Emphysema: Definition, Imaging, and Quantification

Transcription

1 1017 Benjamin Felson Lecture Emphysema: Definition, Imaging, and Quantification William M. Thurlbeck1 and Nestor L. Muller2 This review will discuss imaging of the chest in patients with pulmonary emphysema. Imaging findings must be related to the structure of the lung because emphysema is defined in anatomic terms. Accordingly, we first review the anatomic definitions of emphysema and its consequences and then review the Imaging findings, with emphasis on CT, in patients with this disease. The more severe the morphologic emphysema, the more likely a radiographic diagnosis will be made, no matter what criteria are used. The criterion of arterial deficiency is specific but insensitive. The criteria used to assess overinflation are sensitive but not specific. CT can be used for both qualitative and quantitative assessment of emphysema. The presence and extent of emphysema can be determined by visual assessment of areas of abnormally low attenuation or by objective quantification based on the attenuation values. Statistically significant correlations between emphysema and CT findings have been shown in numerous studies, but mild morphologic emphysema may be missed by CT, and occasionally CT scans give false-positive findings. In patients with moderate to severe emphysema, the severity of emphysema is underestimated on the basis of CT findings by a factor of approximately three when compared directly with results of pathologic examination of lung specimens. In spite of these limitations, CT is the best way of recognizing emphysema in living patients and probably has a significant role in recognizing localized emphysema that is amenable to surgical treatment. Definitions Emphysema is defined by the American Thoracic Society [1] as follows: Emphysema is a condition of the lung characterized by abnormal, permanent enlargement of the air spaces distal to the terminal bronchiole, accompanied by destruction of their walls. Similar definitions have been adopted by the World Health Organization [2] and a Ciba Guest Symposium [3]. The latter, however, included air space enlargement due to dilatation of the alveoli as well as that due to destruction of their walls. Neither abnormal enlargement nor destruction was defined more precisely. Destruction Any definition of destruction should be clear, practical, and sensitive. Several criteria have been suggested, including fenestrae (the presence of abnormal holes), the destructive index, the loss of alveolar surface area, mean linear intercept (Lm) and air space wall per unit volume (AWUV), and loss of alveolar attachments (loss of bronchiolar traction). Fenestrae.-In 1962, Boren [4] described discontinuities of alveolar walls seen on thick sections of human lungs. These consisted of holes normally present in alveolar walls of humans and animals (the pores of Kohn) and abnormal holes referred to by Boren as fenestrae. He suggested that holes larger than 20 iim in diameter were abnormal and constituted evidence of destruction, although why he chose this size is uncertain. Some 30 years later, Nagai et al. [5], using scanning electron microscopy, quantified alveolar wall discontinuities in nonemphysematous human lungs and found that 94% of holes were 10 j.tm or less in diameter and that only 0.2% were more than 20.tm in diameter. These observations substantiated Boren s observation, but clearly standard error of the mean is too expensive and time consuming to be used for practical definition. The destructive index.-the destructive index (DI) is a recent innovation [6] suggested as a criterion for alveolar wall destruction. The DI is assessed using microscopic slides and Received April 12, 1994; accepted after revision June 7, Presented at the meeting of the Society of Thoracic Radiology, Scottsdale, AZ, March This work was supported by the Medical Research Council of Canada, grant MT 7124, and the British Columbia Lung Association. 1Department of Pathology and Laboratory Medicine, Koemer Acute Care Pavilion, GF227, Vancouver Hospital, 2211 Wesbrook Mall, Vancouver, BC V6T 2B5 Canada, and Department of Pathology, British Columbia s Childrens Hospital, Vancouver, BC, Canada. Address correspondence to W. M. Thurlbeck at Vancouver Hospital. 2Department of Radiology, University of British Columbia and Department of Radiology, Vancouver Hospital, 855 W. 12th Ave., Vancouver, BC V5Z 1M9 Canada. AJR 1994;163: X/94/ American Roentgen Ray Society

2 1018 THURLBECK AND MULLER AJR:163, November 1994 has three components: breaks in alveolar walls (DIb), probably corresponding to fenestrae; type II cell metaplasia of alveolar walls, often accompanied by some alveolar wall fibrosis (Dlf); and classic emphysema (DIe), but the criteria for this were not described. Dlf was subsequently found to be irrelevant [7]. The more obvious the macroscopic emphysema, or emphysema on specimens visualized using the unaided eye, the greaterthe component of DIe (Fig. 1). Thus the Dl probably has little value when macroscopic emphysema is obvious [8]. However, Saetta et al. [6] showed that DIb could be increased in smokers in whom air space enlargement was not present (defined as an Lm < 350 iim). This observation still must be explored, because an increased DIb also may be present when there are abnormalities in elastic properties of the lung in smokers in whom macroscopic emphysema is absent [8-10] or when there is an increased number and size of fenestrae in the intervening normal lung in emphysematous lungs [5]. The measurement is tedious. Loss of alveolar surface area.-the popularization by Weibel [11] and Dunnill [12, 13] of morphometric techniques in the lung made measurement of alveolar surface area practical. Loss of alveolar surface area intuitively seems a good method of measuring destruction. However, even when corrected for stature-related variations in alveolar surface area, measurement of loss of alveolar surface area is an insensitive test for emphysema; approximately 43% of lungs with C C) e C) a Dl0 > Emphysema score (Panel) Fig. 1.-Graph shows destructive index (Di) In patients without emphysema (emphysema score, 0), with mild emphysema (emphysema score,> 0-25), and with moderate emphysema (emphysema score, 30-60). Dl consists of breaks In alveolar walls (DIb), alveolar wall type ii cell metaplasia (Dl1), and classic microscopic emphysema (Dlv). When emphysema is absent, total Di consists of breaks in alveolar wails only (Dib). When emphysema is mild, Di increases and is mostly breaks; when emphysema is moderate, Di consists mostly of Die. When emphysema is more severe, Di is aimost entirely DIe. Dif Is irrelevant. obvious macroscopic emphysema have surface area values within normal limits [14]. Mean linear intercept and air space wall per unit volume.- Lm is the length of a test line placed over histologic slides of the lung divided by the number of times the line crosses alveoar walls (not surfaces). It is an approximation of air space (alveolar ducts, alveolar sacs, and alveoli) size; it is not the mean chord length of alveoli. Alveolar surface area (SA) is calculated from the formula 4V/Lm [11-15]. In this formula, V is the volume of the lung in which Lm is measured. Rearrangement ofthis formula shows that SAN (AWUV) is 4/Lm. Lm and AWUV [16, 17] are thus reciprocals. Lm is older and often has been used as a measure of experimental emphysema. Lm will increase if alveolar walls are destroyed, because their loss results in fewer intercepts, or if there is overinflation, which causes a greater distance between intercepts. Neither Lm nor AWUV is a sensitive method for recognizing emphysema. Thirty-two percent of patients with emphysema have normal (nonemphysematous) Lm values [14]. Only 26% of surgically resected lungs have abnormal AWUV values [17], whereas up to 100% [18] of similar lungs have macroscopic emphysema (typically 73-87% [9]), and in the series in which AWUV was measured, this was 49% [17]. Thus, neither AWUV nor Lm can be regarded as the essential defining characteristic of emphysema [19]. Although AWUV and Lm are easy to measure, they require random sampling of the lung inflated to a standard pressure; corrections for shrinkage of lung tissue during fixation and processing; and histologic examination of microscopic fields, which takes at least 1 hr. Loss of alveolar attachments (loss of bronchiolar traction).- Long considered to be a cause of airflow obstruction in emphysema [20, 21], especially on expiration, loss of alveolar attachments to bronchioles has recently been described as representing an early stage of the destruction of lung parenchyma [22]. Bronchioles are thought to be tethered by their attachment to alveolar walls, and associations have been described between loss of alveolar wall attachments and the percentage of alveoli destroyed [23] or the presence and severity of macroscopic emphysema [24, 25]. Although loss of attachments is a cause of airflow obstruction and may be associated with tortuosity and irregularity [25] and ellipticality [26] of bronchioles, it is an indirect way of recognizing destruction and emphysema. The peripheral part ofthe acinus (the gas exchanging unit of the lung distal to the terminal bronchiole) must be the area involved because this is the part ofthe acinus that abuts bronchovascular bundles as well as lobular septa and pleura. If loss of attachments is an early stage of emphysema, distal acinar or paraseptal emphysema may be more important than is usually thought. Bronchiolar inflammation may extend to adjacent penbronchiolar alveolar walls and lead to their destruction [22]. The National Institutes of Health definition of destruction.-a committee of the National Heart, Lung, and Blood Institute stated in 1985 that destruction in emphysema is defined as nonuniformity in the pattern of air space enlargement so that the orderly appearance of the acinus and its components is disturbed and may be lost [27]. This means that the acinus looks odd; thus the definition of emphysema is subjective. The committee also stated that destruction should not be accompanied by obvious

3 AJR:163, November 1994 IMAGING FINDINGS IN EMPHYSEMA 1019 fibrosis. Their meaning is uncertain because fibrosis, assessed 100% specificity and 90% sensitivity between grades 0, I, and histochemically and biochemically, is present in irregular, distal II compared with grades Ill and IV panacinar emphysema), acinar, and centnlobular emphysema [28]. The qualification of different measurements of emphysema have been used both obvious fibrosis was added to exclude the air space enlargement radiologically and pathologically to reach these conclusions. that occurs in end stage ( honeycomb ) lung. In general, in all the studies of radiologic-pathologic correla- We conclude that no good reason exists for altering the original definition of emphysema or macroscopic emphysema. Macroscopic is an unfortunate, perhaps pejorative, term because it implies a casual examination of the lung. Nothing could be further from the truth; careful examination is required. The minimum requirements are careful examination of a slice of inflated lung; knowledge of the normal appearance; and standards to grade the emphysema as mild, moderate, or severe. Paper-mounted whole lung sections [29] show emphysema much better, and these led to the original classic description of centrilobular emphysema by Leopold and Gough [30]. Sensitivity is increased further ifthe lung is exammed with a hand lens or dissecting microscope, and this examination is facilitated if air space walls are rendered opaque by barium sulphate [31] or Bouin s fluid [32]. If many lungs are to be examined for epidemiologic purposes [33], then a simple pictorial grading system in which emphysema is scored from 0 to 100 at intervals of 5 or 10 [34] will suffice. This is usually referred to as the panel-grading system and often has been used to correlate pathologic with CT findings. Scores of 5-25 indicate mild emphysema; 30-50, moderate emphysema; and equal to or greater than 60, severe emphysema. For more detailed purposes such as functional-anatomic correlation, gross examination should be supplemented by histologic examination of random and specific blocks of tissue and by more objective assessments such as Lm measurements. Not unexpectedly, the prevalence of emphysema in random cases depends on the care with which the lung is examined. For example, inclusion of minimal grades of emphysema increased the prevalence of emphysema from 50% to 73% in one autopsy series [35], and another study found that emphysema was almost universally present even in young subjects [36]. Another study has defined a category of trace emphysema [37], and the inclusion or exclusion of this category affects the prevalence of emphysema. Air Space Enlargement The definition of air space enlargement by the National Heart, Lung, and Blood Institute s committee [27] is an increase in air space size as compared with the air space of normal lungs. Data concerning Lm and AWUV measurements are now sufficient for these measurements to be used in practice. The term macroscopic emphysema assumes that the air spaces are enlarged. The values available from nonemphysematous lungs can be used to assess whether a particular lesion shows air space enlargement. The concepts arising from these definitions are shown in Figure 2. Radiographic Diagnosis Radiologic-pathologic correlations have been reviewed in detail [9, 36, 38]. Although considerable precision has been claimed in several studies (including one [39] that showed 0Normal Lint Non-smoker AWUV4 7 I Smoker DI? (Dlbt) Emphysema Di? (Diet Dibt) Lmt AWUV4 Fig. 2.-Diagrams show various concepts of emphysema. A normal nonsmoker (upper left) shows an air space of normal size with three breaks. A smoker (upper right) shows normal air space size with more breaks (Dib), Increasing destructive index (DI). Air space enlargement (lower left) shows increase in mean linear distance (Lm), reduction in air space wall per unit volume (AWUV), and normal number of breaks. In emphysema (lower right), air space shows obvious morphologic abnormalities (wavy lines), Increase In size of air spaces, and more breaks. Die = classic microscopic emphysema. g4 is 6 21 a) > Emphysema severity group Fig. 3.-As emphysema increases in severity (emphysema severity group), radiologic signs of emphysema increase in frequency. Overinf Iation (Inf), expressed as 0-4+, increases through the groups. Arterial deficiency (AD), a qualitative assessment, is present in mild emphysema, but also is present in only minority of most other groups. Percentage of cases with mixed increased markings (IM) and AD emphysema and percentage with im emphysema also are shown. In emphysema severity groups 4 and 5, 100% of cases show radlologic evidence of emphysema [40]. C a) 0 0.

4 1020 THURLBECK AND MULLER AJR:163, November 1994 tions, the severer the emphysema, the more likely the radiologic diagnosis will be made. Figure 3 shows many of the problems. Overinflation is expressed quantitatively (on a scale of 0 to 4+) and increases progressively with increasing severity of emphysema [40]. Arterial deficiency and increased markings were described as present or absent. Arterial def i- ciency was recognized when segmental vessels tapered more rapidly than normal as they proceeded distally and when the peripheral markings were sparse. By contrast, increased markings were recognized when the peripheral pulmonary vascular markings were not only increased in size but were more numerous, particularly peripherally [40]. The presence of arterial deficiency had a low sensitivity but a high specificity [41]. In addition, this sign is poorly reproducible between and within observers, especially by a nonradiologist [42]. When increased markings are included, the diagnosis of emphysema was made in 100% of patients with moderate and severe emphysema. Emphysema with increased markings was found especially in patients with cor pulmonale. The classic appearance of emphysema (arterial deficiency and overinflation) diminishes or disappears in patients with chronic airflow obstruction who have heart failure [43]. Arterial deficiency in the upper zones of the lung indicates centrilobular emphysema; in the lower zones, arterial deficiency indicates panacinar emphysema [41]. The meaning of the radiologic diagnosis of emphysema is also an issue. A radiologic diagnosis may indicate that the chronic airflow obstruction in a given patient is due to emphysema. For a given severity of emphysema, a radiologic diagnosis of emphysema is more likely to be made when obstruction of airflow is symptomatic [41], even if the radiologist is unaware of the clinical features of the case. The radiologic diagnosis of emphysema is usually on-off -a patient either has emphysema or does not. Morphologic emphysema, conversely, recognizes various grades of the disease. The radiologic diagnosis should be quantified in some wayfor example, as a mean of radiologic variables. For instance, the data of Lohela et al. [44] can be analyzed more closely and the mean number of radiologic signs calculated (Table 1). The mean number of radiologic signs increases with the severity of morphologic emphysema, and the assessments of TABLE 1: Mean Number of Radiologic Signs of Emphysema with Increasing Pathologic Emphysema Emphysema Score (Panel-Grading Method) Criteria of Sutinen et al. [45] Criteria of Nicklaus et al. [42}b Note.-Adapted from data of Lohela et al. [44]. acritena of Sutinen et al. include (1) blunting of costophrenic angle and/or diaphragm at or below 11th rib posteriorly, (2) irregular lucencies, (3) increased retrostemal space, and (4) flattening or concavity of the diaphragm. bcritena of Nicklaus et al. include the four used by Sutinen et al. plus arterial deficiency. Lohela et al. include the criteria of both Sutinen et al. [45] and Nicklaus et al. [42]. Lohela et al. also showed a specificity of 74% and a sensitivity of 61% for subjective opinions of the radiologic diagnosis of emphysema. CT Diagnosis Studies have shown that CT findings correlate with the presence and severity of morphologic emphysema better than do results of pulmonary function tests [46, 47]. Our review of the diagnostic accuracy of CT is limited to studies that correlated CT findings with pathologic specimens. CT allows direct visualization of areas of lung destruction and is superior to chest radiography in showing the presence, extent, and severity of emphysema. On CT scans, emphysema is characterized by the presence of areas of abnormally low attenuation (Fig. 4). The first CT-pathologic correlation in emphysema was reported in 1984 by Hayhurst et al. [48], who compared the CT appearance of the lungs with resected lung specimens from six patients who had mild centrilobular emphysema and from five patients without emphysema. They assessed the frequency distribution of the attenuation values on CT scans and showed that patients with emphysema had more pixels with attenuation values between -900 H and H than did patients without emphysema (p <.001, Wilcoxon test). Fig year-old woman with centriiobular emphysema. High-resolution T scan (1.5-mm collimation, high-spatial-frequency algorithm) of left upper lobe of lung shows localized areas of low attenuation near center of secondary lobules.

5 AJR:163, November 1994 IMAGING FINDINGS IN EMPHYSEMA 1021 In 1986, Foster et al. [49] reported findings in 25 patients who had CT while they were alive and who had lungs fixed postmortem by inflation. The CT scans were obtained with 1- cm collimation at 1-cm intervals through the chest. Three radiologists independently assessed the CT scans for the extent of nonperipheral areas of low attenuation, peripheral areas of low attenuation, pulmonary vascular pruning, and pulmonary vascular distortion. The CT criterion that best correlated with the presence and severity of centrilobular emphysema was the presence of nonperipheral areas of low attenuation (r=.84 in the upper part ofthe lung and r=.78 in the lower part of the lung). The criterion of nonperipheral areas of low attenuation allowed recognition of 13 of 15 patients with centrilobular emphysema and produced two of 1 0 false-positive results. Bergin et al. [50] assessed the accuracy of the CT diagnosis of emphysema in 32 patients who had surgery. The CT scans were obtained with 10-mm collimation at 10-mm intervals through the chest and were reviewed independently by two chest radiologists and a chest physician. The presence of areas of low attenuation and vascular disruption on CT scans was used to assess emphysema. The extent of emphysema seen on CT scans was quantified as normal (0), less than 25% involvement of the lung parenchyma (1), between 25% and 50% involvement (2), between 50% and 75% involvement (3), and greater than 75% involvement on all slices (4); the score was expressed as a percentage of maximum. The observers quantified the extent of emphysema for the entire lung and for the resected lobe. Severity of emphysema seen in the pathologic specimens was assessed using a modification of the panel-grading method [34]. The correlations between the CT scores for the total lung and the pathologic emphysema scores were.81 and.88 for the two radiologists and.63 for the chest physician (all p <.001). The correlations between the CT scores for the resected lobes and the pathologic specimens were 0.71 for both the radiologists and.57 for the chest physician. In all cases, the CT scores were lower than the corresponding pathologic scores. Miller et al. [32] obtained preoperative CT scans in 38 patients undergoing lobectomy or pneumonectomy. They Fig year-old man who underwent left upper iobectomy for bronchogenic carcinoma. A, CT scan (10-mm collimation) at level of aortic arch shows only a few questionable 10- callzed areas of low attenuation (arrows) consistent with emphysema, which had not been identified prospectively on CT scan. B, Pathologic specimen cut in same transverse plane at same level as CT scan shows numerous localized areas of centriiobular emphysema (arrows). Even In retrospect, most of these areas cannot be seen on CT scan. compared findings on 10-mm- and 1.5-mm-collimation CT scans with the pathologic findings in the corresponding transverse slice of lung. The CT scans were assessed by two independent observers, and the diagnosis was based on the presence of areas of low attenuation. Emphysema was assessed on CT scans and pathologically by superimposing a grid with squares corresponding to 1 cm2 on the CT scan. The extent of emphysema was expressed by the percentage of grid squares containing emphysema, and the severity was expressed by averaging the severity (0-4) in the grid squares. In addition, the severity of emphysema was assessed pathologically using the panel of standards [34]. Correlation between the CT assessment of emphysema and the pathologic panel score was high for the 10-mm-collimation (r=.81, p <.001) and the 1.5-mm-collimation (r=.85, p <.001) scans. Lower correlations were found between the CT scans and the pathologic grid scores (r =.70 for the 10-mm-collimation scans, r=.72 for the 1.5-mm-collimation scans, p <.001). The extent and severity of emphysema were consistently underestimated on the basis of CT findings. Furthermore, in one of five patients with no emphysema seen on pathologic examination, the CT scan was interpreted by one of the observers as showing mild emphysema; in six of 33 patients with emphysema, the emphysema was missed on CT scans by both observers. The six patients with emphysema missed on CT scans included four with mild centrilobular emphysema and two with mild to moderate panacinar emphysema. The authors concluded that CT consistently underestimated the extent of centriacinar and panacinar emphysema because most of the lesions less than 0.5 cm in diameter were missed [32]. Thus, mild emphysema may be missed on CT scans (Fig. 5), and the severity of emphysema may be underestimated (Fig. 6). Kuwano et al. [18] determined the accuracy of CT scans obtained with 1-mm collimation and with 5-mm collimation in the assessment of emphysema in 42 patients whose lobes were resected. The extent of emphysema was assessed both on CT scans and in the resected lobes using the panel of standards. The diagnosis of emphysema on CT scans was based on the presence of areas of low attenuationand vascu-

6 1022 THURLBECK AND MULLER AJR:163, November 1994 ar disruption. The correlations between the CT emphysema scores and the pathologic specimens were.68 for the 1 -mmcollimation scans and.76 for the 5-mm-collimation scans. The severity of emphysema found pathologically was consistently underestimated on the 5-mm-collimation scans but not on the 1 -mm-collimation scans. Significant interobserver and intraobserver variations were noted, and their average score was used. The authors concluded that high-resolution CT (1-mm collimation) can help identify the presence of mild emphysema. However, a major limitation of this study was that no patients had either a CT or pathology score of less than 10. The authors attributed this to the fact that they examined five slices with both CT and pathology studies and thus were likely to find minimal grades of emphysema. It seems unlikely that this is the whole explanation, because their experience is unique. More to the point, it is clear from the data that there is no correlation between pathology and CT scores of less than 20. Correlations from their data can be calculated, and these show r2 values of.13 for 1-mm-collimation scans and.00 for 5-mm-collimation scans. These findings explain the low correlations in all cases for both types of scans despite apparently good correlation between CT and pathologic findings. Thus, contrary to the claims of the authors, mild emphysema was consistently missed on CT scans. The authors also reported a correlation between CT and the DI; they clearly could not Fig year-oid man who underwent left upper lobectomy for bronchogenic carcinoma. A, High-resolution CT scan (1-mm collimation, high-spatial-frequency reconstruction algorithm) shows localized areas of low attenuation (arrows) consistent with centrilobular emphysema. Nodule represents bronchogenic carcinoma. B, Pathologic specimen cut 1 cm above tumor in same transverse plane as CT scan shows obvious moderately severe centriiobular emphysema. Although emphysema was seen on CT scan, severity was underestimated. recognize breaks in the alveolar wall (DIb) and were probably referring to DIe. Gould et al. [1 9] compared the lowest fifth percentile of Hounsfield numbers with both the lowest AWUV value on five microscopic fields and the percentage of the lung involved by emphysema. They studied 28 patients and found correlations between the Hounsfield numbers and AWUV (r = -.77), and r =.50 for the extent of emphysema. In their hands, diffusing capacity for carbon monoxide provided better correlations with AWUV and the extent of emphysema than did CT findings. Most studies correlating CT findings with results of pathologic examination of resected specimens assess the accuracy of CT in the diagnosis of centrilobular emphysema. Results can be assessed from two studies from different laboratories [32, 49] that together show a specificity of 83% and a sensitivity of 80%. However, the patterns of abnormality seen on CT scans differ for the various forms of emphysema [51]. In centrilobular emphysema, the abnormalities are usually most severe in the upper parts of the lung and include irregular small round or confluent areas of low attenuation interspersed with normal lung (Fig. 4). The localized areas of low attenuation are located near the center of the secondary pulmonary lobule. Panacinar emphysema involves mainly the lower lung zones and is characterized on CT scans by diffuse areas of Fig year-oid woman who underwent left lung transplantation for severe panacinar emphysema due to alpha 1-antiprotease deficiency. A, CT scan of native right lung shows simplification of lung parenchyma with diffuse areas of low attenuation and paucity of vascular marldngs compared with normal transplanted left lung. B, Chest radiograph is shown for comparison with CT scan.

7 AJR:163, November 1994 IMAGING FINDINGS IN EMPHYSEMA 1023 low attenuation with little intervening normal lung (Fig. 7). In severe panacinar emphysema, the simplification of lung parenchyma leads to diffuse areas of low attenuation with a paucity of vascular markings that allow distinction between panacinar emphysema and normal lung parenchyma. However, mild or moderately severe panacinar emphysema is often impossible to distinguish from adjacent normal parenchyma on CT scans [32]. Spouge et al. [52] determined the value of CT in assessing the presence of pathologically proved panacinar emphysema. They studied five patients without emphysema and 10 patients with panacinar emphysema, four of whom had severe emphysema requiring lung transplantation. Conventional 10-mm-collimation CT scans were obtained in 14 patients and high-resolution CT (1.5-mmcollimation high-spatial-frequency reconstruction algorithm) in nine. Emphysema was assessed pathologically using the panel-grading method. The correlation with pathology was r=.90 for conventional CT and r=.96 for high-resolution CT (p <.01 ). The extent of panacinar emphysema was consistently underestimated, although less so on high-resolution CT scans (slope, 0.67) than on conventional CT scans (slope, 0.47). A radiologic diagnosis of emphysema was made in six of nine patients with emphysema using conventional CT (sensitivity, 67%) and in five of seven patients with high-resolution CT (sensitivity, 71%). Thus, the sensitivity of the diagnosis of panacinar emphysema was slightly less than that found in centrilobular emphysema, but underestimation of extent was about the same. Distal acinar emphysema is characterized by areas of low attenuation in the subpleural lung regions and adjacent to vessels and interlobular septa (Fig. 8). Because centrilobular and distal acinar emphysema produce localized areas of low attenuation, they are easier to recognize on CT scans than panacinar emphysema [32, 52]. As expected, the correlation of isolated lung specimen findings on CT scans with pathologic findings is better than that for CT scans obtained in vivo. Hruban et al. [53], using V Fig. 8.-SO-year-old man with distal acinar emphysema. High-resolution CT scan shows localized areas of low attenuation predominantly in subpleural lung regions. Note presence of distal aclnar emphysema along azygos fissure. high-resolution CT, were able to identify even mild emphysema when they scanned 20 postmortem lung specimens. The correlation coefficient between the in vitro score and the pathologic grade was.91, with a slope close to 1. This excellent correlation is not surprising because conventional postmortem radiographs also show emphysema well [54]. Objective Quantification of Emphysema by CT Although the visual assessment of emphysema correlates well with the pathologic scores, the overall extent of emphysema is difficult to estimate visually because of the wide range of volumes represented in different images. This difficulty can be circumvented by highlighting areas of abnormally low attenuation using a computer program. Hayhurst et al. [48] first showed that patients with emphysema had more pixels with attenuation values between -900 H and H than did patients without emphysema (p <.01). The GE 9800 scanner has a standard software program called density mask that highlights voxels within any desired range. Muller et al. [55] compared the density mask with the visual assessment of emphysema in 28 patients undergoing lung resection for tumor. The pathologic score was obtained using a modification of the panel-grading system. Pathologically, seven patients had no emphysema, and 21 had emphysema scores ranging from 5 to 100. In each patient, a single representative CT scan was compared with the corresponding pathologic specimen of tissue. The authors assessed the accuracy of density masks, highlighting all voxels with attenuation values less than -920, -910, and -900 H. Correlation between the three different density mask scores and the pathologic assessment of emphysema was good (all r >.83, p <.001). The best correlations were observed by highlighting all voxels with attenuation equal to or less than -910 H. The correlation between the density mask and the pathologic score of emphysema was By comparison, the correlation between the mean of visual scores by two independent observers and the pathologic score of emphysema was.90 (p <.001). When the density mask at -910 H was used, three cases with emphysema scores of 10 were missed, and emphysema was diagnosed in one normal lung. By comparison, two independent chest radiologists on two separate occasions missed two cases of emphysema with pathologic scores of 10. The first observer missed one additional case and the second observer missed four additional cases with pathologic scores ranging from 5 to 20. Each observer diagnosed mild emphysema (visual score 10) in one healthy subject. The density mask allows objective quantification ofthe total volume of lung showing emphysema on CT scans and the percentage of lung involved with emphysema. Measurements of attenuation on CT have several limitations, including scanner type; calibration; kilovoltage; reconstruction algorithm; volume averaging; patient s size; and the location, environment, and size of the area being assessed. In spite of these limitations, because the visual assessment is based on areas of low attenuation, once this attenuation is determined, these areas can be highlighted and emphysematous changes can be quantified objectively. Recently, computer programs have

8 1024 THURLBECK AND MULLER AJR:163, November 1994 been developed that automatically determine lung volume and the percentage of emphysema from CT scans on the basis of areas with abnormally low attenuation [56]. These programs allow entire lungs to be analyzed within 1 mm. Indications for Use of CT In Assessing Emphysema Although CT is the most accurate method for diagnosing emphysema in vivo, its role in the clinical assessment of patients is limited because of its cost. In most cases, the diagnosis of emphysema can be made with a combination of clinical history, pulmonary function tests, and chest radiographs. Probably the main indication for the use of CT is the preoperative assessment of patients with large bullae who are being referred for bullectomy [57-59]. The best outcome after bullectomy is seen in patients who previously had a rapid progression of dyspnea, restrictive lung function due to compression of normal areas of lung, and absence of generalized emphysema. Large bullae can compress the remaining lung parenchyma and cause further functional and clinical impairment. CT allows assessment not only of the extent of bullae but also of the degree of compression and the severity of emphysema in the remaining lung (Fig. 9). CT has replaced angiography in assessing these patients. CT may also be indicated for assessing patients who have dyspnea and decreased carbon monoxide diffusing capacity without evidence of obstruction of airflow. Klein et al. [60] reported 10 such patients, all smokers with impaired gas exchange and normal findings on chest radiographs, but with evidence of emphysema on high-resolution CT scans. In such patients, high-resolution CT can help differentiate emphysema from pulmonary vascular disease [61]. Fig year-old man with severe shortness of breath. High-resolution CT scan shows large bullae in right lung compressing right middle lobe. Note shift of mediastinum to left. Although high-resolution CT scan shows evidence of emphysema In both lungs, patient underwent buliectomy because of compression of middle lobe and shift of mediastinum. Clinical and functional improvement after surgery was considerable. REFERENCES 1. American Thoracic Society. Chronic bronchitis, asthma and pulmonary emphysema: a statement by the Committee on Diagnostic Standards for Nontuberculous Respiratory Diseases. Am Rev Respir Dis 1962;85: World Health Organization. Chronic cor pulmonale: report of an expert committee. World Health Organization technical report series, No Geneva: World Health Organization, 1961:15 3. Ciba Guest Symposium Report: Terminology, definitions, and classification of chronic pulmonary emphysema and related conditions. Thorax 1959;14: Boren HG. Alveolar fenestrae: relationship to the pathology and pathogenesis of pulmonary emphysema. Am Rev Respir Dis 1962;85: Nagai A, Inano H, Matsuba K, Thuribeck WM. Scanning electron-microscopic observations of emphysema in the human: morphometnc and correlative studies. Am J Res Crlf Care Med (in press) 6. Saetta M, Shiner RJ, Angus GE, et al. Destructive index: a measurement of lung parenchymal destruction in smokers. Am Rev Respir Dis 1985; 131: Saito K, Cagel P, Berend N, Thurlbeck WM. The destructive index in nonemphysematous and emphysematous lungs. Am Rev Respir Dis 1989;139: Osbome 5, Hogg JC, Wright JL, Coppin C, Pare PD. Exponential analysis of the pressure-volume curve: correlation with mean linear intercept and emphysema in human lungs. Am Rev Respir Dis 1988;137: Thuribeck WM. Christie lecture: emphysema then and now. Can Resp J 1994;1: Berend N, Woolcock AJ, Marlin GE. Correlation between the function and structure of the lung in smokers. Am Rev Respir Dis I 979;1 19: Weibel ER. Morphomet,y of the human lung. Heidelberg: Springer, Dunnill MS. Quantitative methods in the study of pulmonary pathology. Thorax 1962;17: Dunnill MS. Evaluation of a simple method of sampling the lung for quantitative histological analysis. Thorax 1 964;19: Thuribeck WM. lntemal surface area and other measurements in emphysema. Thorax 1967;22: Underwood EE. Quantitative sterology. Reading, MA: Addison-Wesley, Gillooly M, Lamb D. Air space size in lungs of lifelong non-smokers: the effect of age and sex. Thorax 1993;48: Gillooly M, Lamb D. Microscopic emphysema in relation to age and smoking habit. Thorax 1993;48: Kuwano K, Matsuba K, lkeda T, et al. The diagnosis of mild emphysema: correlation of computed tomography and pathology scores. Am Rev Respir Dis 1990;141 : Gould GA, MacNee W, McLean A, et al. CT measurements of lung density in life can quantitate distal air space enlargement: an essential defining feature of human emphysema. Am Rev Respir Dis 1988;1 37: Anderson AE Jr, Foraker AG. Relative dimensions of bronchioles and parenchymal spaces in lungs from normal subjects and emphysematous patients. Am J Med 1962;32: Pratt PC, Haque A, Klugh GA. Correlation of postmortem function and structure in normal and emphysematous lungs. Am Rev Respir Dis 1961; 83: Saetta M, Ghezzo H, Kim WD, et al. Loss of alveolar attachments in smokers: a morphometric correlate of lung function impairment. Am Rev Respir Dis 1985;132: Wright JL, Hobson JE, Wiggs B, Pare PD, Hogg JC. Airway inflammation and peribronchiolar attachments in the lungs of nonsmokers, current, and ex-smokers. Lung 1 988;1 66: Petty TL, Silvers GW, Stanford RE. Radial traction and small airways disease in excised human lungs. Am RevRespirDisl986;133: Nagai A, Yamawaki I, Takizawa T, Thuribeck WM. Alveolar attachments in emphysema of human lungs. Am Rev Respir Dis : Lamb D, McLean A, Gillooly M, Warren PM, Gould GA, MacNee W. Relationship between distal air space size, bronchiolar attachments, and lung function. Thorax 1993;48: Snider GL, Kleinerman JL, Thurlbeck WM, Bengali ZH. The definition of emphysema: report of a National Heart, Lung, and Blood Institute, Division of Lung Disease Workshop. Am Rev Respir Dis 1985;132:

9 AJR:163, November 1994 IMAGING FINDINGS IN EMPHYSEMA Cardoso WV, Sekhon HS, Hyde DM, Thurbeck WM. Collagen and elastin in human pulmonary emphysema. Am Rev Respir Dis 1993;147: Gough J, Wentworth J. Thin sections of entire organs mounted on paper. Harvey Lect 1958;53: Leopold JG, Gough J. The centrilobular form of hypertrophic pulmonary emphysema and its relation to chronic bronchitis. Thorax 1957;12: Heard BH. A pathological study of the lungs with chronic bronchitis. Thorax 1958;13: Miller RR, MUller N, Vidal S, Morrison NJ, Staples CA. Limitation of cornputed tomography in the assessment of emphysema. Am Rev Respir Dis 1989;139: Thurlbeck WM, Ryder RC, Stemby N. A comparative study of the severity of emphysema in necropsy populations in three different countries. Am Rev Respir Dis 1974;109: Thurlbeck WM, Dunnill MS, Hartung W, Heard BE, Heppieston AG, Ryder AC. A comparison of three methods of measuring emphysema. Hum Pathol 1970;1 : Thurlbeck WM. The incidence of pulmonary emphysema: with observations on the relative incidence and spatial distribution of various types of emphysema. Am Rev Respir Dis 1963; 87: Thurlbeck WM. Chronic airflow obstruction in lung disease. Philadelphia: Saunders, Ryder AC, Thurlbeck WM, Gough J. A study of interobserver variation in the assessment of emphysema in paper-mounted whole lung sections. Am Rev Respir Dis 1969;99: Pratt PC. Role of conventional chest radiography in diagnosis and exdusion of emphysema. Am J Med 1987;82: Reid L, Millard FJC. Correlations between radiological diagnosis and structural lung changes in emphysema. C/in Radio/1964;15: Thurlbeck WM, Henderson JA, Fraser AG, Bates DV. Chronic obstructive lung disease: a comparison between clinical, roentgenologic, functional and morphologic criteria in chronic bronchitis, emphysema, asthma, and bronchiectasis. Medicine 1970;49: Thurlbeck WM, Simon G. Radiographic appearance of the chest in emphysema. A.JR 1978; 130: Nicklaus TM, Stowell DW, Chnstiansen WA, Renzetti AD Jr. The accuracy of the roentgenologic diagnosis of chronic pulmonary emphysema. Am Rev Respir Dis 1966; 93: Milne ENC, Bass H. Aoentgenologic and functional analysis of combined chronic obstructive pulmonary disease and congestive heart failure. Invest Radio! 1969;4: Lohela P, Sutinen S, P#{225}#{227}kko P, Lahti A, Tienan J. Diagnosis of emphysema on chest radiographs. Fortschr Geb Rontgenstrahlen Nuk!earmed Erganzungsbd 1984;141 : Sutinen S. Chnstofordis AJ, Klugh JA, Pratt PC. Aoentgenologic criteria for the recognition of non-symptomatic pulmonary emphysema: correlation between roentgenologic findings and pulmonary pathology. Am Rev Respir Dis 1965;91 : Morrison NJ, Abboud AT, Ramadan F, et al. Comparison of single breath carbon monoxide diffusing capacity and pressure-volume curves in detecting emphysema. Am Rev Respir Dis 1989;1 39: i Kinsella M, Muller NL, Abboud AT, Morrison NJ, ByBuncio A. Quantitation of emphysema by computed tomography using a density mask program and correlation with pulmonary function tests. Chest 1990;97: Hayhurst MD, Flenley DC, McLean A, et al. Diagnosis of pulmonary emphysema by computerized tomography. Lancet 1984;2: Foster WL Jr. Pratt PC, Roggli vl Godwin JD, Halvorsen RA, Putman CE. Centnlobular emphysema: CT-pathologic correlation. Radiology 1986;159: Bergin C, MUller NL, Nichols DM, et al. The diagnosis of emphysema: a computed tomographic-pathologic correlation. Am Rev Respir Dis 1986; 133: Bergin CJ, MUller NL, Miller AR. CT in the qualitative assessment of emphysema. J Thorac Imaging 1986;1 : Spouge D, Mayo JR, Cardoso W, MUller NL. Panacinar emphysema: CT and pathologic findings. J ComputAssist Tomogr 1993;17: Hruban RH, Meziane MA, zerhouni EA, et al. High resolution computed tomography of inflation-fixed lungs: pathologic-radiologic correlation of centrilobular emphysema. Am Rev Respir Dis 1987;136: Sutinen S, Lohela P, P#{227}#{225}kko P, Lahti A. Accuracy of postmortem radiography of excised air-inflated human lungs in assessment of pulmonary emphysema. Thorax 1982;37: MUller NL Miller AR, Abboud AT. Density mask : an objectve method to quantitate emphysema using computed tomography. Chest 1988;94: Archer DC, Coblentz CL, dekemp RA, Nahmias C, Norman G. Automated in vivo quantification of emphysema. Radiology 1993;188: Morgan MDL, Strickland B. Computed tomography in the assessment of bullous lung disease. BrJ Dis Chest 1984;78: Gaensler EA, Jederlinic PJ, FitzGerald MY. Patient work-up for bullectomy. J Thorac Imaging 1986;1 : Wade JF Ill, Mortenson T, Irgin CG. Physiologic evaluation of bullous emphysema. Chest 1991;100: Klein J, Gamsu G, Webb WA, Golden JA, Muller NL. High-resolution CT diagnosis of emphysema in symptomatic patients with normal chest radiographs and isolated low diffusing capacity. Radiology 1992;182: Webb WA. High-resolution CT of the lung parenchyma. Radio! Clin North Am 1989; 27: