Over the past few years, computed tomography (CT) has become the imaging modality of choice for the lung (1). The removal of overlapping structures, the relatively high spatial resolution, and the high signal-to-noise ratio has vaulted CT imaging to the forefront of medical imaging. CT has seen many changes over the years progressing from thick-slice to thin-slice axial scans, to contiguous spiral scans to the multidetector scanners (MDCT) that make it possible to obtain exquisitely detailed images using isotropic voxels in less than a 10-second breath-hold. The volume of the lung can be calculated with good accuracy by summing the lung CT voxel dimensions (Table 1) (2–4). The contiguous thin slices (≤1 mm) obtained using MDCT also permit image reformatting from the axial plane into transverse, coronal, or sagittal planes (see Figure E1 in the online supplement) and facilitate accurate estimation of volume measurements of the individual lobes. The apparent X-ray attenuation values of the lung, measured in Hounsfield units (HU), can also be used to estimate lung density, which can then be combined with lung volume to calculate lung mass, tissue volume, and airspace volume. More information on these calculations is provided in the online supplement.

There are two common CT-based methods for determining the extent of emphysema: the threshold cutoff (2–9) or the percentile point analysis (8, 10–14). The threshold analysis uses a predetermined cutoff value of attenuation values (i.e., HU) to separate emphysematous from “normal” lungs (see Figure E2). The most common threshold point currently in use is −950 HU (6). The second approach is the percentile method where a specific point on the frequency distribution curve of the X-ray attenuation values (i.e., 15th percentile) is defined and compared between subjects or groups of subjects. Although there is no consensus on the most optimal cutoff point for this type of analysis, the Alpha-1 Foundation–sponsored workshop committee recently stated that, for cross-sectional analysis, both the threshold and the percentile techniques are appropriate, but for longitudinal studies, only the 15th percentile technique was endorsed (15).

Emphysematous lesion size can also be estimated by plotting the cumulative number of low-attenuating voxels that are connected to a neighboring low-attenuation voxel (i.e., a cluster of low-attenuating voxels) against the cumulative size of this cluster on a log–log plot (Figure 1) (16–18). This variable has been shown to correlate with survival (17) and exercise ability after lung volume reduction surgery (LVRS) (16) but not very well with pathologic measurements of emphysema (19).

Figure 1. A pictorial representation of emphysema based on quantitative CT scanning. This figure shows a low attenuation cluster analysis. CT voxels below the defined threshold cutoff are grouped if they are touching a neighboring voxel also below the cutoff. The colored areas in this image represent clusters of connected voxels, or low attenuation cluster areas. Image acquired using Pulmonary Workstation 2.0 (VIDA Diagnostics, Iowa City, IA).

Download Figure | Download Powerpoint

[More] [Minimize]

These CT-based measurements are sensitive to the lung volume at the time of analysis. Thus, emphysema estimation must take into account the extent of breath-hold during the CT scan. Although some authors have shown that lung volume is relatively reproducible in CT studies (20), these same authors have also shown that emphysema calculations are dependent on the lung volumes (21). Therefore, it is recommended that all longitudinal studies apply a lung volume correction factor to their analysis (15). More information on these techniques is provided in the online supplement.

The measurement of airway dimensions is a complicated procedure that uses computerized algorithms including the full-width at half maximum (FWHM) method (22, 23), the maximum-likelihood method (24), the score-guided erosion algorithm (25), and an algorithm where ellipses are fit to the airway lumen and wall (26). There is also considerable interest in measuring airway wall dimensions using three-dimensional reconstruction images of the airway tree. This approach has great advantages over the cross-sectional approach (see Figure E3) because it is possible to identify and measure a specific airway at a specific location. For example, with three-dimensional images, it is possible to identify the right apical segmental bronchus (RB1) or the right lateral basal segmental bronchus (RB9) and follow it from its origin to the limits of reconstruction and measure the length of each segment and the angle of branch points (Figure E4). Additional information on the measurement of airways is provided in the online supplement.

Notwithstanding this progress, many challenges remain. For example, the main site of airflow obstruction in COPD is the small airways (i.e., airways <2 mm). However, with the current limits of CT resolution, the small airways are not well resolved, making it difficult to precisely measure wall thickness relative to the luminal size. Additionally, there is no uniform method of measuring airways (15). One metric of airway wall dimensions is the airway wall area (Aaw) percentage, which is the percentage of the cross-sectional area of the whole airway occupied by wall area. However, the Aaw% contains a reference trap in that, as airways become smaller, the Aaw% becomes larger. Therefore, caution must be exercised when using this metric to assure that airway sampling bias does not influence the results. For example, if in particular subjects, large airways are predominately measured, they will appear to have a lower Aaw% (and thus “less” airway remodeling) compared with other subjects in whom only small airways are measured, regardless of the underlying lung pathology. It is for this reason that more robust methods are needed to facilitate the standardization of airway wall sizes (that are measured and reported) across individuals. Examples of new developments in this field include reporting of data only from specifically named airway segments (27) or using a common reference point such as an (idealized) airway with a specific lumen size (i.e., internal perimeter of 10 mm). A summary of the different airway measurement algorithms currently in use is shown in Table 2.

Routine lung function measurements cannot accurately phenotype patients with COPD because both bronchiolitis and emphysema independently cause reductions in lung function (23). Thoracic CT, on the other hand, has the potential to elucidate the relative contributions of bronchiolitis and emphysema to lung function impairment by allowing direct measurements of airway wall thickness (i.e., remodeling) and the extent of emphysema in the same patients at the same time (23, 28). As mentioned above, investigators have been using CT scans to quantify the extent of emphysema in subjects with COPD for many years (9, 14). Most of these early studies were only cross-sectional in nature, likely because of the availability of scans, but recently many investigators have started to use CT to follow subjects with COPD, particularly in those with α-1 antitrypsin deficiency (29–33), and to investigate interventions in these groups (11, 12, 34). Although many of these studies are limited because of small sample sizes, they have shown that CT is more sensitive than pulmonary function tests in detecting disease progression (10, 11, 30, 33), that the distribution of emphysema is important (35), and that airway measurements may be as important as emphysema measurements (31).

It is also thought that, in studies of COPD, improved phenotyping of subjects may be clinically important for several reasons. First, the natural history and clinical manifestations of subjects with COPD who have predominantly bronchiolitis may be different from those who have predominantly emphysema. For instance, patients with predominantly emphysema tend to be functionally more impaired, demonstrate more air trapping, and have less “bronchitic” symptoms. They may also be less responsive to antiinflammatory therapies than those who are not (36). On the other hand, patients with predominantly bronchiolitis may be at increased risk of exacerbations and hospitalizations (37) and more likely to respond to inhaled corticosteroids and bronchodilators (37, 38). Second, because COPD is a heterogeneous disease, CT imaging can help investigators and clinicians to determine the regional predominance of the disease, which can guide therapy. In the National Emphysema Treatment Trial (NETT), LVRS was found to be most beneficial for patients whose emphysema was located predominately in the upper regions of the lung (17). Third, from a methodological standpoint, by reducing the noise of the study population and making the populations more homogeneous, accurate phenotyping can improve the statistical power of clinical studies to evaluate the natural history and the pathogenesis of the disease and to assess the therapeutic responses of new compounds and interventions. Moreover, with better knowledge of phenotypes, drugs designed specifically for each phenotype can be developed.

Although CT is a powerful tool for the analysis of lung structure, it does have some limitations including technical (imaging) parameters of the CT scanner, disagreements on the best method to analyze the lung parenchyma, no definitive study using airway wall algorithms, exposure of subjects to ionizing radiation, and, most importantly, the lack of longitudinal studies involving sufficient numbers of subjects. These imaging parameters such as slice thickness, the reconstruction algorithm, the exposure of the CT scanner (kVp, mA) and the type of CT scanner (number of detectors, manufacturer) increase the noise within the CT image (39–42). Although this noise has been shown to have little effect on the lung volume and mean lung density measurements, it can have a large effect on measurements of emphysema and therefore need to be very carefully controlled in research studies (39, 40, 42, 43). These confounding factors are summarized in Table 1. For these reasons, many large clinical studies, including the Feasibility of Retinoids for the Treatment of Emphysema (FORTE) study (34, 39), the National Heart Lung and Blood Institute Lung Tissue Research Consortium (www.ltrcpublic.com), Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints (ECLIPSE) Study (32), COPDGene (www.copdgene.org), and Alpha-One Studies (21, 29, 42, 44, 45), have all strictly regulated CT scanners and/or imaging protocols and applied lung phantom indices to compare across centers and CT scanners. Experience in the Alpha-1 multicenter studies showed that it is possible to obtain quality data from large multicenter studies when all of these parameters are carefully controlled, and it is hoped that these types of conformities will also improve the quality of data obtained in these other large longitudinal studies.

One other potential limitation of CT scans that deserves special mention is radiation exposure. The risk of radiation exposure at the level delivered in chest CT remains highly controversial (46, 47). However, there are consistent data indicating that population medical radiation dose is now the second largest source of ionizing radiation exposure in developed countries after natural background radiation and CT is the major source of this exposure (48, 49). In light of the controversy regarding the deleterious effect of CT radiation dose, it is prudent to limit exposure. It is generally accepted that radiation effects are much more substantial in younger compared with older individuals due to the latency of radiation-induced cancers (6 to 25 yr), the increased number of dividing cells in younger patients, and the increased life expectancy available to manifest radiation effects. Therefore, based on radiation exposure issues, CT use should be strongly constrained in children, used cautiously in young adults, and used prudently in older adults (49, 50). In general, if the life expectancy of the patient is less than 10 years, radiation-dose concerns become secondary to assessment of disease activity. However, in all cases it is recommended that CT radiation dose be adjusted on the basis of the size of the patient to be as low as necessary to answer the clinical question posed. Minimizing radiation dose in this fashion conforms to the as low as reasonably achievable (ALARA) principle that governs all medical radiation exposure (50). Because the majority of patients with lung disease who were evaluated using CT are older, and harbor chronic life-limiting diseases, radiation risks are minimized. In this patient population, CT-monitoring disease activity or treatment effects generally have a favorable risk benefit profile. Because CT image noise is also dependent on the radiation dose used, clinical questions involving the lung parenchyma may be answered using a very low radiation dose whereas airway analysis may require a higher dose (20, 40).

Magnetic resonance (MR) imaging (MRI) has developed as a research and diagnostic tool mainly because of the unique tissue contrast it provides with high spatial (1 mm in plane [x and y planes] and 1–5 mm out of plane [z plane]) and temporal resolution (∼1 s per slice). Until very recently, for those anatomical regions low in fat or water (low proton or ¹H, density regions) such as in the lung, MR imaging has been much more challenging because of the inherently low ¹H abundance and corresponding low ¹H signal. Furthermore, the multitude of air-tissue interfaces within the lung also create significant magnetic field distortions, commonly described as susceptibility artifacts, which further diminish the lung MR ¹H signal. Moreover, respiratory and cardiac motion during image acquisition degrades pulmonary MR image quality. Whereas respiratory gating and/or rapid breath-hold imaging methods substantially attenuate the effects of motion, low proton density and susceptibility effects together result in significant technological roadblocks that have hampered the mainstream clinical use of pulmonary MR imaging.

Tissue contrast in conventional proton MR imaging depends both on the ¹H density of the tissue as well as the magnetic environment of the protons within that tissue. The latter effect is generally described in terms of both longitudinal (T₁) and transverse (T₂), relaxation times and these are the principal bases for the tissue contrast observed in most MR images. However, in the lung, transverse relaxation times are generally very short (∼ a few ms) due to the presence of microscopic magnetic field inhomogeneities (i.e., susceptibilities due to air–tissue interfaces), significantly reducing the image signal and the potential for image acquisition based upon T₂-weighted approaches. One way to mitigate this effect is to reduce the time required to acquire the MR signal, referred to as the echo time (TE). By using ultra-short TE methods (51) and new projection reconstruction techniques (52), some of these limitations can be overcome.

The major research applications of ¹H MRI in COPD include perfusion methods to evaluate blood flow and the use of intravenous MR contrast agents to evaluate pulmonary vessel hemodynamics (53). Anatomical or morphological imaging of the airways and parenchyma without contrast is much less common, mainly because of the inherent limitations of ¹H MR in the lung and because of the decreased spatial resolution of MRI compared with MDCT. However, to exploit the unique capabilities of MRI tissue contrast, hybrid T₁-and T₂-weighted imaging as well as intravenous contrast enhancement methods have been developed for airway imaging to allow for the differentiation of inflammation versus smooth muscle remodeling, edema, and mucus deposition (53), none of which is possible using MDCT. These methods provide a way to extract some of the underlying pathologies associated with airway disease and airway remodeling in COPD. On a larger anatomical scale, another potential clinical and research use for ¹H MRI is in the analysis of aberrant respiratory biomechanics and dynamics-monitoring the diaphragmatic and thoracic wall motion in patients with COPD (54–56).

T₁ relaxation times in the lung are also very sensitive to the environment of the protons in the lung, including paramagnetic molecules such as molecular oxygen. Oxygen was first proposed as a contrast agent for proton MRI of ventilation in 1996 (57) because of its ability to alter the T₁ relaxation of lung tissue protons. Although molecular oxygen contains two unpaired electrons and is weakly paramagnetic, its effect in the lung is significant due to the enormous surface area of the lung and the large difference in oxygen partial pressure between breathing room air and breathing pure oxygen. Oxygen-enhanced MR image data are viewed as O₂ wash-in time maps displaying the inverse of the wash-in decay constant of molecular oxygen, and as O₂ enhancement maps of the ventilated lung parenchyma displaying the change in enhanced signal intensity after room air images and pure O₂ images are subtracted. However, these images are still derived from the relatively weak proton signal and can be confounded by misregistration of the two image datasets when subtracted. Although mean O₂ wash-in time can be viewed as related to pulmonary ventilation, the mean relative enhancement ratio may be related to alveolar-capillary oxygen transfer (58). In emphysema, inhomogeneous and varying degrees of decreased oxygen enhancement have been shown to reflect decreased diffusion of molecular oxygen to the capillary bed (see Figure 2). In addition, the maximum mean relative enhancement ratio is significantly lower in patients with emphysema compared with healthy volunteers (59). Dynamic oxygen-enhanced MRI has also been shown to correlate with diffusing capacity of the lung for carbon monoxide (Dl_CO) and FEV₁ in COPD. Despite low proton density and susceptibility artifacts, oxygen-enhanced MRI provides a way to evaluate regional, morphological, and functional changes as well as maps of the oxygen diffusing capacity with high potential for clinical stage stratification of patients with COPD.

Figure 2. A comparison of CT and oxygen (O₂)-enhanced MRI images. (A–C) 56-year-old male healthy control subject; (D–F) 72-year-old male patient with emphysema; (A) Axial thin-section CT demonstrating no low attenuation areas in both lungs. (B) Coronal plane thin-section CT demonstrating no low attenuation areas in both lungs. (C) Left panel shows O₂-enhanced MR image and the right panel shows the corresponding relative enhancement map from a control subject. The relative enhancement map from O₂-enhanced MRI demonstrates relatively homogeneous and high oxygen-enhancement in both lungs. On the relative enhancement map, enhancement in each pixel is expressed color-coded with 0 to 50% enhancement progressing from dark blue to red. (D) Axial thin-section CT demonstrating low attenuation areas in both lungs. (E) Coronal thin-section CT demonstrating low attenuation areas in both lungs. (F) Left panel shows O₂-enhanced MR image and the right panel shows the corresponding relative enhancement map from a patient with COPD. The relative enhancement map demonstrates heterogeneous and reduced oxygen enhancement in both lungs with the relative enhancement in each pixel expressed as color-coded with 0 to 50% enhancement ranging from dark blue to red. Images courtesy of Dr. Yoshiharu Ohno, M.D., Ph.D., Kobe University Graduate School of Medicine, Kobe, Japan.

Download Figure | Download Powerpoint

[More] [Minimize]

MRI of inhaled hyperpolarized noble gases, mainly hyperpolarized helium-3 (³He) and xenon-129 (¹²⁹Xe), typically accomplished using a spin exchange optical pumping method (60), provides nuclear polarization up to five orders of magnitude (100,000 times) compared with that achieved using thermal polarization (60, 61). This increased nuclear polarization compensates for the low density of noble gas nuclei within the lung (compared with the abundance of tissue-based protons) and provides ventilation images of the airways and airspaces of the entire lung with 1 mm in plane and 5 to 10 mm out of plane resolution within a breath-hold interval. ³He gas inhalation provides the strongest MR signal due to its larger gyromagnetic ratio, and currently ³He MRI is most commonly used in research even though the global quantities of ³He are very limited and expensive.

There are two commonly used measurements derived from hyperpolarized ³He MRI, ³He apparent diffusion coefficient (ADC), and the airway functional measurement of ³He ventilation. A brief description of how these structural and functional ³He MRI measurements are quantified is provided in Figure E5S and a comparison of MR, CT, and histology COPD measurements is provided in Table 4.

Diffusion-weighted MR methods have been developed that are sensitive to ³He gas self-diffusion (it is biologically inert and cannot be actively transported across membranes) and provide a measure of the ³He signal that is dependent on the random Brownian motion of the ³He atoms (as opposed to transmembrane gas diffusion). The ³He MRI ADC reflects the decreased diffusion of the gas when inhaled and restricted by the airways and airspaces. The average displacement of helium is the same order of magnitude as alveolar diameters (a few hundred micrometers) and, accordingly, the ³He ADC ranges from 0.8 cm² per second for unrestricted free space (akin to an infinitely large container) to 0.66 cm² per second for an elderly patient with COPD (FEV₁ 26% predicted) and 0.16 cm² per second for a young nonsmoker (FEV₁ 130% predicted) (Figure 3). Quantitative ADC maps derived (62, 63) in previous COPD studies have shown that ADC correlates with pulmonary function (64), histological measurements of lung surface area (65), is highly reproducible in COPD (66) and is dependent on age, position, gravity, and smoking history (67, 68). Short-range diffusion experiments are more sensitive to changes in local microstructure (e.g., alveolar destruction) whereas long-range diffusion experiments are most sensitive to connectivity between airways and larger structural changes (69, 70).

Figure 3. ³He apparent diffusion coefficient (ADC) imaging and maps. (A) ³He ventilation images. (B) ³He ADC maps. (C) ³He ADC gradient vectors. (D) ³He ADC gradient vectors thresholded to 30%, (i) healthy volunteer 60 years of age, (ii) GOLD stage-3 patient, 62 years of age, with chronic obstructive pulmonary disease.

Download Figure | Download Powerpoint

[More] [Minimize]

Hyperpolarized ³He MRI also provides quantitative measurements of those areas of the lung that participate in ventilation and those that do not. As shown in Figure 4, in healthy young adults, a single inhalation of hyperpolarized ³He gas results in homogeneous signal; all areas of the lung are participating equally in ventilation. In contrast, characteristic volumetric “focal” defects are observed in COPD, corresponding to areas of the lung that are not ventilated or are poorly ventilated within the timeframe of a typical 8- to 16-second breath-hold scan. Importantly, when ³He MR ventilation images of a patient with stage-3 COPD are directly compared with CT images, (see Figure E6), there is no anatomical or tissue heterogeneity detected in the CT images that would be predictive of the functional changes revealed by ³He MRI. Focal ³He ventilation defects are quantified as the ³He MRI ventilation defect volume (VDV) and percentage of ventilation volume (PVV) (71) shown to be significantly different in healthy volunteers, healthy asymptomatic smokers, and subjects with COPD (72). ³He MRI VDV was also shown to be sensitive to small functional changes over short periods of time in stage-3 COPD (73). As shown in Figure 4, ³He MRI ADC and ventilation measures provide a way to stratify patients based on these phenotypes.

Figure 4. Different ³He MRI phenotypes of chronic obstructive pulmonary disease (COPD) of two patients with similar lung function. (A) ³He MR images. (B) ³He apparent diffusion coefficient (ADC) maps. (C) ³He ADC histograms; (i) 52-year-old male with stage-3 COPD, FEV₁ 49% predicted, FEV₁/FVC 42%, residual volume (RV) 4.3 L, TLC 8 L, inspiratory volume (IC) 2.9 L, mean ADC = 0.58 cm²/s, mean VDV = 102 cm³; (ii) 72-year-old male with stage-3 COPD, FEV₁ = 49% predicted, FEV₁/FVC 54%, RV 4.1 L, TLC 7.7 L, IC 3.0 L, mean ADC = 0.30 cm²/s, mean VDV = 360 cm³.

Download Figure | Download Powerpoint

[More] [Minimize]

OCT imaging has been applied to the study of bronchial and lung tissues and has been shown to be promising (77–79). The first proof-of-principle study in assessing airway remodeling of COPD was reported by our team (80). In this study, we used a prototype device developed in collaboration between Light Lab Imaging (Boston) and PENTAX (Tokyo, Japan). In this study, we found a strong correlation between CT and OCT measurements of lumen and wall area (80). Importantly, although the correlation between FEV₁% predicted and CT and OCT measured wall area (as a percentage of the total area) of fifth-generation airways was similarly strong, we found that the slope of the relationship was much steeper using OCT than using CT indicating greater sensitivity of OCT in detecting changes in wall measurements that relate to FEV₁ (80). An example of the difference in airway structure between subjects with moderately severe COPD and a smoker of similar age but with normal lung function is shown in Figure 5. In addition to increase in airway wall thickness, differences in the subepithelial matrix and number of alveoli attachment to the airway wall can be detected. Changes in airway lumen diameter between maximal inspiration and tidal breathing may reflect the degree of airway remodeling and elastic recoil property of the lung parenchyma (see Video E1).

Figure 5. Freeze-framed images from optical coherence tomography demonstrating airway wall remodeling of small airways in chronic obstructive pulmonary disease (COPD). The image to the left is from a small airway of a heavy smoker with FEV₁ of 102% of predicted, while the image to the right is from a heavy smoker with FEV₁ of 64% of predicted and a FEV₁/FVC of 68%. The COPD airway has a thicker wall compared to a normal airway of the same size. The innermost ring represents the optical coherence tomography probe and the middle ring represents the lumen of the airway and the outer ring represents airway wall. The small holes in the outer ring represent alveolar attachments (see Video E1 for details).

Download Figure | Download Powerpoint

[More] [Minimize]

However, before OCT technology can be advocated widely for COPD research, standards for proper measurements of airway dimensions need to be developed. OCT is a dynamic in vivo imaging modality and, as such, it is influenced by swings in the intrathoracic pressure during the respiratory cycle (see online videoclip). Deep inspiration increases the size of airway lumen (and decrease the relative airway wall), whereas full expiration decreases the luminal size. Thus, it is imperative that the measurements be “standardized” or “normalized” to a particular phase of the respiratory cycle. Additionally, because there is a large variation in the bronchial anatomy among subjects, it may be a challenge to accurately and consistently identify airways based on the generational number (e.g., fifth-generation airways). See the online supplement for more information on airway standardization.