulmonary hypertension (PH), is a common manifestation of SSc; ILD affects up to 50% of SSc patients, while PH affects 7%–9% [2, 3, 4]. While PH and ILD are only diagnosed upon clinically important disease, post‐mortem analysis of lungs from deceased SSc patients suggests that interstitial and vascular changes are nearly universally present, even in the absence of clinically significant disease [5]. These findings support the concept that subclinical pulmonary involvement is present early in the disease course and may precede abnormalities detectable by conventional clinical testing.

Current approaches to the evaluation of pulmonary disease in SSc—including pulmonary function test (PFT), echocardiography, and high‐resolution computed tomography (HRCT)—are limited in their ability to detect early functional impairment or to disentangle contributions from alveolar, interstitial, and vascular pathology [6, 7]. HRCT provides high spatial resolution but is insensitive to early microvascular disease, while pulmonary function tests and echocardiography offer only global or indirect measures of lung involvement. As a result, there remains a critical unmet need for non‐invasive imaging biomarkers capable of detecting early pulmonary dysfunction and monitoring the development of disease in these patients.

Xenon MRI (Xe‐MRI) is a promising technique for evaluating impaired lung function on a regional basis [8]. In Xe‐MRI, the gaseous contrast agent “hyperpolarized ¹²⁹Xe” is inhaled and imaged during a short (< 15 s) breath‐hold. Imaging can be tailored to generate 3D maps of ventilation, airspace size, or gas exchange. Gas exchange is measured through the imaging of xenon dissolved in the alveolar capillary membrane (“Membrane”) and red blood cells (RBCs) [9, 10]. In ILD, patients consistently exhibit significantly increased membrane uptake on Xe‐MRI, as well as a corresponding decrease in RBC transfer [11, 12, 13, 14, 15]. In PH, patients show reduced RBC transfer and abnormal RBC oscillation amplitude [16]. Thus, the most common pulmonary manifestations of SSc—ILD and PH—are known to be detectable by Xe‐MRI.

The aims of this study were to evaluate the presentation of Xe‐MRI signatures in systemic sclerosis and compare Xe‐MRI measures to clinical standard measures of lung structure and function. This initial classification is an essential first step toward the wider use of Xe‐MRI in this population to evaluate disease mechanisms, identify early disease, and evaluate treatment response.

2. Materials and Methods

2.1. Participants

This study was a prospective, observational, cross‐sectional study examining Xe‐MRI features in patients with SSc. This study was performed with approval from the local Institutional Review Board (IRB) (University of Kansas Medical Center IRB, STUDY000146119) and the Food and Drug Administration (IND #151978). All participants provided written informed consent.

SSc Participants were recruited by physician referral from the Divisions of Allergy, Clinical Immunology and Rheumatology and Pulmonary, Critical Care, and Sleep Medicine. Inclusion criteria included: [1] Diagnosis of systemic sclerosis based on American College of Rheumatology/European League Against Rheumatism (ACR/EULAR) classification at the time of imaging; [2] Age ≥ 18; [3] CT imaging and pulmonary function testing (PFT) completed clinically at any time prior to Xe‐MRI [17]. Exclusion criteria included: [1] Standard MRI contraindications; [2] Pregnancy or breastfeeding. Participants were recruited irrespective of their lung health status. SSc participants were recruited between May 2022 and August 2024.

Clinical variables were collected from participants' medical records, including diagnosis (inclusive of limited vs. diffuse extent of disease), CT images, radiologist interpretation, PFT results, serology, echocardiography results, and right heart catheterization results. Radiologist interpretation included the assessment of CT pattern (normal, nonspecific, nonspecific interstitial pneumonia [NSIP], or usual interstitial pneumonia [UIP]). PFT results were extracted, selecting for [1] the PFTs closest in time to the Xe‐MRI date and [2] the PFTs closest in time to 1‐year follow‐up from the Xe‐MRI date. PFTs closest in time to Xe‐MRI date were used for comparison to Xe‐MRI features, while PFTs closest in time to 1‐year follow‐up were used as a coarse measure of disease progression.

Healthy volunteers were recruited through community advertising and imaged using Xe‐MRI for comparison with SSc participants. Inclusion criteria: [1] no history of clinical disease and [2] minimal smoking history (< 5 pack years and no smoking in the last 1 year). Exclusion criteria: [1] Standard MRI contraindications; [2] pregnancy or breastfeeding. Imaging was performed identically for SSc participants and healthy volunteers. Healthy volunteers were imaged between September 2021 and January 2026. Spirometry was acquired during the Xe‐MRI study visit. CT imaging and complete pulmonary function testing were not available in healthy volunteers.

2.2. Imaging Acquisition

Healthy and SSc participants completed hyperpolarized ¹²⁹Xe MRI at a single timepoint using an identical gas exchange imaging protocol. Isotopically enriched (85% ¹²⁹Xe) xenon gas (Linde Specialty Gases) was polarized to ~30% using a Polarean 9820 xenon hyperpolarizer (Polarean Imaging Plc., Durham, NC). The total dose volume delivered to participants was equal to 20% of FVC as measured at their most recent clinic visit (SSc participants), or 20% of FVC estimated based on Global Lung Initiative (https://gli‐calculator.ersnet.org/; healthy volunteers) [18]. Doses consisted of 100% xenon gas. Participants inhaled xenon doses from approximate functional residual capacity (FRC).

Participants were imaged using a 3 T scanner (MAGNETOM Skyra, Siemens Healthineers, Erlangen, Germany) using a flexible ¹²⁹Xe vest coil (Clinical MR Solutions LLC, Brookfield, WI). Before imaging, scanning parameters were calibrated using the protocol recommended by the ¹²⁹Xe MRI Clinical Trials Consortium [18]. Briefly, the calibration protocol included 500 free induction decay (FID) acquisitions centered at the RBC resonance (218 ppm), followed by 20 FID acquisitions centered at the gas resonance (0 ppm). Acquisition parameters for the FID acquisitions at both frequencies were identical and included: TR/TE = 15 ms/0.45 ms; Bandwidth = 25.6 kHz, Acquisition Points = 256; Total Duration = 8.4 ms. Following calibration, participants were imaged using a single‐breath ventilation/gas‐exchange sequence [19] Briefly, this imaging sequence collects a 3D radial 1‐point Dixon acquisition centered at the RBC frequency (218 ppm) interleaved with a higher resolution 3D spiral (FLORET) acquisition at the gas frequency [20]. Relevant sequence parameters included: Dissolved/Gas interleaf TR = 2.7/5.3 ms (total TR between like‐contrasts = 8 ms); Field of View = 400 × 400 × 400 mm³; Dissolved/Gas Matrix Size = 64 × 64 × 64/96 × 96 × 96; Dissolved/Gas Flip Angle = 14.6°/1°; Dissolved/Gas Bandwidth = 686/1635 Hz/Pixel; Number of Views = 1238; Total Duration = 9.9 s. The TE was set to the echo time at which membrane and RBC signals were 90° out of phase, as determined by the calibration sequence (0.45–0.50 ms). Within the same breath‐hold after xenon imaging, a geometry‐matched ¹H anatomic image set was acquired using a 3D spiral (FLORET) acquisition. Relevant parameters included: TR/TE = 3.3/0.3 ms; Field of View = 400 × 400 × 400; Matrix Size = 96 × 96 × 96; Bandwidth = 2957 Hz/Pixel; Number of Views = 906; Total Duration = 3.0 s. This anatomic image was used to assist in the generation of a thoracic cavity mask and for visualization of images. The total duration of breath‐hold for gas exchange Xe‐MRI and ¹H anatomic MRI was 14 s.

The clinically‐acquired CT image with the closest temporal proximity to Xe‐MRI was used to compare structural features with Xe‐MRI. CT scans were done with multi‐channel scanners with following scan parameters: Slice thickness: 2 mm (N = 16), 3 mm (N = 1); In‐plane resolution: Mean 0.68 × 0.68 mm² (Range 0.60 × 0.60 mm² to 0.95 × 0.95 mm²). Scanners were: Siemens SOMATOM Definition AS (N = 10), Siemens SOMATOM Definition FLASH (N = 3), GE LightSpeedVCT (N = 2), GE LightSpeed16 (N = 1), GE Optima CT660 (N = 1); Reconstruction Kernel: B31f (N = 10), I31f\3 (N = 2), Bf37f\2 (N = 1) for Siemens scanners, STANDARD (N = 2), CHST (N = 2) for GE scanners. For the subset of CT images used for vascular segmentation (N = 6), Parameters for this subset included: Slice thickness 0.6 mm; In‐plane resolution, mean 0.76 × 0.76 mm² (Range 0.59 × 0.59 mm² to 0.88 × 0.88 mm²). Scanner: SOMATOM Definition AS (N = 5) and SOMATOM Definition FLASH (N = 1) (Siemens Healthineers); Reconstruction Kernel: B60f (N = 5), Br58f (N = 1). Kilovolt peak (kVp) was 120 V and vendor provided AEC was adopted for all CTs used in analysis.

2.3. Image Analysis

Xe‐MRI Images were analyzed using open source Xe‐MRI analysis code for quantifying xenon spectroscopy and gas exchange (xenon‐dynamic‐spectroscopy‐consortium Version 1.0, xenon‐gas‐exchange‐consortium Pipeline Version 4.0.1) [21]. Briefly, calibration spectra were fit in the time domain to generate a time‐series evolution of RBC signal and quantify oscillation amplitude [22], which was corrected for estimated capillary blood volume [23]. Gas and dissolved images were reconstructed using regridding to a nominal resolution of 6.25 × 6.25 × 6.25 mm³. Dissolved images were separated into their constituent membrane and RBC images using the 1‐point Dixon technique [9]. Subsequently, images were binned based on a standard, multi‐site healthy reference distribution [24]. Ventilation Defect Percent (VDP), Membrane/Gas ratio (“membrane uptake”), RBC/Gas ratio (“RBC transfer”), and RBC defect percent were computed from images [25]. Signal ratios (Membrane/Gas, RBC/Gas, RBC/Membrane) were corrected for hemoglobin levels in participants [26].

The lungs were segmented from CT images using the Chest imaging platform [27, 28] in 3D Slicer. For lung density analysis and artery ratios, coronal CT images with 2 mm or 3 mm slice thickness, in‐plane resolution between 0.60 × 0.60 mm² and 0.95 × 0.95 mm², and medium reconstruction kernel were used. The pulmonary artery to ascending aorta (PA/AA) ratio was calculated by manually measuring the maximal diameters of the main PA and AA at the level of the PA bifurcation. Whole‐lung masks were used to calculate mean lung density and high attenuation area percent (HAA%; percentage of the lungs with −600 < intensity < −250 H.U.) [29, 30]. For vascular analysis, axial CT images with 0.6 mm slice thickness, in‐plane resolution between 0.59 × 0.59 mm² and 0.88 × 0.88 mm², and sharp reconstruction kernel were used. Vascular masks were generated using TotalSegmentator [31]. Vasculature masks were quantified to evaluate the total blood volume (TBV) and the volume of vessels with a cross‐sectional area less than 5 mm² (BV5), between 5 and 10 mm² (BV5‐10), and greater than 10 mm² (BV10) [32]. The ratios BV5/TBV and BV10/TBV provide measures of vascular pruning that may be relevant in the context of SSc lung disease. CT and Xe‐MRI features were compared on the whole‐lung level.

2.4. Statistical Analysis

Statistical analysis was performed using MATLAB 2023b (Mathworks, Natick, MA). Because of the known sex‐based differences in Xe‐MRI gas exchange measures, male and female participants were analyzed separately [33, 34]. Statistical analysis is limited to female participants due to small numbers of male participants enrolled. Male participant data is presented descriptively throughout.

Demographic features, lung function testing, and Xe‐MRI quantitative measures were tested for normality using Lilliefors Tests. Descriptive statistics were used to summarize measures as mean ± standard deviation (normally distributed) or median (interquartile range; IQR) (not normally distributed). Values were compared between the female healthy and SSc groups using t‐tests for normally distributed continuous data, Wilcoxon tests for non‐normally distributed continuous data, and Fisher's exact tests for categorical data. To further explore relationships between clinical measures and Xe‐MRI, SSc patients were placed in subgroups based on disease extent reported in each patient's medical record (limited vs. diffuse) and presence of interstitial features on CT (nonspecific/normal vs. NSIP/UIP) [35]. Xe‐MRI features within these subgroups were tested for normality using Lilliefors tests and compared between subgroups using parametric or non‐parametric testing as appropriate. Other possible subgroupings (e.g., PH‐status) were not examined due to imbalanced numbers between categories.

Xe‐MRI measures were correlated with PFTs, including %predicted FVC, %predicted diffusing capacity of the lungs for carbon monoxide (DLCO), and FVC%/DLCO%, a measure sometimes used in SSc to discriminate between impairment to gas exchange due to vasculature vs. due to restriction [36, 37]. To assess the relationship between pulmonary hemodynamics and Xe‐MRI, Xe‐MRI measures were correlated with pulmonary artery systolic pressure (PASP) estimated from clinical echocardiography (when available). Similarly, Xe‐MRI measures were correlated with basic quantitative CT measures: HAA, mean lung density, and vascular pruning measures (exploratory due to limited numbers). Correlation analysis was performed only for female SSc participants using Pearson's correlation using a 2‐sided hypothesis test.

For all statistical testing, p < 0.05 was considered significant.

3. Results

A total of 25 participants (21 female) with SSc were initially imaged for this study. There were 3 participants who were excluded due to a revised diagnosis away from SSc to a different autoimmune condition between the time of imaging and analysis. An additional 1 participant was excluded due to an inability to access clinical data including PFT and CT imaging. Participant flow is visualized in Figure 1. After exclusions, a total of 18 female participants with SSc (Age 60.1 ± 10.8 years) were available for analysis. A total of 13 healthy female participants (Age 59.3 ± 11.9 years) were enrolled in the study. Based on our analytical strategy, the 3 male SSc participants (Age 53, IQR: 22.5) and the 3 male healthy volunteers (Age 37, IQR: 18.8) were not statistically analyzed but are presented descriptively throughout. CT images for density analysis were available in 17/18 female participants. Xe‐MRI was completed at a median of 36 days removed from CT imaging for these participants (Range 0 to 601 days). CT imaging was completed at a median of 15 days (Range 4 to 55 days) removed from Xe‐MRI. Xe‐MRI was completed at a median of 36 days removed from pulmonary function testing (Range 0 to 250 days). Approximate 1‐year follow‐up PFTs were available in 16/18 participants, with a median of 354 days following MRI (Range 198 to 795 days).

FIGURE 1.

Consort diagram demonstrating the flow of patient enrollment. Around 25 patients were initially recruited for Xe‐MRI imaging. Three patients were excluded following chart review revealing that participants were more accurately diagnosed with alternative autoimmune conditions between the time of imaging and analysis. One additional patient was excluded due to an inability to access clinical data including PFTs and CT imaging. Three remaining male subjects were additionally excluded from primary analyses due to small numbers. Xe‐MRI features are sex‐dependent, and thus including males in the statistical analysis would confound results. Ultimately, 18 participants were included in the primary analyses for this study.

Age (p = 0.849), BMI (p = 0.477), and smoking status (p > 0.999) were not different between SSc and healthy groups (Table 1). SSc participants had a significantly lower FVC than healthy volunteers (80.0, IQR: 19.0 vs. 104.0, IQR: 8.8). The SSc group was heterogeneous in disease presentation and severity. 50% of the population exhibited interstitial abnormalities on CT imaging, while the remaining 50% exhibited normal CT based on radiologist report. Similarly, 50% of the group had forced vital capacity (FVC) < 80%, and 50% had diffusing capacity of the lung for carbon monoxide (DLCO) < 80%. A total of 11 of 18 female participants (1 of 3 male participants) had limited cutaneous SSc, with the remainder having diffuse cutaneous SSc. Five total SSc participants (4 female, 1 male) had CT evidence of PH (PA/AA > 1.0), and 5 (4 female, 1 male) had a clinical diagnosis of PH (2 classified as WHO group 1 (PAH); 3 unclassified). A majority of SSc participants (15/18 females, 2/3 males) were on medication to treat SSc. At approximate 1‐year follow‐up, FVC% changed by an average of 1.9% (Range −9% to 14%), and DLCO% changed by an average of −1.5% (Range −14% to 6%). This small change in PFTs over ~1 year suggests stable disease in these participants.

TABLE 1.

Demographic and Clinical data for Participants enrolled in this study.

	Healthy volunteers (male) (n = 3)	SSc (male) (n = 3)	Healthy volunteers (female) (N = 13)	SSc (female) (N = 18)	p values (female, healthy vs. SSc)
Demographics
Age (years)	37.0 (18.8)	53.0 (22.5)	59.3 ± 11.9	60.1 ± 10.8	0.849
BMI (kg/m2)	27.3 (2.3)	30.0 (6.8)	27.5 ± 6.9	25.8 ± 5.8	0.477
Smoking Status—N ever smoker (%)	1 (33%)	1 (33%)	5 (38%)	7 (39%)	> 0.999
Years w SSc Dx	NA	3.0 (2.2)	NA	4.0 (9.5)	—
Extent of Dx—N Limited (%)	NA	1 (33%)	NA	11 (61%)	—
PFTs
FVC (% predicted)	99.5 (41.0) a	71.0 (18.8)	104.0 (8.8) a	80.0 (19.0)	< 0.001
TLC (% predicted)	NA	71.0 (14.2)	NA	77.0 (12.8)	—
DLCO (% predicted)	NA	78.0 (39.0)	NA	76.0 (34.0)	—
FVC%/DLCO%	NA	0.91 (0.28)	NA	1.15 (0.53)	—
CT Characteristics
Pattern (NSIP, UIP, Nonspecific/Normal)—N (%)	NA	NSIP = 1 (33%) Nonspecific/Normal = 2 (67%)	NA	NSIP = 7 (39%) UIP = 2 (11%) Nonspecific/Normal = 9 (50%)	—
HAA%	NA	6.6 (2.2)	NA	7.1 (3.1)	—
PA/AA Ratio > 1.0—N (%)	NA	1 (33%)	NA	4 (22%)	—
PH—n with disease (%)	NA	1 (33%)	NA	4 (22%)	—
Autoantibody Profile
ANA—N positive (%)	NA	2 (67%)	NA	15 (83%)	—
RNP—N positive (%)	NA	0 (0%)	NA	1 (6%)	—
SS‐A—N positive (%)	NA	0 (0%)	NA	3 (17%)	—
SS‐B—N positive (%)	NA	0 (0%)	NA	0 (0)	—
Centromere—N positive (%)	NA	1 (33%)	NA	6 (33%)	—
RNAPol III—N positive (%)	NA	0 (0%)	NA	3 (17%)	—
Scl‐70—N positive (%)	NA	2 (67%)	NA	3 (17%)	—
Therapeutics b	NA	MMF = 2 (67%) Other = 1 (33%) None = 1 (33%)	NA	MMF = 7 (39%) Other = 8 (44%) None = 3 (17%)	—
Xe MRI Data
VDP (%)	2.9 (1.3)	5.0 (4.4)	7.4 (8.5)	9.1 (14.7)	0.535
Membrane uptake (%)	0.84 (0.08)	0.75 (0.51)	0.79 ± 0.14	0.91 ± 0.29	0.129
RBC transfer (%)	0.41 (0.08)	0.25 (0.13)	0.28 ± 0.07	0.22 ± 0.07	0.035
RBC/membrane	0.48 (0.14)	0.24 (0.25)	0.34 (0.05)	0.25 (0.10)	< 0.001
RBC defect percent (%)	3.3 (5.2)	30.6 (16.0)	11.9 (13.7)	22.8 (16.9)	0.004
RBC oscillation amplitude (%)	9.2 (0.0)	5.4 (6.6)	10.0 ± 3.2	10.2 ± 3.1	0.901

Note: Data from male participants is presented descriptively, but statistical analysis was not performed due to small numbers. Normality was tested using Lilliefors test. Categorical measures are reported as N (%). Normally distributed measures are reported as mean ± standard deviation. Non‐normally distributed measures are reported as median (interquartile range). p values compare across healthy and SSc female participants and were calculated using t‐tests for normally distributed continuous data, Wilcoxon test for non‐normally distributed continuous data, or Fischer's Exact test for categorical data. Bold p values indicate statistically significant difference.

Abbreviations: ANA, antinuclear antibody; BMI, body mass index; DLCO, diffusing capacity for carbon monoxide; Dx, diagnosis; FVC, forced vital capacity; HAA, high attenuation area; MMF, Mycophenolate Mofetil; PH, pulmonary hypertension; ratio, RBC to membrane ratio; RBC Defect %, red blood cell defect percent; RBC, red blood cell transfer; RNAPol III, anti‐RNA polymerase III; RNP, anti‐ribonucleoprotein antibody; Scl‐70, anti‐topoisomerase I antibody; SS‐A, anti‐Sjogren‐Syndrome‐related‐antigen A; SS‐B, anti‐Sjogren‐Syndrome‐related‐antigen B; SSc, systemic sclerosis; std, Standard deviation; TLC, total lung capacity; VDP, ventilation defect percent.

^a N = 1 male healthy volunteers and N = 4 female healthy volunteers did not have spirometry available due to malfunctioning equipment during the imaging visit.

^b Some participants were on more than one therapeutic, so numbers do not necessarily add to 100%.

In SSc group, serology was similarly heterogeneous with ANA positivity predominance (15/18 female and 2/3 male) and ~1/3 positivity of anti‐centromere antibodies in both male and female (6/18 female and 1/3 male). Other antibodies were not commonly positive (Table 1).

Representative Xe‐MRI images from SSc participants are shown in Figure 2. Xe‐MRI features were compared across healthy and SSc groups (female only) (Figure 3). RBC oscillation amplitude was not different between SSc and healthy participants (10.2 ± 3.1 vs. 10.0 ± 3.2; p = 0.901). Membrane uptake appeared to be mildly elevated in SSc participants as compared to healthy volunteers (0.91 ± 0.29 vs. 0.79 ± 0.14; p = 0.129), but this difference was not statistically significant. RBC/Membrane (0.25 (0.10) vs. 0.34 (0.05)), RBC Transfer (0.22 ± 0.07 vs. 0.28 ± 0.07), and RBC defect percent (22.8% (16.9%) vs. 11.9 (13.7)) were significantly different across groups.

FIGURE 2.

Representative Xe‐MRI scans for one healthy participant and four SSc participants. Images are binned based on signal intensity ratios, with red and orange indicating values lower than a healthy reference population and blue and purple indicating values higher than a healthy reference population (as reported in Leewiwatwong et al.). A representative CT coronal slice is also provided for each of the systemic sclerosis participants. The percentage of the lungs binned to defect (red), low (orange), and high (blue/purple) are shown in boxes on the right of each panel. Notably, many of the participants in this study demonstrate minimal radiographic disease on CT imaging but substantial Xe‐MRI RBC abnormalities. Only the far‐right patient was identified as having an NSIP pattern, while all others were noted as having “normal” CT features. None of the participants shown have diagnosed pulmonary hypertension.

FIGURE 3.

Comparison of Xe‐MRI markers between healthy and SSc participants. Data for male SSc and healthy subjects are displayed descriptively on the righthand side of each boxplot but not included in statistical calculations. Overall membrane uptake (A) and RBC oscillation amplitude (E) were not significantly different between healthy and SSc participants. However, RBC transfer (B) and RBC/Membrane Ratio (C) were significantly reduced, and RBC defect percent (D) was significantly increased in SSc participants.

In subgroup analysis, Xe‐MRI measures were not found to significantly differentiate between either diffuse (N = 7)/limited (N = 11) disease extent or CT abnormalities present (N = 9)/not present (N = 9) (See Table 2).

TABLE 2.

Xe‐MRI measures within SSc extent (Diffuse vs. Limited) and CT abnormalities (present vs. not present) subgroups.

Subgroup analysis
SSc extent	Diffuse (N = 7)	Limited (N = 11)	p
RBC/membrane	0.26 ± 0.12	0.24 ± 0.07	0.688
Membrane uptake	0.77 ± 0.12	0.89 ± 0.19	0.198
RBC transfer	0.21 ± 0.08	0.23 ± 0.08	0.637
RBC defect percent	29.8 ± 13.3	24.2 ± 12.7	0.398
Oscillation amplitude	10.6 ± 2.8	9.1 ± 3.6	0.385

CT abnormalities	Present (N = 9)	Not present (N = 9)	p
RBC/membrane	0.25 ± 0.08	0.25 ± 0.10	0.869
Membrane uptake	0.89 ± 0.17	0.82 ± 0.19	0.410
RBC transfer	0.23 ± 0.08	0.22 ± 0.08	0.770
RBC defect percent	26.7 ± 13.1	25.4 ± 13.2	0.843
Oscillation amplitude	11.1 ± 2.7	8.3 ± 3.5	0.091

Note: Xe‐MRI measures were not significantly different between subgroups.

To visualize correlations between Xe‐MRI measures and clinical variables, a correlation matrix was generated (Figure 4). Similar to previous Xe MRI studies in ILD [12, 15] there was a moderate significant correlation between DLCO% and RBC/membrane (N = 17; R = 0.70). A moderate significant correlation was observed between FVC% and membrane uptake (N = 18; R = −0.61). FVC%/DLCO% exhibited correlations with RBC‐related Xe‐MRI features, including RBC/membrane (N = 17; R = −0.66), RBC defect percent (N = 17; R = 0.60), and RBC transfer (N = 17; R = −0.61). There were no significant correlations between PASP and Xe‐MRI features (N = 12). Membrane uptake correlated strongly with CT density measures (HAA%: N = 17; R = 0.81, mean lung density: N = 17; R = 0.80), while RBC‐related measures were not correlated with CT density.

FIGURE 4.

Correlation Matrix for Xe‐MRI, PFTs, and CT findings. Metrics from Xe‐MRI, pulmonary function testing, echocardiography, and mean lung density measures from CT images. Bold correlation coefficients indicate significant p‐values, and the number of data points for each correlation is shown below the correlation coefficient. Membrane uptake correlated with FVC and CT density measures. RBC transfer, RBC/Membrane, and RBC defect percent correlated with FVC%/DLCO%, which supports an interpretation of Xe‐MRI RBC‐related features as being reflective of pulmonary vascular health. No Xe‐MRI features were strongly correlated with pulmonary artery systolic pressure (PASP) measured from echocardiography.

Most female SSc participants (17/18) had CT images available that were suitable for density analysis. HAA% was relatively low in this population (median 7.1, IQR: 3.1). Around 13 of 17 had HAA% of less than 10%, a threshold which has been reported to be sensitive and specific to clinically relevant abnormalities [38]. Only 6 participants had CT images that were suitable for measurement of vascular features based on the clinical CT protocol and reconstruction kernel available within the medical record (Figure 5). Participants exhibited a median small vessel volume fraction (BV5/TBV) of 0.58 (range 0.41–0.62), which is very similar to published values for healthy volunteers using similar methods [39]. In this limited sample (N = 6), RBC Transfer had a strong correlation with BV5/TBV (R = −0.85) and BV10/TBV (R = 0.87).

FIGURE 5.

CT Vascular analysis and Xe‐MRI. In a subset of SSC participants (N = 6), CT images had adequate quality to generate high quality vascular segmentations for quantification. A and B show representative vascular maps and representative coronal slices of RBC transfer maps for two SSc participants. C shows a correlation matrix comparing Xe‐MRI to vascular measures from CT imaging. In this limited dataset, RBC transfer correlates strongly with small (BV5/TBV) and large (BV10/TBV) vessel volume. Bold coefficients indicate significant p‐values, and the number of data points for each correlation is shown below the correlation coefficient.

4. Discussion

Given that Xe‐MRI is highly sensitive to the functional abnormalities that are common to the expected pulmonary pathophysiology in SSc, we sought to test the utility of Xe‐MRI for characterization of SSc pulmonary involvement in a larger and more clinically characterized cohort of SSc patients than that previously reported [40]. Despite efforts to recruit an adequate population of male SSc patients for this study, we were only able to recruit 4 male participants and thus discussion and conclusions are specific only to female SSc patients.

Consistent with the expected pathophysiology in SSc lung disease, ventilation measured using Xe‐MRI was largely normal in this population of SSc participants. While many participants exhibited some degree of ventilation heterogeneity, it was typically in the ranges expected for older adults [41]. However, gas exchange Xe‐MRI revealed a high burden of abnormalities in gas exchange, most notably in features related to RBC transfer (RBC transfer, RBC/Membrane ratio, and RBC defect percent).

Contrary to prior Xe‐MRI studies in ILD, elevated membrane uptake was uncommon in this SSc cohort. This could be due to the relatively low burden of interstitial abnormalities in the group, with most patients exhibiting only a small HAA. The participant with the most extensive interstitial abnormalities also had the highest membrane uptake, which correlated strongly with CT mean lung density, though the lack of patients with extensive interstitial disease limits the conclusion about this relationship. No significant membrane signal difference was found between CT‐abnormal and CT‐normal participants, consistent with prior reports showing no clear association between interstitial abnormality extent and membrane uptake [12, 14, 15]. The current study data show no significant difference between membrane signal in patients with radiologist‐identified interstitial abnormalities and those with normal‐appearing CT images.

While the mechanism is not fully understood, mounting evidence suggests that elevated membrane uptake reflects an active disease process rather than solely tissue density [15, 42, 43], and may be predictive of prognosis or treatment response [14, 44]. In this SSc cohort, FVC% and DLCO% remained stable over ~1 year follow‐up, so the lack of elevated membrane in this population could be reflective of stable disease. Whether dominantly normal membrane is characteristic of SSc or simply of mild, stable disease requires study in more severe cohorts.

In contrast to membrane uptake, RBC transfer was reduced in many SSc participants. A likely explanation is vascular damage, which is proposed as the primary origin of SSc lung disease [45]. Such directionality of damage would be in accordance with the observed reduced RBC signal with no associated membrane increase, particularly given the relatively mild interstitial disease in these patients. In further support of this vascular hypothesis, RBC‐related Xe‐MRI features strongly correlated with FVC%/DLCO% [36, 37]. However, CT findings from this study do not seem to support this hypothesis: while RBC transfer correlated with small blood vessel volume (BV5/TBV) and large vessel volume (BV10/TBV), the direction of these correlations (negative for BV5, positive for BV10) was opposite to what would be expected if RBC transfer is reduced with reduced small vessel volume. It is likely that the small sample size available for this correlation analysis contributes to the unexpected direction of correlation, though it is possible that these findings are due to changes to pulmonary capillaries beyond the resolution of CT that do not track with BV5. Alternative explanations for the observed imaging features could include subtle interstitial lung disease or combined interstitial/microvascular abnormality.

Finally, the lack of difference in RBC oscillation amplitude between SSc participants and healthy volunteers was at first unexpected. SSc patients commonly exhibit loss of small vessels in the lungs, similar to pulmonary arterial hypertension (PAH) patients [46]. It was expected that SSc patients, particularly those with SSc‐PH, would exhibit reduced RBC oscillation amplitude, as is commonly observed in PAH [23, 47]. However, SSc‐PH patients in this study did not exhibit a marked reduction in RBC oscillation amplitude, which was corrected for estimated capillary blood volume and thus should not be confounded by reduced RBC transfer [23]. Likely, the etiology of PH in SSc is sufficiently distinct from isolated PAH that RBC oscillation amplitude is unaffected. Specifically, upstream impedance to blood flow in the pulmonary arterioles and reduced capillary compliance are thought to drive reduced oscillation amplitude in PAH [23]. In contrast, SSc‐PH patients typically exhibit small vessel intimal fibrosis, some degree of confounding interstitial fibrosis, and both upstream and downstream impedance to pulmonary blood flow [48, 49]. Fibrosis and downstream impedance both could result in elevated oscillation amplitude. The distinctive pathophysiology of SSc‐PH is very likely the explanation for the oscillation amplitudes in the normal range observed in this study. Additional work in a larger number of patients will be necessary to understand whether Xe‐MRI RBC oscillation amplitude will have sensitivity to pulmonary hypertension in this population.

5. Limitations

This study has several limitations. First, the female predominance of SSc prevented statistical analysis in males, so generalizability across sexes is uncertain; given that male SSc patients often have more severe disease [50], validation in a larger cohort including male patients is needed. Second, the single‐site, single‐scanner, single‐vendor, 3 T design limits generalizability across platforms. Third, the lack of standardized CT protocol limits quantitative analysis and regional CT and Xe‐MRI correlation. The CT quantification used may be suboptimal to differentiate between different ILD characteristics. In addition, supporting standard clinical tests were collected from the medical record, and thus there was variability in the time between pulmonary function testing, CT imaging, echocardiography, and Xe‐MRI. There may have been changes to structural or functional measures in the intervening time between Xe‐MRI and clinical measures. These limitations, as well as the limited sample size, prevent firm conclusions regarding the mechanisms of functional impairment observed on Xe‐MRI. Finally, healthy volunteers did not have complete pulmonary function testing or CT imaging available, so these standard clinical measures could not be compared across groups.

6. Conclusion

In this dominantly female SSc cohort, Xe‐MRI revealed that impairments to RBC‐related features (namely RBC transfer, RBC/Membrane ratio, and RBC defect percent) were the hallmark features of SSc despite relatively mild pulmonary disease, whereas membrane uptake remained largely normal. These findings support that SSc lung disease has a primarily vascular origin and suggest Xe‐MRI is a promising technique for identifying subclinical vascular impairment in SSc lung disease.

Funding

Research reported in this publication was supported by the National Scleroderma Foundation and by the American Heart Association (Career Development Award 930177).

Conflicts of Interest

P.J.N. is a consultant for Polarean Imaging, Plc. C.S.H. is a consultant for Polarean Imaging. M.C. sits on an advisory board for Polarean Imaging, Plc.