Using Computed Tomographic Scanning to Advance Understanding of Chronic Obstructive Pulmonary Disease
http://www.100md.com
《美国胸部学报》
Harvard Medical School, and Pulmonary and Critical Care Medicine, Brigham and Women's Hospital, Boston, Massachusetts
ABSTRACT
Chronic obstructive pulmonary disease is a syndrome that encompasses a variety of pathologies and underlying mechanisms. Progress in understanding mechanisms, testing therapies, and identifying contributing genetic factors will be facilitated by the availability of techniques to characterize patients with respect to the nature and extent of both parenchymal and airway diseases. This review discusses the applicability of computed tomographic scan analysis to this problem. The current state of the field is briefly reviewed and future directions for the field are proposed.
Key Words: airways;chronic obstructive pulmonary disease;computed tomographic scanning;density mask;emphysema
THE PROBLEM
Chronic obstructive pulmonary disease (COPD) has been defined by the Global Initiative for Chronic Obstructive Lung Disease (GOLD) as "a disease state characterized by airflow limitation that is not fully reversible. The airflow limitation is usually both progressive and associated with an abnormal inflammatory response of the lungs to noxious particles or gases" (1). The creation of GOLD was in response to the fact that COPD is widely prevalent but relatively under-recognized as a source of significant morbidity and mortality for adults around the world. It has long been recognized that COPD encompasses a broad spectrum of clinical presentations and pathologic abnormalities. The European Respiratory Society definition of COPD (1995) notes the heterogeneity of the pathobiology underlying COPD: "Chronic obstructive pulmonary disease (COPD) is a disorder characterized by reduced maximum expiratory flow and slow forced emptying of the lungs; features which do not change markedly over several months. Most of the airflow limitation is slowly progressive and irreversible. The airflow limitation is due to varying combinations of airway disease and emphysema; the relative contribution of the two processes is difficult to define in vivo" (2).
COPD is therefore a syndrome that encompasses a heterogeneous group of conditions that share the common characteristic of chronic airflow obstruction with varying contributions from emphysema, small airway disease, and chronic bronchitis. Cigarette smoking is the major known risk factor for the development of COPD; however, the development of chronic airflow obstruction is markedly variable among smokers (3). A study by Burrows and colleagues found that only 15% of the variability in FEV1 was explained by pack-years of smoking (3). A study of lung pathology in smokers demonstrated that microscopic emphysema was present in only 26% of smokers (4). The low percentage of variance in pulmonary function explained by smoking and the variable development of microscopic emphysema in smokers suggest that differences in genetic susceptibility to the effects of smoking may be present. Wedzicha notes, "we now know that COPD is a largely heterogeneous condition, consisting of a number of pathological processes whose effects are modified by varied host susceptibility" (5). There is wide variability in the rate of decline of lung function among smokers (6) and poor correlation between disabling symptoms and airflow obstruction. Among patients presenting in a primary care setting who are diagnosed with COPD, O'Brien and colleagues reported a broad range of physiologic abnormalities and a variety of anatomic changes on computed tomographic (CT) scanning, concluding that "COPD in primary care is a heterogeneous condition" (7). There is similar heterogeneity with respect to response to therapy. For example, a meta-analysis performed by Callahan and colleagues concluded that only 10% of patients had a 20% or greater improvement in FEV1 with chronic oral corticosteroid therapy (8). Similarly, Paggiaro and associates noted that improvement of FEV1 by more than 10% occurred in 11% more patients treated with inhaled corticosteroids than placebo (9). This heterogeneity in susceptibility, clinical course, and response to therapy suggests that improved understanding of the mechanisms of airflow obstruction and the ability to more precisely categorize patients will facilitate progress in the field.
Airflow measured during forced exhalation at iso-volume is universally reduced in patients with COPD compared with healthy individuals, indicating the presence of dynamic flow limitation. Recoil pressures at iso-volume are also reduced, demonstrating loss of lung elasticity, a decrease in airway distending pressures, and presumably loss of airway tethering. However, measures of static airway properties are also abnormal in many patients, indicating that only a portion of airflow limitation in advanced COPD is attributable to dynamic airway compression and loss of elastic recoil. Conductance values at zero flow obtained from iso-volume pressure–flow relationships have confirmed abnormal airway physiology in patients with COPD relative to healthy individuals, even among those patients with radiologic evidence of significant tissue destruction and emphysema. These findings are supported by direct in vitro measurements. Macklem and colleagues (10) and Hogg and colleagues (11) measured airway resistance using retrograde catheters in isolated human COPD lungs, and showed marked elevations in airway resistance at physiologic inflation pressures, with the majority of resistance ( 75%) occurring in the peripheral airways (< 2–3 mm). However, only a very small number of such measurements have been made, and the physiology recorded reflects that of only the sickest patients, since lungs were harvested postmortem from individuals dying of their disease.
Airway pathology of tissue specimens obtained from patients with COPD undergoing lung resection or volume reduction surgery confirms varying extents of airway inflammation, destruction, fibrosis, and thickening. However, correlations with physiology and detailed studies of pathophysiology have not been reported (12–14). In certain studies, a close relationship has been demonstrated between the extent of airway pathology and spirometric readouts, whereas in others, little or no relationship appears to exist. Indeed, as pointed out by Macklem and colleagues, the mere presence of airway pathology does not, a priori, imply a physiologic consequence, since a subset of patients with emphysema-predominant COPD appear to have relatively normal airway conductance values despite demonstrating pathologic changes in the small airways (10). This confusing, sometimes contradictory, array of data indicates that, although airway disease is likely an important contributor to the pathophysiology of COPD, precisely how much it contributes, and in whom it contributes, is less clear. The pathobiology of airway and parenchymal disease in COPD, and the clinical and physiologic consequences of each, are not well characterized.
Differences in airway and parenchymal contributions to airflow obstruction may have important implications with respect to genetic or immunologic determinants of disease progression, environmental risk factors, disease severity, rates of progression, and response to specific treatments. This is particularly the case if airway disease and parenchymal destruction have different genetic determinants and/or differing underlying mechanisms. This phenotypic heterogeneity has been cited as a reason for the lack of replication of COPD susceptibility genes (15). Unfortunately, it is difficult to accurately identify the relative contributions of tissue destruction and airway pathology to airflow obstruction using conventional pulmonary function testing. Whether flow limitation is caused by permanent airway narrowing due to scarring and thickening, a sustained increase in airway smooth muscle tone, loss of parallel conducting pathways due to tissue destruction, loss of airway distending pressures with relatively normal airways, or a combination of these factors, the major clinical and physiologic features of COPD can be similar, if not indistinguishable. Our studies demonstrate that, among patients chosen for lung volume reduction surgery by criteria designed to exclude patients with significant airway disease, in many patients the airway disease is responsible for flow limitation (16). Specifically, abnormalities in FEV1, FVC, RV, RV/TLC ratios, exercise capacity, and health-related quality-of-life scores are expected to be similar. Alternative approaches are required to characterize the extent to which each of these factors contributes to flow limitation in any given patient.
This heterogeneity of COPD has slowed progress in the understanding of the pathogenesis of the disease, the development of therapies, and the identification of genetic factors important in the development of COPD. The airflow limitation characteristic of COPD may result from the effects of parenchymal destruction (emphysema) or primary airway pathology, including inflammation, mucus hypersecretion, smooth muscle hypertrophy and constriction, and subepithelial fibrosis. As illustrated in Figure 1, the relative contributions of parenchymal and airway pathology to airflow obstruction varies markedly from patient to patient.
THE NEED FOR BETTER TOOLS TO ASSESS HUMANS WITH COPD
The factors described above indicate that there are a variety of reasons to attempt to develop tools to more specifically characterize the type(s) of COPD present in an individual patient. The ability to identify patients with homogenous underlying pathology will facilitate the investigation of the relationship of in vitro and animal model findings to the pathogenesis of human disease, trials to establish whether therapies directed at specific pathways are effective in humans, and identification of genetic factors contributing to the susceptibility to develop different types of COPD.
Traditionally, the marker used to assess disease severity has been a measure of expiratory flow: the FEV1. It has the advantage of being relatively simple to measure, highly repeatable, noninvasive, and relatively inexpensive. In addition to being used as a classification tool, the longitudinal change in FEV1 has been used to assess disease progression and/or response to therapeutic interventions. Because COPD is a disease that progresses over many years, and the rate of change in FEV1 is modest, studies to date have required relatively large numbers of patients followed for relatively long periods of time.
Among the candidates for a better tool is quantitative analysis of CT scans of the chest. There are two closely related trends in technology that combine to offer the opportunity to develop an approach to use CT scans to accurately characterize patients. The first is the evolution of CT scanning technology with the implementation of multidetector CT scanning. The second is the well-described trend in computer technology with rapid advancements in computational power and falling prices for storage of large volumes of data (17, 18).
DATA ACQUISITION
The widespread availability of multidetector CT scanners and the use of helical ("spiral") CT scanning protocols provide the capability of obtaining a large amount of data on a particular patient. With a single breath-hold of approximately 15 s, the current generation of scanners allows the acquisition of data on the entire thorax in approximate 1-mm slices. The resultant images have much higher resolution, but also more noise, then previous generations of CT scanners. CT scanning uses the pattern and amount of attenuation of a radiation beam passing through the patient to generate a value of signal intensity for a particular unit of volume in the patient. This signal intensity is processed by a reconstruction algorithm to create a two-dimensional image (slice) in which the pixel density (or voxel, considering the third dimension of slice thickness) is directly related to tissue density. The result of this process is a set of two-dimensional gray-scale images used for clinical and research purposes. It is important to recognize that the presentation of CT data is based on the requirements and preferences of the individuals reviewing the scans. The visual appearance of the scan is a function of the algorithm used to generate the visual image. Multidetector CT scanners provide a larger body of data from which to construct thinner, and consequently more numerous, slices. The opportunity and the challenge of the availability of these data are the transmission, storage, display, and analysis of a large body of tissue density data.
The issue of reconstruction/presentation algorithms, also known as "kernels" or convolution filters, is an important methodologic consideration in the evaluation of CT data. The kernel is a mathematical algorithm that takes the intensity of a particular pixel and assigns it a visual value on a gray scale. The visual value is calculated after considering the intensity values of surrounding pixels. Certain approaches emphasize contrast, accentuate noise, and produce sharp edges (so-called high-resolution scans), whereas others average with the surrounding pixels to produce less well defined edges and boundaries but result in an image with qualities that are preferred by most clinicians reviewing the scans. If one is interested in measurements that involve the boundaries of structures, such as airway wall thickness, the choice of reconstruction kernel will affect the measurement. Further complicating the issue is that the algorithms used in CT scanning are largely proprietary and vary from manufacturer to manufacturer. Currently, there is no open standard or agreed-upon "gold" standard for CT reconstruction algorithms.
DATA ANALYSIS
Traditionally, analysis of chest CT scans has been performed by radiologists and clinicians, with the emphasis on describing abnormalities. These include areas of increased tissue density, such as lung masses, interstitial fibrosis, or mediastinal adenopathy, or areas of decreased density, such as bullae or pneumothoraces. With the exception of defining the size of certain abnormal findings, particularly nodules or masses, the emphasis has been on subjective, rather than quantitative, assessment. Although the amount of data available in a multidetector chest CT scan is large and can be used to generate several hundred contiguous slices through the thorax, reviewing such a large number of images is too cumbersome for routine clinical use. Consequently, most CT scans obtained for clinical purposes are presented in a format with fewer slices, each of increased thickness, or a series of noncontiguous thin slices at predetermined intervals. In almost all settings, the raw data obtained from the CT scanning process are discarded shortly after the image reconstruction is completed. This means that data analysis is performed on the reconstructed images and that the reconstruction protocol becomes an important variable to consider in data analysis.
If the CT scan is to be used to generate information about parenchymal and airway characteristics, then a protocol that provides the thinnest contiguous slices that encompass the region of interest is preferred. Such a large body of data is best analyzed by specially designed software capable of processing large amounts of voxel intensity data. Several such tools are available and under active development (14, 19, 20).
WHAT ARE THE DATA OF INTEREST AVAILABLE FROM CHEST CT SCANS?
The two obvious sets of data of interest to investigators interested in lung disease are characterization of parenchymal tissue density (fibrosis/emphysema) and characterization of airways. These will be considered separately in the following sections.
Parenchymal Tissue Density
There is a body of work, with prominent contributions from Hogg and colleagues, relating anatomically determined tissue density with Hounsfield units (HU) measured on CT scans of the chest (12–14, 21–23). The most commonly used approach to characterizing "normal" and "abnormal" lung is a dichotomous one that involves applying a density threshold for differentiating the two. After first limiting the characterization to the lung parenchyma, which involves removing the chest wall, trachea, and mediastinal and hilar structures from the field of interest (known as "segmentation"), an HU threshold is applied and the lung is divided into "normal" voxels and "abnormal" voxels. This is most commonly performed in the quantitation of emphysema, in which voxels below the threshold are classified as emphysema and those above as normal. It is immediately obvious that the amount of emphysematous lung reported from a particular scan is dependent on the threshold used to dichotomize the lung. To date, there is no universal agreement on a single HU threshold that differentiates normal from emphysematous lung. Most reports use values between –910 and –950 HU. An additional factor that will affect the calculated extent of emphysema is the reconstruction kernel used to perform the scan (24).
These data can then be reported in a number of ways: the percentage of the lung that is occupied by emphysema and the visual display of the data with emphysema voxels presented in one color and normal voxels in another are the most common examples of the density mask approach (see Figure 2). Additional refinements of this approach include dividing the lung into regions and comparing the characteristics of each region with the others to reach conclusions about the uniformity, or lack thereof, of the anatomic distribution of disease. An example of the utility of this type of characterization is the assessment of patients being considered for lung volume reduction surgery. The National Emphysema Treatment Trial demonstrated that patients with "upper lobe predominant" emphysema, in which the proportion of lung classified as emphysema is higher in the upper one-third than the lower portions of the lung, receive greater benefits from the surgery than patients with more uniformly distributed disease.
Additional analytic approaches applied to these data include the calculation of "hole size" by grouping adjacent emphysematous voxels into a "hole." The distribution of the number and size of holes can then be reported. The slope of a log-log plot of hole number and size is termed the "alpha" and has been used to summarize these data (25).
Airways
The methodology for the characterization of the airways is less well developed than that of parenchymal tissue density. The challenge is severalfold: reliable identification of airways, particularly small airways; accurate delineation of the internal and external edges of the airway wall; and the development of a useful way to present the data characterizing the airways. Also unaddressed as of yet is whether characteristics of the voxels within the airway walls provide useful information about the nature of airway pathology.
Identification of airways.
A major challenge is the identification of airways for analysis. Although the data in CT scanning is acquired in three dimensions, it is typically reconstructed and presented as a series of two-dimensional images. The plane of these images is most often one that is orthogonal to the central axis of the body. Use of data presented in this manner means that most of the airways are "cut" at a nonperpendicular angle to the plane of the scan. One exception to this is the apical segment of the right upper lobe (RUL), which is therefore commonly used to assess airway wall thickness in reports in the literature (26). Most tangentially cut airways are not usable for analysis, because the nonorthogonal plane of projection distorts the proportions of the airway wall to the lumen.
The use of two-dimensional planar data for airway analysis has several limitations. As noted above, the number of airways available for analysis is small. The analysis requires a significant amount of human input to identify the airway, select the region of interest, and edit the computer-generated airway wall dimensions to eliminate artifacts caused by adjacent blood vessels (27). Although some data are available to suggest that the findings in large airways, which are the most easily analyzed by these methods, are representative of pathology found in small airways, few data are available about the heterogeneity of airway abnormalities in an individual patient. The requirement for extensive human operator input has also limited its applicability, because the use of these techniques is not feasible in studies involving large numbers of chest CT scans.
An important issue concerning the definition of airway characteristics are the methods used to define the boundaries of the airway wall and the reconstruction kernel used to generate the data being analyzed. The most commonly used method for the quantitation of airway wall dimensions is the "full width half maximum" method, in which the inner and outer airway wall boundaries are defined as the point corresponding to the half maximal intensity of the airway wall voxels (reviewed in Reference 27). As outlined above, the choice of reconstruction kernel will affect the shape of the intensity profile of the airway wall, which in turn will affect the calculated airway dimension derived from the CT scan. Nakano and colleagues have demonstrated that this approach results in large fractional errors in small airways (14). This has led to efforts to develop alternative computational methods to define airway wall boundaries. Ideally, these techniques will result in more accurate measurements of the small airways, which are the site of airflow limitation in COPD, and produce consistent results that are independent of the reconstruction kernel used to generate the scan images.
Adding a new dimension.
With apologies to Edward Tufte (28), the challenge for researchers in this field is "Escaping Flatland," that is to say the development of approaches to analyze the data in three dimensions, rather than two. The ability to create a volumetric representation of the airway tree requires two components: a sufficient density of data (number of slices) that allows reliable identification of a continuous airway lumen and robust tools that accurately and reliably identify the airway lumen and the inner and outer boundaries of the airway walls. The farther apart and thicker CT slices are, the less certainty an observer, or computer program, can have that an airway lumen seen on one slice corresponds to the same airway as the lumen seen on another slice. The development and introduction of 32- and 64-detector CT scanners provides the technology to deal with this issue.
Still in evolution is the development of automated tools to create a volumetric representation of the airway tree (an example is shown in Figure 3). Most approaches involve identifying the lumen of a large airway and then "growing" the airway lumen by looking for adjacent voxels of extremely low (air) density. Once a centerline of the lumen is established, a series of rays can be drawn at any point in a plane orthogonal to the axis of the lumen and an attempt made to define the inner and outer boundary of the airway wall along these rays. These data can be captured quantitatively and also used to generate a visual image of the bronchial tree. The exact dimensions of this image are a function of the reconstruction kernel, as outlined above. The analysis of such a three-dimensional image is still in evolution. There is no consensus as to how to report on the quantitative airway data that are derivable from such an image.
A common function of the software programs under development is that they are observer/operator independent. Any person trained to use the program will arrive at the same results as others in the analysis of the same set of data. This is in marked distinction to subjective, semiquantitative assessment.
FUTURE DIRECTIONS IN THE APPLICATION OF CT SCANS IN COPD
CT scanning holds great promise as a tool for the use in the characterization of patients with COPD. It is widely available, relatively inexpensive, and has the potential to be used in both cross-sectional and longitudinal studies. To fully capitalize on the potential of CT scans in this context, a number of challenges need to be addressed.
These include the following:
The development of widely available software programs that will automatically segment the lung, perform parenchymal characterization, accurately identify airways, and characterize airway dimensions and airway wall characteristics. The ideal application will be one that performs an accurate characterization independent of the reconstruction kernel used to generate the images. Table 1 summarizes the characteristics of an ideal program.
The development of standard nomenclature for describing the airways being characterized. For example, airway diameter can refer to either lumenal diameter or the diameter of the outer boundary of the airway wall; in addition, there is not agreement as to whether to characterize airways by size or by generation.
A systematic correlation of CT characteristics, physiology, and microscopic anatomy derived from patients encompassing the full spectrum of COPD, as assessed by severity of airflow obstruction, degree of airways disease, and amount and distribution of emphysema.
Longitudinal data describing the changes of CT characteristics of COPD over time. This will allow development of agreement as to what constitutes a significant change in parenchymal and/or airway characteristics. This will facilitate the use of CT scanning as an endpoint in observational and therapeutic human studies in COPD.
More transparency in reconstruction kernels and an improved ability to capture and archive the raw data. Included in this concept is a continually updated database that allows comparison of the properties of different kernels.
The development of an approach to present the vast amount of data obtained from a CT scan in a concise manner that is meaningful and comprehensible to an interested medical audience.
Although at first appearance, the list appears daunting, the pace of progress in the field is currently very rapid, in part because of the ability to draw on application development that has been performed in other organ systems, such as the central nervous system. It is likely that many of these goals will be reached over the next several years.
FOOTNOTES
Conflict of Interest Statement: J.R. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript.
REFERENCES
Aerdts SJ, van Dalen R, Clasener HA, Festen J, van Lier HJ, Vollaard EJ. Antibiotic prophylaxis of respiratory tract infection in mechanically ventilated patients. Chest 1991;100:783–791.
Siafakas NM, Vermeire P, Pride NB, Paoletti P, Gibson J, Howard P, Yernault JC, Decramer M, Higgenbottam T, Postma DS, et al. Optimal assessment and management of chronic obstructive pulmonary disease (COPD). The European Respiratory Society Task Force. Eur Respir J 1995;8:1398–1420.
Burrows B, Knudson RJ, Cline MG, Lebowitz MD. Quantitative relationships between cigarette smoking and ventilatory function. Am Rev Respir Dis 1977;115:195–205.
Gillooly M, Lamb D. Microscopic emphysema in relation to age and smoking habit. Thorax 1993;48:491–495.
Wedzicha JA. The heterogeneity of chronic obstructive pulmonary disease. Thorax 2000;55:631–632.
Anthonisen NR, Connett JE, Kiley JP, et al. Effects of smoking intervention and the use of an inhaled anticholinergic bronchodilator on the rate of decline of FEV1: the Lung Health Study. JAMA 1994;272:11497–11505.
O'Brien C, Guest PJ, Hill SL, Stockley RA. Physiological and radiological characterisation of patients diagnosed with chronic obstructive pulmonary disease in primary care. Thorax 2000;55:635–642.
Callahan CM, Dittus RS, Katz BP. Oral corticosteroid therapy for patients with stable chronic obstructive pulmonary disease: a meta-analysis. Ann Intern Med 1991;114:216–223.
Paggiaro PL, Dahle R, Bakran I, Frith L, Hollingworth K, Efthimiou J. Multicentre randomised placebo-controlled trial of inhaled fluticasone propionate in patients with chronic obstructive pulmonary disease. International COPD Study Group. Lancet 1998;351:773–780.
Macklem PT, Fraser RG, Brown WG. The detection of the flow-limiting bronchi in bronchitis and emphysema by airway pressure measurements. Med Thorac 1965;22:220–230.
Hogg JC, Macklem PT, Thurlbeck WM. Site and nature of airway obstruction in chronic obstructive lung disease. N Engl J Med 1968;278:1355–1360.
Hogg JC, Chu F, Utokaparch S, et al. The nature of small-airway obstruction in chronic obstructive pulmonary disease. N Engl J Med 2004;350:2645–2653.
Nakano Y, Muller NL, King GG, et al. Quantitative assessment of airway remodeling using high-resolution CT. Chest 2002;122:271S–275S.
Nakano Y, Wong JC, de Jong PA, et al. The prediction of small airway dimensions using computed tomography. Am J Respir Crit Care Med 2005;171:142–146.
Hersh CP, Demeo DL, Lange C, et al. Attempted replication of reported chronic obstructive pulmonary disease candidate gene associations. Am J Respir Cell Mol Biol 2005;33:71–78.
Ingenito EP, Evans RB, Loring SH, et al. Relation between preoperative inspiratory lung resistance and the outcome of lung-volume-reduction surgery for emphysema. N Engl J Med 1998;338:1181–1185.
Moore G. Cramming more components onto integrated circuits. Electronics 1965;38:114–117.
Moore G. Lithography and the future of Moore's law. Proc SPIE 1995;2440:2–17.
Hoffman E. Available from: http://dpi.radiology.uiowa.edu/vida/vidahome.html (accessed April 2, 2006).
Kikinis RJF, Grimson E, Wells W. Available from: www.slicer.org (accessed April 2, 2006).
Muller NL, Staples CA, Miller RR, Abboud RT. "Density mask": an objective method to quantitate emphysema using computed tomography. Chest 1988;94:782–787.
Coxson HO, Rogers RM, Whittall KP, et al. A quantification of the lung surface area in emphysema using computed tomography. Am J Respir Crit Care Med 1999;159:851–856.
Gevenois PA, de Maertelaer V, De Vuyst P, Zanen J, Yernault JC. Comparison of computed density and macroscopic morphometry in pulmonary emphysema. Am J Respir Crit Care Med 1995;152:653–657.
Kemerink GJ, Kruize HH, Lamers RJ, van Engelshoven JM. CT lung densitometry: dependence of CT number histograms on sample volume and consequences for scan protocol comparability. J Comput Assist Tomogr 1997;21:948–954.
Coxson HO, Whittall KP, Nakano Y, et al. Selection of patients for lung volume reduction surgery using a power law analysis of the computed tomographic scan. Thorax 2003;58:510–514.
Nakano Y, Muro S, Sakai H, et al. Computed tomographic measurements of airway dimensions and emphysema in smokers: correlation with lung function. Am J Respir Crit Care Med 2000;162:1102–1108.
de Jong PA, Muller NL, Pare PD, Coxson HO. Computed tomographic imaging of the airways: relationship to structure and function. Eur Respir J 2005;26:140–152.
Tufte E. Envisioning information. Cheshire, CT: Graphics Press; 1990.(John Reilly)
ABSTRACT
Chronic obstructive pulmonary disease is a syndrome that encompasses a variety of pathologies and underlying mechanisms. Progress in understanding mechanisms, testing therapies, and identifying contributing genetic factors will be facilitated by the availability of techniques to characterize patients with respect to the nature and extent of both parenchymal and airway diseases. This review discusses the applicability of computed tomographic scan analysis to this problem. The current state of the field is briefly reviewed and future directions for the field are proposed.
Key Words: airways;chronic obstructive pulmonary disease;computed tomographic scanning;density mask;emphysema
THE PROBLEM
Chronic obstructive pulmonary disease (COPD) has been defined by the Global Initiative for Chronic Obstructive Lung Disease (GOLD) as "a disease state characterized by airflow limitation that is not fully reversible. The airflow limitation is usually both progressive and associated with an abnormal inflammatory response of the lungs to noxious particles or gases" (1). The creation of GOLD was in response to the fact that COPD is widely prevalent but relatively under-recognized as a source of significant morbidity and mortality for adults around the world. It has long been recognized that COPD encompasses a broad spectrum of clinical presentations and pathologic abnormalities. The European Respiratory Society definition of COPD (1995) notes the heterogeneity of the pathobiology underlying COPD: "Chronic obstructive pulmonary disease (COPD) is a disorder characterized by reduced maximum expiratory flow and slow forced emptying of the lungs; features which do not change markedly over several months. Most of the airflow limitation is slowly progressive and irreversible. The airflow limitation is due to varying combinations of airway disease and emphysema; the relative contribution of the two processes is difficult to define in vivo" (2).
COPD is therefore a syndrome that encompasses a heterogeneous group of conditions that share the common characteristic of chronic airflow obstruction with varying contributions from emphysema, small airway disease, and chronic bronchitis. Cigarette smoking is the major known risk factor for the development of COPD; however, the development of chronic airflow obstruction is markedly variable among smokers (3). A study by Burrows and colleagues found that only 15% of the variability in FEV1 was explained by pack-years of smoking (3). A study of lung pathology in smokers demonstrated that microscopic emphysema was present in only 26% of smokers (4). The low percentage of variance in pulmonary function explained by smoking and the variable development of microscopic emphysema in smokers suggest that differences in genetic susceptibility to the effects of smoking may be present. Wedzicha notes, "we now know that COPD is a largely heterogeneous condition, consisting of a number of pathological processes whose effects are modified by varied host susceptibility" (5). There is wide variability in the rate of decline of lung function among smokers (6) and poor correlation between disabling symptoms and airflow obstruction. Among patients presenting in a primary care setting who are diagnosed with COPD, O'Brien and colleagues reported a broad range of physiologic abnormalities and a variety of anatomic changes on computed tomographic (CT) scanning, concluding that "COPD in primary care is a heterogeneous condition" (7). There is similar heterogeneity with respect to response to therapy. For example, a meta-analysis performed by Callahan and colleagues concluded that only 10% of patients had a 20% or greater improvement in FEV1 with chronic oral corticosteroid therapy (8). Similarly, Paggiaro and associates noted that improvement of FEV1 by more than 10% occurred in 11% more patients treated with inhaled corticosteroids than placebo (9). This heterogeneity in susceptibility, clinical course, and response to therapy suggests that improved understanding of the mechanisms of airflow obstruction and the ability to more precisely categorize patients will facilitate progress in the field.
Airflow measured during forced exhalation at iso-volume is universally reduced in patients with COPD compared with healthy individuals, indicating the presence of dynamic flow limitation. Recoil pressures at iso-volume are also reduced, demonstrating loss of lung elasticity, a decrease in airway distending pressures, and presumably loss of airway tethering. However, measures of static airway properties are also abnormal in many patients, indicating that only a portion of airflow limitation in advanced COPD is attributable to dynamic airway compression and loss of elastic recoil. Conductance values at zero flow obtained from iso-volume pressure–flow relationships have confirmed abnormal airway physiology in patients with COPD relative to healthy individuals, even among those patients with radiologic evidence of significant tissue destruction and emphysema. These findings are supported by direct in vitro measurements. Macklem and colleagues (10) and Hogg and colleagues (11) measured airway resistance using retrograde catheters in isolated human COPD lungs, and showed marked elevations in airway resistance at physiologic inflation pressures, with the majority of resistance ( 75%) occurring in the peripheral airways (< 2–3 mm). However, only a very small number of such measurements have been made, and the physiology recorded reflects that of only the sickest patients, since lungs were harvested postmortem from individuals dying of their disease.
Airway pathology of tissue specimens obtained from patients with COPD undergoing lung resection or volume reduction surgery confirms varying extents of airway inflammation, destruction, fibrosis, and thickening. However, correlations with physiology and detailed studies of pathophysiology have not been reported (12–14). In certain studies, a close relationship has been demonstrated between the extent of airway pathology and spirometric readouts, whereas in others, little or no relationship appears to exist. Indeed, as pointed out by Macklem and colleagues, the mere presence of airway pathology does not, a priori, imply a physiologic consequence, since a subset of patients with emphysema-predominant COPD appear to have relatively normal airway conductance values despite demonstrating pathologic changes in the small airways (10). This confusing, sometimes contradictory, array of data indicates that, although airway disease is likely an important contributor to the pathophysiology of COPD, precisely how much it contributes, and in whom it contributes, is less clear. The pathobiology of airway and parenchymal disease in COPD, and the clinical and physiologic consequences of each, are not well characterized.
Differences in airway and parenchymal contributions to airflow obstruction may have important implications with respect to genetic or immunologic determinants of disease progression, environmental risk factors, disease severity, rates of progression, and response to specific treatments. This is particularly the case if airway disease and parenchymal destruction have different genetic determinants and/or differing underlying mechanisms. This phenotypic heterogeneity has been cited as a reason for the lack of replication of COPD susceptibility genes (15). Unfortunately, it is difficult to accurately identify the relative contributions of tissue destruction and airway pathology to airflow obstruction using conventional pulmonary function testing. Whether flow limitation is caused by permanent airway narrowing due to scarring and thickening, a sustained increase in airway smooth muscle tone, loss of parallel conducting pathways due to tissue destruction, loss of airway distending pressures with relatively normal airways, or a combination of these factors, the major clinical and physiologic features of COPD can be similar, if not indistinguishable. Our studies demonstrate that, among patients chosen for lung volume reduction surgery by criteria designed to exclude patients with significant airway disease, in many patients the airway disease is responsible for flow limitation (16). Specifically, abnormalities in FEV1, FVC, RV, RV/TLC ratios, exercise capacity, and health-related quality-of-life scores are expected to be similar. Alternative approaches are required to characterize the extent to which each of these factors contributes to flow limitation in any given patient.
This heterogeneity of COPD has slowed progress in the understanding of the pathogenesis of the disease, the development of therapies, and the identification of genetic factors important in the development of COPD. The airflow limitation characteristic of COPD may result from the effects of parenchymal destruction (emphysema) or primary airway pathology, including inflammation, mucus hypersecretion, smooth muscle hypertrophy and constriction, and subepithelial fibrosis. As illustrated in Figure 1, the relative contributions of parenchymal and airway pathology to airflow obstruction varies markedly from patient to patient.
THE NEED FOR BETTER TOOLS TO ASSESS HUMANS WITH COPD
The factors described above indicate that there are a variety of reasons to attempt to develop tools to more specifically characterize the type(s) of COPD present in an individual patient. The ability to identify patients with homogenous underlying pathology will facilitate the investigation of the relationship of in vitro and animal model findings to the pathogenesis of human disease, trials to establish whether therapies directed at specific pathways are effective in humans, and identification of genetic factors contributing to the susceptibility to develop different types of COPD.
Traditionally, the marker used to assess disease severity has been a measure of expiratory flow: the FEV1. It has the advantage of being relatively simple to measure, highly repeatable, noninvasive, and relatively inexpensive. In addition to being used as a classification tool, the longitudinal change in FEV1 has been used to assess disease progression and/or response to therapeutic interventions. Because COPD is a disease that progresses over many years, and the rate of change in FEV1 is modest, studies to date have required relatively large numbers of patients followed for relatively long periods of time.
Among the candidates for a better tool is quantitative analysis of CT scans of the chest. There are two closely related trends in technology that combine to offer the opportunity to develop an approach to use CT scans to accurately characterize patients. The first is the evolution of CT scanning technology with the implementation of multidetector CT scanning. The second is the well-described trend in computer technology with rapid advancements in computational power and falling prices for storage of large volumes of data (17, 18).
DATA ACQUISITION
The widespread availability of multidetector CT scanners and the use of helical ("spiral") CT scanning protocols provide the capability of obtaining a large amount of data on a particular patient. With a single breath-hold of approximately 15 s, the current generation of scanners allows the acquisition of data on the entire thorax in approximate 1-mm slices. The resultant images have much higher resolution, but also more noise, then previous generations of CT scanners. CT scanning uses the pattern and amount of attenuation of a radiation beam passing through the patient to generate a value of signal intensity for a particular unit of volume in the patient. This signal intensity is processed by a reconstruction algorithm to create a two-dimensional image (slice) in which the pixel density (or voxel, considering the third dimension of slice thickness) is directly related to tissue density. The result of this process is a set of two-dimensional gray-scale images used for clinical and research purposes. It is important to recognize that the presentation of CT data is based on the requirements and preferences of the individuals reviewing the scans. The visual appearance of the scan is a function of the algorithm used to generate the visual image. Multidetector CT scanners provide a larger body of data from which to construct thinner, and consequently more numerous, slices. The opportunity and the challenge of the availability of these data are the transmission, storage, display, and analysis of a large body of tissue density data.
The issue of reconstruction/presentation algorithms, also known as "kernels" or convolution filters, is an important methodologic consideration in the evaluation of CT data. The kernel is a mathematical algorithm that takes the intensity of a particular pixel and assigns it a visual value on a gray scale. The visual value is calculated after considering the intensity values of surrounding pixels. Certain approaches emphasize contrast, accentuate noise, and produce sharp edges (so-called high-resolution scans), whereas others average with the surrounding pixels to produce less well defined edges and boundaries but result in an image with qualities that are preferred by most clinicians reviewing the scans. If one is interested in measurements that involve the boundaries of structures, such as airway wall thickness, the choice of reconstruction kernel will affect the measurement. Further complicating the issue is that the algorithms used in CT scanning are largely proprietary and vary from manufacturer to manufacturer. Currently, there is no open standard or agreed-upon "gold" standard for CT reconstruction algorithms.
DATA ANALYSIS
Traditionally, analysis of chest CT scans has been performed by radiologists and clinicians, with the emphasis on describing abnormalities. These include areas of increased tissue density, such as lung masses, interstitial fibrosis, or mediastinal adenopathy, or areas of decreased density, such as bullae or pneumothoraces. With the exception of defining the size of certain abnormal findings, particularly nodules or masses, the emphasis has been on subjective, rather than quantitative, assessment. Although the amount of data available in a multidetector chest CT scan is large and can be used to generate several hundred contiguous slices through the thorax, reviewing such a large number of images is too cumbersome for routine clinical use. Consequently, most CT scans obtained for clinical purposes are presented in a format with fewer slices, each of increased thickness, or a series of noncontiguous thin slices at predetermined intervals. In almost all settings, the raw data obtained from the CT scanning process are discarded shortly after the image reconstruction is completed. This means that data analysis is performed on the reconstructed images and that the reconstruction protocol becomes an important variable to consider in data analysis.
If the CT scan is to be used to generate information about parenchymal and airway characteristics, then a protocol that provides the thinnest contiguous slices that encompass the region of interest is preferred. Such a large body of data is best analyzed by specially designed software capable of processing large amounts of voxel intensity data. Several such tools are available and under active development (14, 19, 20).
WHAT ARE THE DATA OF INTEREST AVAILABLE FROM CHEST CT SCANS?
The two obvious sets of data of interest to investigators interested in lung disease are characterization of parenchymal tissue density (fibrosis/emphysema) and characterization of airways. These will be considered separately in the following sections.
Parenchymal Tissue Density
There is a body of work, with prominent contributions from Hogg and colleagues, relating anatomically determined tissue density with Hounsfield units (HU) measured on CT scans of the chest (12–14, 21–23). The most commonly used approach to characterizing "normal" and "abnormal" lung is a dichotomous one that involves applying a density threshold for differentiating the two. After first limiting the characterization to the lung parenchyma, which involves removing the chest wall, trachea, and mediastinal and hilar structures from the field of interest (known as "segmentation"), an HU threshold is applied and the lung is divided into "normal" voxels and "abnormal" voxels. This is most commonly performed in the quantitation of emphysema, in which voxels below the threshold are classified as emphysema and those above as normal. It is immediately obvious that the amount of emphysematous lung reported from a particular scan is dependent on the threshold used to dichotomize the lung. To date, there is no universal agreement on a single HU threshold that differentiates normal from emphysematous lung. Most reports use values between –910 and –950 HU. An additional factor that will affect the calculated extent of emphysema is the reconstruction kernel used to perform the scan (24).
These data can then be reported in a number of ways: the percentage of the lung that is occupied by emphysema and the visual display of the data with emphysema voxels presented in one color and normal voxels in another are the most common examples of the density mask approach (see Figure 2). Additional refinements of this approach include dividing the lung into regions and comparing the characteristics of each region with the others to reach conclusions about the uniformity, or lack thereof, of the anatomic distribution of disease. An example of the utility of this type of characterization is the assessment of patients being considered for lung volume reduction surgery. The National Emphysema Treatment Trial demonstrated that patients with "upper lobe predominant" emphysema, in which the proportion of lung classified as emphysema is higher in the upper one-third than the lower portions of the lung, receive greater benefits from the surgery than patients with more uniformly distributed disease.
Additional analytic approaches applied to these data include the calculation of "hole size" by grouping adjacent emphysematous voxels into a "hole." The distribution of the number and size of holes can then be reported. The slope of a log-log plot of hole number and size is termed the "alpha" and has been used to summarize these data (25).
Airways
The methodology for the characterization of the airways is less well developed than that of parenchymal tissue density. The challenge is severalfold: reliable identification of airways, particularly small airways; accurate delineation of the internal and external edges of the airway wall; and the development of a useful way to present the data characterizing the airways. Also unaddressed as of yet is whether characteristics of the voxels within the airway walls provide useful information about the nature of airway pathology.
Identification of airways.
A major challenge is the identification of airways for analysis. Although the data in CT scanning is acquired in three dimensions, it is typically reconstructed and presented as a series of two-dimensional images. The plane of these images is most often one that is orthogonal to the central axis of the body. Use of data presented in this manner means that most of the airways are "cut" at a nonperpendicular angle to the plane of the scan. One exception to this is the apical segment of the right upper lobe (RUL), which is therefore commonly used to assess airway wall thickness in reports in the literature (26). Most tangentially cut airways are not usable for analysis, because the nonorthogonal plane of projection distorts the proportions of the airway wall to the lumen.
The use of two-dimensional planar data for airway analysis has several limitations. As noted above, the number of airways available for analysis is small. The analysis requires a significant amount of human input to identify the airway, select the region of interest, and edit the computer-generated airway wall dimensions to eliminate artifacts caused by adjacent blood vessels (27). Although some data are available to suggest that the findings in large airways, which are the most easily analyzed by these methods, are representative of pathology found in small airways, few data are available about the heterogeneity of airway abnormalities in an individual patient. The requirement for extensive human operator input has also limited its applicability, because the use of these techniques is not feasible in studies involving large numbers of chest CT scans.
An important issue concerning the definition of airway characteristics are the methods used to define the boundaries of the airway wall and the reconstruction kernel used to generate the data being analyzed. The most commonly used method for the quantitation of airway wall dimensions is the "full width half maximum" method, in which the inner and outer airway wall boundaries are defined as the point corresponding to the half maximal intensity of the airway wall voxels (reviewed in Reference 27). As outlined above, the choice of reconstruction kernel will affect the shape of the intensity profile of the airway wall, which in turn will affect the calculated airway dimension derived from the CT scan. Nakano and colleagues have demonstrated that this approach results in large fractional errors in small airways (14). This has led to efforts to develop alternative computational methods to define airway wall boundaries. Ideally, these techniques will result in more accurate measurements of the small airways, which are the site of airflow limitation in COPD, and produce consistent results that are independent of the reconstruction kernel used to generate the scan images.
Adding a new dimension.
With apologies to Edward Tufte (28), the challenge for researchers in this field is "Escaping Flatland," that is to say the development of approaches to analyze the data in three dimensions, rather than two. The ability to create a volumetric representation of the airway tree requires two components: a sufficient density of data (number of slices) that allows reliable identification of a continuous airway lumen and robust tools that accurately and reliably identify the airway lumen and the inner and outer boundaries of the airway walls. The farther apart and thicker CT slices are, the less certainty an observer, or computer program, can have that an airway lumen seen on one slice corresponds to the same airway as the lumen seen on another slice. The development and introduction of 32- and 64-detector CT scanners provides the technology to deal with this issue.
Still in evolution is the development of automated tools to create a volumetric representation of the airway tree (an example is shown in Figure 3). Most approaches involve identifying the lumen of a large airway and then "growing" the airway lumen by looking for adjacent voxels of extremely low (air) density. Once a centerline of the lumen is established, a series of rays can be drawn at any point in a plane orthogonal to the axis of the lumen and an attempt made to define the inner and outer boundary of the airway wall along these rays. These data can be captured quantitatively and also used to generate a visual image of the bronchial tree. The exact dimensions of this image are a function of the reconstruction kernel, as outlined above. The analysis of such a three-dimensional image is still in evolution. There is no consensus as to how to report on the quantitative airway data that are derivable from such an image.
A common function of the software programs under development is that they are observer/operator independent. Any person trained to use the program will arrive at the same results as others in the analysis of the same set of data. This is in marked distinction to subjective, semiquantitative assessment.
FUTURE DIRECTIONS IN THE APPLICATION OF CT SCANS IN COPD
CT scanning holds great promise as a tool for the use in the characterization of patients with COPD. It is widely available, relatively inexpensive, and has the potential to be used in both cross-sectional and longitudinal studies. To fully capitalize on the potential of CT scans in this context, a number of challenges need to be addressed.
These include the following:
The development of widely available software programs that will automatically segment the lung, perform parenchymal characterization, accurately identify airways, and characterize airway dimensions and airway wall characteristics. The ideal application will be one that performs an accurate characterization independent of the reconstruction kernel used to generate the images. Table 1 summarizes the characteristics of an ideal program.
The development of standard nomenclature for describing the airways being characterized. For example, airway diameter can refer to either lumenal diameter or the diameter of the outer boundary of the airway wall; in addition, there is not agreement as to whether to characterize airways by size or by generation.
A systematic correlation of CT characteristics, physiology, and microscopic anatomy derived from patients encompassing the full spectrum of COPD, as assessed by severity of airflow obstruction, degree of airways disease, and amount and distribution of emphysema.
Longitudinal data describing the changes of CT characteristics of COPD over time. This will allow development of agreement as to what constitutes a significant change in parenchymal and/or airway characteristics. This will facilitate the use of CT scanning as an endpoint in observational and therapeutic human studies in COPD.
More transparency in reconstruction kernels and an improved ability to capture and archive the raw data. Included in this concept is a continually updated database that allows comparison of the properties of different kernels.
The development of an approach to present the vast amount of data obtained from a CT scan in a concise manner that is meaningful and comprehensible to an interested medical audience.
Although at first appearance, the list appears daunting, the pace of progress in the field is currently very rapid, in part because of the ability to draw on application development that has been performed in other organ systems, such as the central nervous system. It is likely that many of these goals will be reached over the next several years.
FOOTNOTES
Conflict of Interest Statement: J.R. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript.
REFERENCES
Aerdts SJ, van Dalen R, Clasener HA, Festen J, van Lier HJ, Vollaard EJ. Antibiotic prophylaxis of respiratory tract infection in mechanically ventilated patients. Chest 1991;100:783–791.
Siafakas NM, Vermeire P, Pride NB, Paoletti P, Gibson J, Howard P, Yernault JC, Decramer M, Higgenbottam T, Postma DS, et al. Optimal assessment and management of chronic obstructive pulmonary disease (COPD). The European Respiratory Society Task Force. Eur Respir J 1995;8:1398–1420.
Burrows B, Knudson RJ, Cline MG, Lebowitz MD. Quantitative relationships between cigarette smoking and ventilatory function. Am Rev Respir Dis 1977;115:195–205.
Gillooly M, Lamb D. Microscopic emphysema in relation to age and smoking habit. Thorax 1993;48:491–495.
Wedzicha JA. The heterogeneity of chronic obstructive pulmonary disease. Thorax 2000;55:631–632.
Anthonisen NR, Connett JE, Kiley JP, et al. Effects of smoking intervention and the use of an inhaled anticholinergic bronchodilator on the rate of decline of FEV1: the Lung Health Study. JAMA 1994;272:11497–11505.
O'Brien C, Guest PJ, Hill SL, Stockley RA. Physiological and radiological characterisation of patients diagnosed with chronic obstructive pulmonary disease in primary care. Thorax 2000;55:635–642.
Callahan CM, Dittus RS, Katz BP. Oral corticosteroid therapy for patients with stable chronic obstructive pulmonary disease: a meta-analysis. Ann Intern Med 1991;114:216–223.
Paggiaro PL, Dahle R, Bakran I, Frith L, Hollingworth K, Efthimiou J. Multicentre randomised placebo-controlled trial of inhaled fluticasone propionate in patients with chronic obstructive pulmonary disease. International COPD Study Group. Lancet 1998;351:773–780.
Macklem PT, Fraser RG, Brown WG. The detection of the flow-limiting bronchi in bronchitis and emphysema by airway pressure measurements. Med Thorac 1965;22:220–230.
Hogg JC, Macklem PT, Thurlbeck WM. Site and nature of airway obstruction in chronic obstructive lung disease. N Engl J Med 1968;278:1355–1360.
Hogg JC, Chu F, Utokaparch S, et al. The nature of small-airway obstruction in chronic obstructive pulmonary disease. N Engl J Med 2004;350:2645–2653.
Nakano Y, Muller NL, King GG, et al. Quantitative assessment of airway remodeling using high-resolution CT. Chest 2002;122:271S–275S.
Nakano Y, Wong JC, de Jong PA, et al. The prediction of small airway dimensions using computed tomography. Am J Respir Crit Care Med 2005;171:142–146.
Hersh CP, Demeo DL, Lange C, et al. Attempted replication of reported chronic obstructive pulmonary disease candidate gene associations. Am J Respir Cell Mol Biol 2005;33:71–78.
Ingenito EP, Evans RB, Loring SH, et al. Relation between preoperative inspiratory lung resistance and the outcome of lung-volume-reduction surgery for emphysema. N Engl J Med 1998;338:1181–1185.
Moore G. Cramming more components onto integrated circuits. Electronics 1965;38:114–117.
Moore G. Lithography and the future of Moore's law. Proc SPIE 1995;2440:2–17.
Hoffman E. Available from: http://dpi.radiology.uiowa.edu/vida/vidahome.html (accessed April 2, 2006).
Kikinis RJF, Grimson E, Wells W. Available from: www.slicer.org (accessed April 2, 2006).
Muller NL, Staples CA, Miller RR, Abboud RT. "Density mask": an objective method to quantitate emphysema using computed tomography. Chest 1988;94:782–787.
Coxson HO, Rogers RM, Whittall KP, et al. A quantification of the lung surface area in emphysema using computed tomography. Am J Respir Crit Care Med 1999;159:851–856.
Gevenois PA, de Maertelaer V, De Vuyst P, Zanen J, Yernault JC. Comparison of computed density and macroscopic morphometry in pulmonary emphysema. Am J Respir Crit Care Med 1995;152:653–657.
Kemerink GJ, Kruize HH, Lamers RJ, van Engelshoven JM. CT lung densitometry: dependence of CT number histograms on sample volume and consequences for scan protocol comparability. J Comput Assist Tomogr 1997;21:948–954.
Coxson HO, Whittall KP, Nakano Y, et al. Selection of patients for lung volume reduction surgery using a power law analysis of the computed tomographic scan. Thorax 2003;58:510–514.
Nakano Y, Muro S, Sakai H, et al. Computed tomographic measurements of airway dimensions and emphysema in smokers: correlation with lung function. Am J Respir Crit Care Med 2000;162:1102–1108.
de Jong PA, Muller NL, Pare PD, Coxson HO. Computed tomographic imaging of the airways: relationship to structure and function. Eur Respir J 2005;26:140–152.
Tufte E. Envisioning information. Cheshire, CT: Graphics Press; 1990.(John Reilly)