COPD Patients Have Short Lung Magnetic Resonance T₁ Relaxation Time

Abstract

Magnetic resonance imaging (MRI) may provide attractive biomarkers for assessment of pulmonary disease in clinical trials as it is free from ionizing radiation, minimally invasive and allows regional information. The aim of this study was to characterize lung MRI T₁ relaxation time as a biomarker of chronic obstructive pulmonary disease (COPD); and specifically its relationship to smoking history, computed tomography (CT), and pulmonary function test (PFT) measurements in comparison to healthy age-matched controls. Lung T₁ and inter-quartile range (IQR) of T₁ maps from 24 COPD subjects and 12 healthy age-matched non-smokers were retrospectively analyzed from an institutional review board approved study. The subjects underwent PFTs and two separate MR imaging sessions at 1.5 tesla to test T₁ repeatability. CT scans were performed on the COPD subjects. T₁ repeatability (intraclass correlation coefficient) was 0.72 for repeated scans acquired on two visits. The lung T₁ was significantly shorter (p < 0.0001) and T₁ IQR was significantly larger (p = 0.0002) for the COPD subjects compared to healthy controls. Lung T₁ significantly (p = 0.001) correlated with lung density assessed with CT. Strong significant correlations (p < 0.0001) between lung T₁ and all PFT measurements were observed. Cigarette exposure did not correlate with lung T₁ in COPD subjects. In conclusion, lung MRI T₁ mapping shows potential as a repeatable, radiation free, non-invasive imaging technique in the evaluation of COPD.

Keywords:

• biomarker
• chronic obstructive pulmonary disease (COPD)
• longitudinal relaxation time
• lung imaging
• magnetic resonance imaging (MRI)
• smoking

Introduction

Chronic obstructive pulmonary disease (COPD) is a heterogeneous lung disease that is a major cause of morbidity and mortality. The Global Obstructive Lung Disease (GOLD) guidelines characterize COPD by limitation of airflow, tissue destruction and an abnormal inflammatory response, most often caused by cigarette smoking (1). Characterization of COPD relies on spirometry based pulmonary function test (PFT) measurements, such as forced expiratory volume in 1 s (FEV₁). PFTs are inexpensive, relatively easy to perform and therefore commonly used in clinical trials. However, these methods only measure global lung function, resulting in a loss of sensitivity in early/mild disease (2, 3). The discrepancy between spirometric measurements and COPD symptoms, as well as the nonspecific differentiation of lung disorders, is motivating the development of novel techniques to obtain regional lung function information and improved disease characterization (1, 4–6). Improved disease characterisation will allow the use of personalised medicine approaches to COPD treatment, an emerging field in which imaging biomarkers are likely to play an important role (7).

Regional biomarkers of emphysema in COPD lungs can be obtained from computed tomography (CT) (8) or single-photon emission computed tomography (SPECT) (9). The clinical benefits of CT and SPECT for diagnosis of COPD clearly outweigh the potential harmful effects due to ionizing radiation. However for clinical trials, particularly those including a placebo cohort, repeated exposure to ionizing radiation needs to be considered carefully given that there may be no clinical benefit of the examination to the patient. Magnetic resonance imaging (MRI) may provide attractive biomarkers for assessment of pulmonary disease in clinical trials as it is free from ionizing radiation, minimally invasive and allows regional information (10–12).

Lung MRI has well-known limitations because of the low density of the lung and the fast signal decay due to susceptibility differences between tissue and air in lung parenchyma. Nevertheless, several lung MRI applications have been developed and interest in MRI of the lungs has recently increased (10–12). One technique is oxygen-enhanced MRI, in which regional and global lung function can be assessed by measuring MR signal before and after a challenge with oxygen (13–17). The lung water proton longitudinal relaxation time (subsequently called T₁), a tissue specific MRI biomarker, is affected by interactions of water with macromolecules (18) and paramagnetic agents such as molecular oxygen (13). Moreover, T₁ measurements of the lung have been used to enable lung perfusion measurements using MRI contrast agents and to measure regional partial pressure of oxygen in the alveolar airspaces using hyperpolarised gases (10, 16, 19).

However, only a small number of studies have published data on changes in native T₁ in emphysema and fibrosis patients relative to healthy controls (20, 21). Little is known about lung T₁ for different GOLD stage defined COPD patients, especially in relation to well-established markers such as smoking history, CT and PFTs. In contrast to other studies of T₁ in the lungs (20, 21), we have enrolled subjects with both moderate and severe COPD, have age-matched controls, included smoking history, added CT lung density scans and complete PFT information.

The objective for our study, therefore, was to test T₁ repeatability and characterize the lung T₁ relationship to smoking history, CT density scans and systematically performed PFTs in moderate and severe GOLD stage defined COPD subjects and healthy age-matched ­controls.

Materials and Methods

Subjects and study design

Lung T₁ measurements from 24 COPD subjects and 12 healthy age-matched non-smokers (Table 1) were retrospectively analyzed from an institutional review board-approved single center lung MRI study conducted in Manchester, United Kingdom. All COPD subjects were current or former smokers; 12 subjects with moderate COPD, defined by FEV₁ > 50% predicted (GOLD stage 1–2) (22), and 12 with severe COPD, FEV₁ < 50% predicted (GOLD stage 3–4). A week prior to MRI scanning, all subjects underwent PFTs, whereas only the COPD subjects were scanned using CT, in order to not expose the healthy participants to ionizing radiation. Subsequently, a second lung MRI T₁ measurement was performed one week later to test T₁ repeatability.

Table 1.  Demographic, pulmonary function, CT density and lung MRI T₁ data

	Healthy	Moderate COPD	Severe COPD
No. of subjects	12	12	12
No. of men	8	9	10
Age (years)	63 ± 12 (42–79)	66 ± 9 (49–80)	65±6 (52–74)
Body mass index (kg/m^2)	25.3 ± 4.1 (18.1–32.1)	28.7 ± 4.3 (23.3–38.6)	25.6 ± 3.4 (20.6–32.9)
Smoking index (pack-years)	0.3 ± 1 (0–3.6)	52 ± 26 (17–102)	47 ± 19 (12–91)
Pulmonary function measurement
FEV1 (% predicted)	121 ± 14 (104–147)	69 ± 9 (51–82)	41 ± 6 (26–49)
FEV1/FVC (% predicted)	78 ± 4 (72–85)	51 ± 8 (40–64)	33 ± 5 (25–44)
DLCO/VA (% predicted)	88 ± 11 (71–109)	69 ± 12 (51–90)	48 ± 15 (26–77)
CT density scan
PD15 (g/L)	ND	58 ± 23 (29–89)	30 ± 22 (2–77)
RA950 (%)	ND	14 ± 9 (4–30)	30 ± 15 (6–55)
Lung MRI measurement
T1 (ms)	1053 ± 55 (953–1167)	947 ± 56 (882–1029)†	911 ± 64 (811–991)†
IQR of T1 (ms)	207 ± 45 (155–291)	324 ± 80 (231–460)†	291 ± 62 (182–385)†

Note: Data are means ± standard deviations, with ranges in parentheses. ND = no data.

†Significant difference with healthy group, p < 0.001.

Pulmonary function test

PFTs were performed to assess FEV₁, FEV₁/forced vital capacity (FVC) and the diffusing capacity of carbon monoxide divided by alveolar volume (DL_CO/VA). The measurements were carried out using a computerized spirometer system (V62J SensorMedics Plethysmograph, Viasys Healthcare, Carefusion, UK) by a trained test administrator according to ERS standards (23).

CT protocol

CT was carried out for the COPD subjects to allow quantification of global lung density using a LightSpeed Plus scanner (GE Medical Systems, Amersham, UK). Volumetric images of the lung were acquired at full inspiration with following parameters: 120 kVp, 40 mAs, matrix size 512² and 1.25-mm slice thickness. Images were reconstructed using the GE standard algorithm and the data was measured using the VIDA software package (Vida Diagnostics, Inc., Coralville, IA). The 15th percentile density (PD₁₅) and relative lung area with attenuation values less than −950 Hounsfield units (HU), RA₉₅₀, were calculated as measures of lung density. PD₁₅ is expressed as grams per litre (g/L) by adding 1000 to the HU values.

MRI protocol

MR imaging was performed on a 1.5 tesla Philips Achieva MR system (Philips Healthcare, Best, the Netherlands). Subjects were positioned supine and the body coil was used for radio frequency transmission and reception. Cardiac triggering, respiratory gating and respiratory manoeuvres were not employed, allowing normal tidal breathing. Localization scans were acquired to allow reproducible slice positioning at each visit, selecting a posterior two-dimensional coronal slice placed across the descending aorta to permit good lung coverage and avoid the heart.

Static T₁ mapping scans were performed during air breathing using a two-dimensional half Fourier acquired single shot turbo spin echo (HASTE) sequence preceded by an adiabatic non-selective inversion pulse. One week later the subjects were recalled and the repeatability of the T₁ measurements tested. The imaging parameters were: repetition time (TR) = 5500 ms, echo time (TE) = 
3 ms, field of view (FOV) = 450² mm², matrix size = 128², 68 phase-encoding steps, coronal section with slice thickness = 10 mm, flip angle (FA) = 90°, with inversion times (TI) of 50, 300, 1100, 2000 and 5000 ms and 5 repeats acquired at each TI, resulting in a 4 min T₁ mapping scan time. T₁ maps were calculated on a voxelwise basis from inversion recovery HASTE data by fitting the inversion recovery signal equation

where SI = signal intensity, S0= signal expected without the application of an inversion pulse, f = inversion efficiency set to 1, and n = echo train length set to 68 (24).

MR imaging analysis

All data analysis was performed using Pulmolux:OE software (Bioxydyn Ltd., Manchester, United Kingdom). Image registration was used to correct for breathing-induced motion. The software uses a non-linear diffeomorphic registration algorithm with a neighbourhood cross correlation similarity metric to account for the large elastic deformations in the lung. Lung segmentation was performed using a semi-automatic method where an intensity threshold was defined in order to extract the lungs from the image. The intensity threshold was individual-specific to account for both physiological and technical variation between scans. The segmentation procedure also allowed the large pulmonary vessels to be excluded in the quantification.

All segmentations were verified by a second observer to ensure the correct lung regions were defined for analysis. T₁ was calculated by fitting the inversion recovery signal equation pixel-by-pixel over each lung image and used to calculate the median and inter-quartile range (IQR) of each subject using RStudio (version 0.98.507). Average lung T₁ was compared between groups. Histograms and IQR of T₁ were assessed to reveal the inherent structure in the maps as median T₁ values are poorly suited to report image heterogeneity. Additionally, the signal to noise ratio (SNR), mean lung signal/standard deviation (SD) in background, was assessed for each TI image.

Statistical analysis

The T₁ repeatability was assessed by intraclass correlation coefficient (ICC) and coefficient of variation (SD/mean, CV) analysis. Mean T₁ from the two scanning visits was associated with pack-years (number of years or equivalent years in which 20 cigarettes a day were smoked, PY), CT density (RA_950, PD₁₅) and PFTs (FEV₁, FEV₁/FVC, DL_CO/VA) by Pearson's correlation. Differences in demographic and lung T₁ variables between groups were tested using one-way ANOVA with Tukey's multiple comparison test. All p-values are nominal and 0.05 was used as a significance level throughout the study. If not stated otherwise, the reported values are given as the mean ± one SD. Analyses were performed using RStudio (version 0.98.507).

Results

The means ± SD of demographic, PFT, CT and lung T₁ biomarkers for all participants are shown in Table 1. There was no significant difference in age or body mass index across the three groups. The COPD smokers had smoking histories ranging from 12 to 102 PY; including 18 former and 6 current smokers (mean 49 ± 22 PY). One present healthy non-smoker had a past smoking history of 3.6 PY.

Representative lung T₁ maps from each of the three groups – healthy, moderate and severe COPD subjects– are shown in Figure 1. Heterogeneity of T₁ values is evident in the maps of those with COPD, whereas the T₁ map of a healthy subject appears relatively homogenous. Histograms of T₁ further reveal this heterogeneity with an asymmetric broadening of T₁ in the COPD subjects. Also, the mean T₁ for the COPD subjects appears to be shorter than seen in the healthy subject.

Figure 1.  Representative coronal lung MRI T₁ maps overlaid on a signal intensity image with corresponding normalised T₁ histograms for healthy, moderate COPD and severe COPD subjects. Both T₁ maps and histograms reflected the severity of COPD by reduced T₁ and increased heterogeneity.

Repeated T₁ calculations in 27 participants were usable, 9 were excluded due to poor alignment, missing visits and data quality issues. The median lung T₁ of each subject was satisfactorily repeatable between visits (ICC = 0.72). The mean T₁ difference and CV assessed from the median T₁ maps obtained from the two visits were 5 ms and 2.5%, respectively. Thus, T₁ values were averaged for each individual across the two scanning visits before testing for significant group differences and correlations. Typical mean SNRs in lung for TIs of 50, 300, 1100, 2000 and 5000 ms were 28, 17, 8, 28 and 56, respectively.

The lung T₁ (mean ± SD) was significantly shorter (10%, p < 0.0001; 14%, p < 0.0001) for the moderate COPD subjects (947 ± 56 ms) and severe COPD subjects (911 ± 64 ms) than in the healthy controls (1053 ± 55 ms) (Figure 2a; Table 1). No statistically significant differences of T₁ were found between the subjects with different grades of COPD.

Figure 2.  Global lung T₁ (a) and regional IQR of T₁ (b) for healthy (), moderate COPD (▲) and severe COPD () subjects. T₁ was significantly shorter for the moderate COPD (p < 0.0001) and severe COPD subjects (p < 0.0001) than in the healthy controls. IQR of T₁ was significantly larger for the moderate COPD (p < 0.0001) and severe COPD subjects (p = 0.0002) than in the healthy controls. No statistically significant differences in T₁ and IQR of T₁ were found between the COPD subjects. The error bars indicate standard deviation of the mean.

IQR of T₁ was significantly larger (57%, p < 0.0001, 41%, p = 0.0002) for the moderate COPD (324 ± 80 ms) and severe COPD subjects (291 ± 62 ms) than in the healthy controls (207 ± 45 ms) (Figure 2b; Table 1). No statistically significant differences in IQR of T₁ were found between the subjects with different grades of COPD.

Significant correlations between lung T₁ and CT density measurements; RA₉₅₀ (r = −0.63, p = 0.001) (Figure 3a) and PD₁₅ (r = 0.50, p = 0.017) (Figure 3b) were found, showing that lung T₁ decreases with lower lung density. Moreover, strong significant correlations (p < 0.0001) between lung T₁ and all PFT parameters; FEV₁ (r = 0.74) (Figure 4a), FEV₁/ FVC (r = 0.77) (Figure 4b) and DL_CO/VA (r = 0.75) (Figure 4c) were observed, indicating that lung T₁ decreases with reduced lung function. There was no significant correlation between lung T₁ and PY in the COPD subjects (p = 0.87).

Figure 3.  Lung T₁ as a function of CT density measurements RA₉₅₀ (a) and PD₁₅ (b), for moderate COPD (▲) and severe COPD () subjects. Significant correlations between lung T₁ and RA₉₅₀ (r = −0.63, p = 0.001) and PD₁₅ (r = 0.50, p = 0.017) were found.

Figure 4.  Lung T₁ as a function of PFT parameters FEV₁ (a), FEV₁/ FVC (b) and DL_CO/VA (c), for healthy (), moderate COPD (▲) and severe COPD () subjects. Strong significant correlations (p < 0.0001) between lung T₁ and all PFT parameters; FEV₁ (r = 0.74), FEV₁/FVC (r = 0.77) and DL_CO/VA (r = 0.75) were observed.

Discussion

Data from this study demonstrates significantly shorter lung T₁ in moderate and severe COPD subjects than in healthy age-matched non-smoking controls. Additionally, IQR of T₁ significantly increased for the COPD subjects compared to healthy controls, indicating enhanced heterogeneity in the diseased groups. The T₁ protocol possessed good repeatability between the two visits in COPD and healthy subjects. Moreover, the significant T₁ correlations to CT density and PFT parameters indicate that lung T₁ is a potential biomarker for COPD staging. To the best of our knowledge, no other study has provided evidence of the lung T₁ relationship to PY, CT and PFTs in COPD subjects in conjunction with healthy age-matched controls.

There are a number of reasons why lung T₁ may be reduced in COPD subjects. It could be explained by the presence of tar or other substances which enhance dipolar relaxation in the extracellular tissue water and which accumulate in the lung as a direct consequence of smoking. However, we did not observe any T₁ correlation to PY in the COPD subjects, so therefore it is unlikely that smoke particles from the cigarette smoke is the cause of shorter T₁. Lung T₁ may also be changed due to reduced oxygen concentrations and increased inflammation with edema in the COPD subjects. However, these alterations would be expected to increase T₁ in the lung rather than reduce it as found in our study (13, 25). The observed finding with reduced T₁ in COPD patients, therefore, most likely reflects smoking-induced lung pathology, specifically emphysema which is supported by our PFT and CT measurements. Emphysema may reduce lung T₁ due to the elimination of long T₁ compartments such as water and blood (26, 27) and an increased fraction of fibrotic tissue (20). Fibrotic tissue is known to have short T₁ (28).

Stadler et al. found shorter lung T₁ values for 11 emphysema and 14 fibrosis patients (20) than 10 healthy volunteers (29). Dasenbrook et al. showed a lowered normalized T₁ for 10 cystic fibrosis patients than 5 healthy volunteers with unspecified age (21). Our observed T₁ changes for the COPD subjects are in the same order of magnitude as these studies. The results presented here extend the previous work (20, 21) by using larger study groups, different disease grades, age-matched healthy controls, smoking history information, CT lung density scans and systematically performed PFTs. Moreover, our MRI T₁ protocol allowed normal tidal breathing, while the studies mentioned above performed breath-hold protocols. A free-breathing MRI examination has a practical advantage in diseased subjects that might have difficulties in holding their breath. Moreover, capturing the lung during normal breathing is probably more likely to give representative information about lung function.

Additionally, a previous publication (30) analyzed the same individuals as in this study and retrieved proton density values from the T₁ mapping fitting routine (S₀, Equation 1). Normalized lung proton density strongly correlated with CT density scans (30). The addition of lung T₁ to proton density may provide extra information on the underlying biology. Thus, radiation-free lung MRI T₁ together with proton density quantification would be of considerable interest as an assessment of lung disease especially for longitudinal studies.

A limitation with this retrospective study was the two-dimensional MRI protocol that was restricted to one slice and did not cover the whole lung. A multi-slice or three-dimensional protocol would be preferred for improved regional analysis of the heterogeneous COPD disease. The present study showed good repeatability; however, reproducibility of lung T₁ across sites will need to be determined in a multi-center setting with carefully designed study protocols (31). No correlation between lung T₁ and cigarette exposure was seen in the COPD subjects. However, we did not monitor intervals between last cigarette smoked and the imaging sessions, so there may be some effect of smoke particles on lung T₁. To test this hypothesis a study in healthy smokers should be performed. Other limitations with this study were the small sample size, lack of mild COPD subjects and lack of baseline CT scans from the healthy controls.

The T₁ data was presented as the change of median T₁ and IQR, two quantities which describe the general change as well as the heterogeneity of T₁ in the COPD subjects. As indicated above, the relationship between T₁ and the underlying pathophysiology is not clearly outlined. In future studies it might be appropriate to compare other measures such as the 15th percentile of T₁ to CT PD₁₅, however first we need to increase our understanding of lung T₁. The mean lung density from CT was not used in the comparison with T₁, since it will represent a mixture of larger structures such as bronchi and vessels, emphysema and inflammation and will be of little value to increase the understanding of our findings (32, 33).

This was an exploratory study to investigate a potential new biomarker for COPD. The main application for the imaging biomarker would be in clinical trials where there are needs to perform multiple scans and ionizing radiation would be a limitation. The validation of MRI biomarkers is in some ways more challenging than for more familiar molecular biomarkers (31). For MRI measurements used as biomarkers, the quality and validity of the imaging measurement as a biomarker depends crucially on the use of a diagnostic imaging device, in the presence of the patient, in a manner for which the device was not primarily designed, and may be unfamiliar to the user in the trial site. It is essential to establish repeatability and variation before investigating response to therapy.

In conclusion, MRI T₁ mapping of the lung showed significant differences between COPD subjects and healthy controls, and lung T₁ correlated with CT density and PFTs, indicating the potential role of T₁ quantification in the evaluation of COPD and emphysema. The good repeatability, and the radiation-free and non-invasive nature of MRI, make T₁ mapping an attractive imaging biomarker of COPD for future longitudinal studies.

Acknowledgments

Thanks to the University of Manchester Magnetic Resonance Imaging Facilities and NIHR/Welcome Trust Clinical Research Facility for access to scanning facilities.

Funding

This work was supported by AstraZeneca and Allmänna Sjukhusets i Malmö stiftelse för bekämpande av Cancer.

Declaration of Interests Statement

The authors have no conflicts of interest to declare. The authors alone are responsible for the content and writing of the paper.