Pulmonary CT and MRI Phenotypes that help explain COPD Pathophysiology and Outcomes

Abstract

Pulmonary X-ray computed tomographic (CT) and magnetic resonance imaging (MRI) research and development has been motivated, in part, by the quest to sub-phenotype common chronic lung diseases such as chronic obstructive pulmonary disease (COPD). For thoracic CT and MRI, the main COPD research tools, disease biomarkers are being validated that go beyond anatomy and structure to include pulmonary functional measurements such as regional ventilation, perfusion and inflammation. In addition, there has also been a drive to improve spatial and contrast resolution while at the same time reducing or eliminating radiation exposure. Therefore, this review focuses on our evolving understanding of patient-relevant and clinically-important COPD endpoints and how current and emerging MRI and CT tools and measurements may be exploited for their identification, quantification and utilization. Since reviews of the imaging physics of pulmonary CT and MRI and reviews of other COPD imaging methods were previously published and well-summarized, we focus on the current clinical challenges in COPD and the potential of newly emerging MR and CT imaging measurements to address them. Here we summarize MRI and CT imaging methods and their clinical translation for generating reproducible and sensitive measurements of COPD related to pulmonary ventilation and perfusion as well as parenchyma morphology. The key clinical problems in COPD provide an important framework in which pulmonary imaging needs to rapidly move in order to address the staggering burden, costs as well as the mortality and morbidity associated with COPD.

INTRODUCTION -COPD MEASUREMENTS AND ENDPOINTS ARE URGENTLY NEEDED

Despite decades of research, therapies that modify chronic obstructive pulmonary disease (COPD) progression or mortality are lacking.¹ Despite considerable efforts to discover and develop new COPD interventions, progress has been slow. This is in large part due to limited and suboptimal patient phenotyping that relies on spirometry measurements made at the mouth that cannot account for the regional and inter-subject variability of COPD. Moreover, while COPD is still diagnosed and classified on the basis of symptoms related to the presence of persistent airflow limitation,¹ these measurements correlate weakly with important clinical outcomes.²

A wide variety of imaging methods may be used to study the pulmonary system, including those that rely on tissue absorption of x-ray radiation (chest x-ray and CT), radiofrequency stimulation (magnetic resonance imaging: MRI) or signals generated from injected or inhaled radioactive particles (simple gamma emission projection imaging, single photon emission tomography (SPECT) and Positron Emission Tomography (PET)) and inhaled or injected contrast agents. New insights into the basis for disease initiation and progression using these methods have the potential to break through the current conundrum, which is the fact that image acquisition and analysis tools are slow to be adopted because of a lack of meaningful interventions and yet the interventions are slow in coming because of a lack of understanding disease mechanisms and etiology. Therefore, as shown in Figure 1, we frame this review of CT and MRI measurements and phenotypes as potential solutions to the following major COPD problems: 1) COPD treatments are required that improve outcomes, not just symptoms, 2) a better understanding of COPD disease onset and mechanisms is urgently required in order to better design drugs targeted at underlying pathophysiology, 3) more sensitive measurements are required to better understand the links between ventilation, perfusion and inflammation, and, 4) better predictive measurements of COPD exacerbations and progression are critically needed. Here we summarize and compare MRI and CT tools and measurements of COPD because of their near-universal availability that makes their use in COPD multicenter clinical trials and COPD patient care, both practical and timely. Taken together, MRI and CT methods provide a way to identify the underlying pathologies associated with COPD and previous work has been summarized and previously reviewed^3–5

Figure 1. Current COPD challenges.

Three distinct COPD phases are outlined in schematic: 1) Early disease when patients are asymptomatic, clinical measurements typically do not reflect disease but imaging measurements provide evidence of mild emphysema, airways disease, perfusion heterogeneity, LV filling defects, etc. 2) Mild-moderate COPD as patients become symptomatic, clinical measurements are modestly abnormal while imaging measurements can be markedly abnormal revealing regional disease, LV filling defects can continue to worsen, co-morbidities can begin to appear including aortic aneurysms, coronary disease, lung nodules, osteoporosis, 3) Severe COPD with patients reporting severe symptoms and activity impairment, clinical measurements of airflow limitation, diffusing capacity of carbon monoxide and gas trapping are markedly abnormal and yet patients still can be differentiated into those with predominantly airway or predominantly parenchymal disease with marked differences in the distribution of parenchymal destruction.

PULMONARY CT MEASUREMENTS OF COPD

CT Structural Measurements and Phenotypes

Computer-based methods^6–9 for the objective quantitation of CT images are increasingly used in multi-center studies^10–15 that aim to interrogate phenotype-genotype linkages and identify intermediary endpoints for the assessment of potential interventions. A cornerstone of this approach is the CT lung tissue attenuation or density mask that provides a way to estimate regional COPD-related emphysema. By empirically defining regional emphysematous lung using selected threshold tissue attenuation measurements (Hounsfield Unit: HU) at full inspiration (total lung capacity: TLC), one can count the number of voxels in the whole lung and express this measure. In this way, the percent of voxels reflecting emphysema can be expressed relative to the total volume of the lung or lung region.^16–23 The CT density mask is particularly useful in classifying mild/moderate and severe emphysema^24,25 and has been used in the National Emphysema Treatment Trial (NETT) to identify subgroups of patients who show benefit from lung volume reduction surgery.²⁶ In the left panel of Figure 2, (using Apollo software provided by VIDA Diagnostics, Coralville, IA, USA), the spatially-heterogeneous distribution of emphysema, shown in spheres and regional clustering (sphere size) is shown by lung lobe. Similarly, a density mask may be used on the expiratory dataset (RV or FRC) to identify regions of air trapping as shown in the middle panel of Figure 2. In both the inspiratory and expiratory images, airways may also be segmented and labeled^9,27,28 using commercially-available software such that relationships can be determined between the airway path, airway remodeling features, and parenchymal destruction or peripheral airway closure. CT-based density metrics have been used in numerous large-scale studies^12–15, producing a wealth of literature over the past several years. These measures have identified correspondence of quantitative CT measures of emphysema and air trapping to, for example: genotypes,^29,30 left ventricular (LV) filling,³¹ physiologic measures,^32,33 environmental smoke exposure in childhood as a risk factor for emphysema,³⁴ predictors of bronchoreversibility,³⁵ association of cigarette smoking with sub-clinical disease,³⁶ and trapped gas in severe asthma.³⁷ It is also important to note that in addition to tissue attenuation or density masks, texture analyses may be employed to evaluate clusters of emphysema such³⁸ as estimated using low attenuating clusters. Image analysis methods based on such local feature patterns provide a way to objectively differentiate between these subtypes and some of these methods have better classification rates than expert radiologists³⁹. So-called low attenuation clusters are visually obvious, correlating with histopathology measurements⁴⁰ and more closely reflecting scoring performed by a radiologist⁴¹

Figure 2.

CT Measurements: Threshold-based evaluation of the extemt and distribution of emphysema at full inspiration (far left panel), amount and air trapping at expiration to functional residual capacity (FRC) or residual volume (RV) (middle panels), airway geometry assessed in conjunction with distribution patterns of emphysema and air trapping (middle panels), vascular anatomy (total pulmonary vascular volume and total pulmonary arterial volume assessed from full inspiratory non-contrast enhanced CT scans (far right panel). Colour-coding differentiate between lung lobes.

However, CT quantification of the lung parenchyma is also challenging. Scanner mis-calibration, inconsistent use of reconstruction kernels, differences in reconstruction algorithms between manufacturers and poor coaching of the patient to the targeted lung volume can result in measurement variability.^42–45 Accurate, quantitative CT requires image acquisition consistency⁴⁶ and even then, subject age, sex, race/ethnicity and height influence the normal range of these measures (similar to lung function) as does weight.⁴⁷ Current smoking status has a large and paradoxical effect on these measures,⁴⁷ making precise measurements of smoking status (eg. Cotinine or other objective measure) important in longitudinal studies. Reference equations for both percent emphysema and total lung volume on CT are now available to account for most of these differences.⁴⁷ The limitations due to concerns of radiation dose are being addressed with recent advances in CT technologies, including improvements in the x-ray tube, detector technology, adaptive exposure ^48,49 and iterative image reconstruction,^50–54 leading to clinically-adequate image quality with sub-mSv doses.^48,55 Finally, despite considerable successes using quantitative CT to assess the presence and distribution of emphysema and airway wall remodeling^56–61, critical underlying differences among disease sub-populations continue to emerge. For example, in an apparently discordant result, spatially-matched airway segmentation demonstrated that airway walls may become thinner rather than thicker in COPD, leaving open the possibility that airway wall remodeling may, itself have multiple phenotypes.⁹

Functional CT Measurements and Phenotypes

Pulmonary CT Vascular or Perfusion

It has long been thought that a better understanding of the pathologic response to inflammation in COPD and asthma will be important for the design of new therapies. What is well-understood is that in response to noxious particles and gases in cigarette smoke, the lung reacts by recruiting inflammatory cells. Pulmonary vascular changes, including thickening of the vessel walls, have been characterized early in the history of COPD.^62,63 More recently it was observed that in the presence of inflammation, there is an enhanced delivery of progenitor cells to the lung.^64,65 Remy-Jardin and colleagues recently observed an increased propensity for the lung to develop emphysema in regions of suspected inflammatory processes as defined by ill-defined ground glass opacities and micro nodules.⁶⁶ In more than half of smokers, the inflammatory process was able to resolve itself with repair and maintenance of normal parenchymal anatomy and function. For the remainder (30–45% of smokers) there was parenchymal destruction. There is evidence in the literature suggesting that in humans and animals, hypoxic pulmonary vasoconstriction (HPV) is normally blocked in the presence of inflammation.^25,67–69 For example, Alford et al.⁷⁰ demonstrated that smokers with normal pulmonary function, but small visibly obvious signs of localized, apical centrilobular emphysema, have an increase in coefficient of variation (CV) of CT-based regional pulmonary blood flow (PBF) and mean transit time (MTT), supporting a hypothesis that one etiology of smoking-associated emphysema may be failure to maintain PBF to inflamed lung regions. Recent work published together with an associated editorial^31,71 also demonstrated and discussed the strong relationship between impaired LV filling and percent emphysema in a group of non-smokers, ex-smokers and current smokers. The correlation between impaired LV filling with emphysema was greatest in current smokers compared with previous smokers, consistent with the notion that the inflammatory effects of smoking may determine this effect rather than emphysema alone.

With the observation that pulmonary perfusion heterogeneity may serve as a biomarker in smokers susceptible to centrilobular emphysema,⁷⁰ it has been hypothesized that the lung normally inhibits hypoxic pulmonary vasoconstriction (HPV) in the presence of inflammation. Conversely, in patients unable to block HPV in inflamed lung regions, perfusion is reduced, thus prolonging inflammation and limiting repair mechanisms. With the use of single breath methods in conjunction with dual energy CT, efforts have been directed towards maintaining sensitivity to increased perfusion heterogeneity while simplifying imaging methodologies.

Dual energy CT (DECT) provides a way to double the temporal resolution of CT making it possible to image the entire lung field with a 0.62 mm slice thickness in 0.6 seconds. By having two x-ray guns together with two sets of detector rows, both capable of acquiring 128 slices of image data, there is now the possibility of dual energy CT which allows for sensitive discrimination between tissue types and contrast agents such as iodine for perfusion and xenon for ventilation. At two different peak kilo-voltages (kVp), the reconstructed CT densities for iodine or xenon are shifted significantly between the two resultant image data sets, but the body tissues are not. Thus, because the two image sets were acquired simultaneously, assuring alignment, with a modified form of image subtraction (material decomposition) it is possible to assess perfused blood volume or regional ventilation respectively. For regional assessment of perfused blood volume, when blood is equilibrated with a concentration of iodine (or gadolinium) through the slow infusion of contrast agent, an image of iodine (or gadolinium) can be directly related to regional blood volume.

Other studies have demonstrated the equivalency of perfused blood volume (PBV), as assessed by dual energy CT, and true perfusion. Perfusion blood volume and perfused blood flow (PBF) likely reflect one another because peripheral vascular beds dilate and capillary beds are recruited with increased PBF. PBV may be evaluated using DECT during a slow infusion of x-ray contrast. As shown in Figure 3 modified⁷² from a report on interventions performed in pigs, regional perfusion (colour-coded from red to blue as percent of total perfused blood volume or total perfusion) was perturbed either by incrementally pulling back a balloon catheter placed in a pulmonary artery or by imaging the lung at various static inflation pressures which effectively reduces perfusion in the non-dependent lung regions.⁷² Under all conditions, PBF maps and PBV maps are strikingly similar and a strong relationship for PBF and PBV heterogeneity was also demonstrated.

Figure 3.

Gray-scale (top row), PBV (middle row), and PBF (lower row) MDCT scans. (Columns 1–3) Color map comparison of CT-derived PBF and PBV from pig imaged at 3 different lung volumes, used to achieve a range of pulmonary perfusion values. (Column 4) Color map comparison of CT-derived PBF (dynamic axial scanning) and PBV (dual energy spiral scanning) from pig studied with a balloon partially inflated in a left lower lobe pulmonary artery. Color coding is the same for each condition: percent of total PBV or PBV with low values in blue and high values in red. Modified from ⁷².

Pulmonary CT Ventilation

Xenon gas regional wash-in and wash-out kinetic studies were also explored using CT imaging^73–75 and these methods have been translated clinical studies.^76–82 Importantly, with the more recent development of dual-energy CT, it is possible to simplify the assessment of regional ventilation via use of single breath methods. As shown in Figure 4, by adjusting xenon inhalation of a Xe/O₂ gas mixture, gas flow to the central airways may be monitored and compared with gas flow to the parenchyma. As shown in Figure 4 left panel, the central airway tree is identified by having the subject inhale to TLC, exhale the central dead space volume and re-inhale the same volume of a Xenon gas mixture. This leaves just the central airways filled with Xenon gas. In the right panel, results are shown from of a single inhalation of a Xenon gas mixture. It was also demonstrated that with a slow inhalation of Xenon gas mixtures, geometry and gas density influence or dominate gas distribution causing increased ventilation heterogeneity while with a rapid inspiration, ventilation is more homogeneous.⁸³ Furthermore, the gravity-driven distribution of Xe gas can be eliminated by mixing Xe with helium (He), likely due to the change in the gas density mixture. Another contrast gas agent, Krypton (Kr) while less radio-dense than Xe^84,85, has no anesthetic effects. As multi-spectral CT technologies, including Photon Counting CT^86–88 evolve, it is expected that sensitivity to Kr gas contrast will improve, making it a potential translational/clinical method.

Figure 4.

Dual Energy CT scans of the airway tree (left) and lung parenchyma (right) of an anesthetized pig. For the scan in the left panel, the lung was inflated to 25 cmH2O airway pressure using room air. An amount of air approximately equal in volume to the central airway tree was removed and replaced with xenon gas. This provided a way to identify the central airway tree without the use of more conventional airway segmentation methods. In the right panel, the lungs were inflated from functional residual capacity to total lung capacity via a gas mixture of 80%xenon and 20%oxygen. Material decomposition image processing was used to generate an image representing the regional distribution of the inhaled xenon gas.

As an alternative to inhaled contrast agents, recent studies^89,90 have shown that regional measures of lung ventilation can be assessed using CT images acquired at different volumes without the need for inhaled contrast agents. A two-lung volume (TLC or full inspiration and RV or full expiration) protocol has been standardized as part of “SPIROMICS”¹³ CT protocol that obtains isotropic sub-millimeter images of the entire lungs at TLC and RV. Using this protocol together with advanced image registration methods^90–95, regional maps of lung ventilation can be obtained at spatial resolutions close to the size of a pulmonary acinus. Image matching of an inspiratory/expiratory lung image pair has also been employed⁷ to help differentiate air trapping from emphysema. Functional measurements may be directly generated from expiratory CT^96,97 or a combination of inspiratory and expiratory CT, including those generated using parametric response maps.⁷ All of these important approaches have the potential to test the hypotheses generated using micro-CT⁹⁸ about the pathological mechanisms that accompany the earliest airway and parenchyma changes in COPD.

There is well-developed software commercially available for evaluating the relationship of airway microstructural abnormalities with ventilation and emerging software tools are now being developed to probe the geometry of the pulmonary arterial and venous trees and vascular-structure-function relationships.^99–101 The extracted pulmonary vascular tree from a non-contrast inspiratory MDCT volume of the lung is shown in the right panel of Figure 2, and the volume of this combined arterial and venous tree – the Total Pulmonary Vascular Volume or TPVV has been generated developed using methods that segment the arterial tree.¹⁰¹ The cross sectional area of the pulmonary trunk relative to the aorta has been shown¹⁰² to correlate with acute exacerbations in COPD patients and TPVV (normalized to total lung volume) is currently being used as an upstream marker of down-stream endothelial dysfunction.¹⁰³

PULMONARY MRI MEASUREMENTS OF COPD

Conventional ¹H MRI Structural and Functional Phenotypes

As pulmonary CT continues to advance with new capabilities and lower radiation doses, pulmonary MRI has also advanced to provide complimentary tools for the quantitative evaluation of lung structure and function. However, pulmonary MRI using conventional hardware platforms (¹H methods) is very technically challenging and therefore currently, its clinical use, has been limited. These technical demands stem from the inherently low pulmonary ¹H abundance and corresponding low ¹H signal that can be measured using conventional MRI approaches. Furthermore, the multitude of lung air-tissue interfaces generate significant magnetic field distortions or susceptibility artifacts, further diminishing pulmonary ¹H MRI signal. For these reasons, and until recently, the major applications of conventional pulmonary ¹H MRI included intravenous contrast agents to evaluate pulmonary blood flow and vessel hemodynamics.¹⁰⁴ Methods have also been devised that combine relaxation signals and intravenous contrast that provide a way to differentiate inflammation¹⁰⁵, smooth muscle remodeling, edema and mucus deposition.^104,105 Taken together, these methods provide a way to identify the underlying pathologies associated with COPD. However, because signal intensity is generally very low using ¹H MRI, careful calibration with other organs in the same field of view is required. In addition, because of gravity-dependent changes occur, signal averaging over time may not provide physiologically-relevant information in the dependent lung regions, where atelectasis may occur in minutes while supine.

Another way to address these challenges is to reduce the time (echo time or TE) required to acquire the pulmonary MR signal. As shown in Figure 5, conventional, ultra-short TE (UTE) pulse sequences as suggested 2 decades ago¹⁰⁶ significantly improve pulmonary MR signal so that emphysematous regions can be identified and quantified, based upon the difference between tissue-poor emphysematous bullae and more normal parenchyma. So-called ultrashort-echo or zero echo time MRI methods help address the inherent challenges of low tissue and ¹H density¹⁰⁷ by minimizing the effects of rapid MR signal decay. The relationship between MRI signal intensity and tissue density¹⁰⁸, as previously shown for pulmonary CT, is clearly important for further development of the method. The first studies employing UTE methods reported that the tissue density was related to MR signal¹⁰⁹ and T₂*. UTE MRI was also used to measure signal intensity and T₂* in emphysema¹¹⁰ and showed good correlations with histological measurements while T₂* correlated with pulmonary function measurements¹¹¹ and pulmonary signal intensity was related to tissue density, pulmonary function and CT density measurements.¹⁰⁸ Very recently methods have been developed that exploit optimized¹¹² and so-called zero echo time (ZTE)¹¹³ approaches and this has resulted in excellent, MR image quality and signal in pulmonary images that is very similar to CT. Another method for measuring regional lung function, involves using inhaled oxygen in combination with ¹H MRI¹¹⁴ and this exploits the alteration of lung tissue ¹H relaxation times by molecular O₂. In this manner, wash-in or difference maps can be generated by using the inherent signal differences that stem from breathing room air and pure O₂. While pulmonary ventilation is reflected by O₂ wash-in maps, the alveolar-capillary transfer may be reflected by the O₂-enhancement ratio.¹¹⁵ Like all ¹H-based MRI, O₂-enhanced MRI is limited by the weak proton signal and the fact that O₂-enhanced measurements still require histopathological validation. In fact, for all MRI measurements of COPD, validation of the pathologies directly or indirectly measured is still pending.

Figure 5.

MRI measurements of COPD for different GOLD stages including: emphysema (ADC, ¹H MRI signal intensity), and ventilation (³He MRI and FDMRI).

To tackle the challenge of very low pulmonary ¹H signal intensity, Bauman and colleagues^116–118 proposed another ingenious approach that relies on MRI signal oscillations that occur with the differences in lung volume during normal tidal breathing. It was hypothesized that the ¹H MRI signal oscillations during breathing could be employed to generate both ventilation and perfusion images. They developed a way¹¹⁹ using Fourier decomposition (FD) of oscillating ¹H signal intensity¹²⁰ related to the compression and expansion of the lung parenchyma and blood flow¹²¹ to generate pulmonary ventilation and/or perfusion measurements. Importantly, this method was recently shown in COPD subjects with emphysema ¹²² and contrast can be generated using static volumetric ¹H MRI methods.¹²³ Figure 5 shows very recent examples of Fourier-decomposition of pulmonary magnetic resonance imaging (FDMRI) in COPD subjects across GOLD grades and excellent comparisons were also shown in animal studies.¹²⁴ FDMRI exploits free-breathing 1H MRI and non-rigid registration to generate ventilation images. Importantly, although image registration and analysis is complex, and again dependent on weak ¹H MRI contrast, this approach does not depend on inhaled gas or injected contrast agents and therefore there is strong potential for clinical translation. Again, a more complete validation still needs to be undertaken.

Inhaled ¹⁹F and hyperpolarized ¹²⁹Xe/³He MRI Structural and Functional Phenotypes

Similar to FD and O₂-enhanced ¹H MRI methods, MRI using inhaled ¹⁹F gas (after multiple breaths of perfluoropropane mixed with oxygen), or hyperpolarized noble gases such as ³He and ¹²⁹Xe (after a single breath mixed with ⁴He or N₂) provides a way to visualize pulmonary ventilation by taking advantage of high gas density in the lung. For inhaled ³He and ¹²⁹Xe, increased nuclear polarization generates ventilation images of the airways and airspaces and to measure apparent diffusion coefficients that estimate gas displacement. For ¹²⁹Xe, similar to Xe-CT methods, the fractional solubility of Xe gas in biological tissues¹²⁵ has additional applications for measuring gas exchange, alveolar surface area and perfusion. For all three gases, the nuclear proton provides the MRI signal. In the case of ¹⁹F, MR signal is inherently strong because of the high gyromagnetic ratio of ¹⁹F such that extra polarization (and polarization equipment) is not required. For both ³He and ¹²⁹Xe, however, nuclear polarization is achieved^126,127 using laser polarization equipment. As shown in Figure 5, ³He and ¹²⁹Xe MRI provide superior signal-to-noise ratio ventilation images in COPD subjects, and typically, image quality using ³He MRI is the greatest compared with polarized ¹²⁹Xe and unpolarized ¹⁹F due to larger gyromagnetic ratio and high polarization rates for ³He. For these reasons, currently ³He MRI is most commonly used in research even though the global quantities of ³He are very limited and expensive. Although experience with ¹²⁹Xe and ¹⁹F is still limited,^128,129 these inhaled gas MRI methods provide the strongest translational potential because of the relative abundance and low cost of these gases. Notwithstanding these challenges, recent acute therapy studies in COPD post-salbutamol¹³⁰ and post-COPD exacerbation therapy requiring hospitalization¹³¹ show the potential for inhaled gas MRI ventilation defect measurements to represent the subtle (post-exacerbation) and not so subtle (post-salbutamol) ventilation improvements after therapy. It is important to note that the visibly obvious and statistically-significant improvements in ventilation in these cases were not related to FEV₁ improvements. In addition, in a number of these previous COPD studies¹³², and as shown in Fig. 6 ventilation measured using MRI was directly related to CT-derived measurements of abnormally-remodeled airways and emphysema. Importantly, previous work also showed the relationship of ventilation abnormalities with subclinical disease in smokers with normal lung function.¹³³ Moreover, regional ventilation worsening in COPD has been measured over a follow-up period of two years¹³⁴ even in patients with no change in FEV₁. The Polarized Helium Imaging of the Lung (PHIL) study, was performed in three European centres in 122 COPD patients and is yet the largest COPD MRI-CT multicenter comparison study completed to date.¹³² This important study showed the potential for COPD biomarkers stemming from a comprehensive, prospectively planned, multi-centre and multimodality imaging approach. Another innovative example¹³⁵ showed the potential of ventilation MRI to interrogate the role of regional collateral ventilation in COPD patients with bullous emphysema. Finally MRI measurements of ventilation in mild and moderate patients (GOLD I/II COPD)¹³⁶ were also predictive of COPD exacerbation requiring hospital care, and in these patients, previous exacerbation and FEV₁ were not.

Figure 6.

MR ventilation imaging reflecting the effects of both emphysema (slow filling units) and airways disease (airway obstruction) – a unique predictor of COPD exacerbations in mild disease.

For all inhaled gas MRI methods, diffusion due to random Brownian motion within the lung airways and airspaces can be measured using diffusion-weighted imaging similar to that used in conventional MRI.^{137 3}He or ¹²⁹Xe diffusion-weighted methods provide measurements of parenchyma microstructures, including the alveoli and acini that define the boundaries of the fundamental units for gas exchange.¹³⁸ The ADC map provides quantitative regional airspace measurements information that is in agreement with the presence of emphysematous damage.^139,140 Such measurements are consistent with alveolar changes related to differences in lung volumes¹⁴¹ gravitational dependence^141–143 and ageing.¹⁴⁴ Previous COPD studies have shown that ADC correlates with pulmonary function,^142,145,146 histological measurements of lung surface area¹⁴⁷ and is highly reproducible in COPD,¹⁴¹ while sensitive to subclinical disease^148,149 and disease progression.¹⁵⁰ Novel approaches have also been used to measure diffusion in longer time frames¹⁵¹, providing a way to generate information about acinar duct and airway connectivities, including communication and collateral ventilation.^152,153 Long-range ADC appears to be more than twice as sensitive to parenchymal differences associated with COPD than short-range ADC^154–156. Unfortunately this important information has not yet been exploited in clinical research studies.

Gadolinium-enhanced MRI Pulmonary Perfusion Phenotypes

Given the relative lack of ¹H signal in the lung using conventional approaches, gadolinium-enhanced imaging of first-pass pulmonary perfusion provides sensitive and relatively reproducible measurements of signal increase and its change over time (slope). Fourier transformation of the signal change allows for the assessment of pulmonary blood flow, mean transit time and pulmonary blood volume. These measures can be attained over the whole lung or the lung periphery in order to evaluate microvascular perfusion.^157–159 Pulmonary perfusion deficits were observed more frequently in COPD patients as compared to healthy volunteers; the largest study to date noted largely diminished pulmonary microvascular blood flow in mild to severe COPD with strong spatial correlations in severe disease to centrilobular and panlobular emphysema.¹⁶⁰

CONCLUSIONS

As summarized in Table 1, pulmonary CT and MRI are on the threshold of providing regional, non-invasive measurements of ventilation, perfusion and parenchymal destruction as intermediate endpoints of COPD. However, whereas pulmonary CT measurements have been evaluated in COPDGene,¹² ECLIPSE,¹⁴ MESA Lung,¹⁶¹ CanCOLD,¹⁵ and SPIROMICS,¹³ and cardiac MRI has been utilized in over 5000 participants in MESA, pulmonary MRI has not been exploited in large-scale studies. There are many reasons for this even though pulmonary MRI is rapid, well-tolerated and radiation-free, so it serves as an ideal platform for serial and longitudinal evaluations in patients. More widespread use of all imaging biomarkers has been limited for a number of key reasons including: 1) lack of support to harmonize image acquisition software, 2) universally available image analysis software, 3) regulatory boundaries for emerging approaches, and, 4) historically weak links between respiratory and radiology clinical programs. Not-withstanding these issues, CT and MRI measurements of COPD are being developed to provide a better understanding of disease onset, develop COPD treatments that improve outcomes, and to provide better predictive measurements of COPD exacerbations and progression.

Table 1.

MRI and CT COPD Phenotypes

	Airways Disease	Imaging Biomarkers of COPD
		Emphysema	Perfusion abnormalities
CT	Lumen area Wall area % Pi10 PRM-gas trapping Xe gas ventilation	Low attenuating clusters Low attenuating area RA950 RA856 PRM-emphysema	Total pulmonary vascular volume Iodine perfusion
MRI	Ventilation defect percent Percent ventilated volume	ADC	Gadolinium perfusion Xe perfusion/diffusion

The unique and complementary ability of both MRI and CT to measure disease morphological and functional consequences and explore mechanisms of disease pathophysiology has not yet translated to improved COPD patient care. In COPD, there is an increasing recognition that different phenotypes exist^162–164 and that these patient groups may have different responses to therapy. Moreover as therapies become more targeted and patient-specific, imaging will likely be one of the only ways to verify response and efficacy for an individual patient or group of patients. This will require imaging to become more quantitative, sensitive and accessible to justify its current cost and complexity. Although the risks related to tobacco smoking will decrease over time in the developed world, as the world becomes more industrialized and polluted, and smoking habits change in the developing world, respiratory illnesses will continue to increase in prevalence, morbidity and overall mortality. It is in this clinical context that development and appropriate utilization of pulmonary MRI and CT remain critically important.