Computed Tomographic Measurements of Airway ... - ATS Journals

Computed Tomographic Measurements of Airway Dimensions and Emphysema in Smokers Correlation with Lung Function YASUTAKA NAKANO, SHIGEO MURO, HIROAKI SAKAI, TOYOHIRO HIRAI, KAZUO CHIN, MITSUHIRO TSUKINO, KOICHI NISHIMURA, HARUMI ITOH, PETER D. PARÉ, JAMES C. HOGG, and MICHIAKI MISHIMA Department of Respiratory Medicine, Graduate School of Medicine, Kyoto University, Kyoto, Japan; Departments of Physical Therapeutics, and Radiology and Nuclear Medicine, Kyoto University Hospital, Kyoto, Japan; and University of British Columbia Pulmonary Research Laboratory, St. Paul’s Hospital, Vancouver, British Columbia, Canada

Chronic obstructive pulmonary disease (COPD) is characterized by the presence of airflow obstruction caused by emphysema or airway narrowing, or both. Low attenuation areas (LAA) on computed tomography (CT) have been shown to represent macroscopic or microscopic emphysema, or both. However CT has not been used to quantify the airway abnormalities in smokers with or without airflow obstruction. In this study, we used CT to evaluate both emphysema and airway wall thickening in 114 smokers. The CT measurements revealed that a decreased FEV1 (%predicted) is associated with an increase of airway wall area and an increase of emphysema. Although both airway wall thickening and emphysema (LAA) correlated with measurements of lung function, stepwise multiple regression analysis showed that the combination of airway and emphysema measurements improved the estimate of pulmonary function test abnormalities. We conclude that both CT measurements of airway dimensions and emphysema are useful and complementary in the evaluation of the lung of smokers.

Chronic obstructive pulmonary disease (COPD) is characterized by the presence of airflow obstruction caused by emphysema or airway narrowing, or both (1). Cigarette smoking, which is the major risk factor of COPD, causes the abnormalities by inducing inflammation in the lung parenchyma and peripheral airways. A large number of studies have been done to detect and quantify pulmonary emphysema using computed tomography (CT) (2–18). Low attenuation areas (LAA) on CT scans in vivo have been shown to represent macroscopic or microscopic emphysematous changes in the lungs of patients (2–7). Although the CT measurement of LAA correlates well with diffusing capacity, the relationship to measurements of airflow obstruction is less significant (5–15) presumably because airflow obstruction is related to both loss of recoil and inflammatory narrowing of the airways. Recently some studies have been done to measure airway dimensions using CT (19–27). The measurement of airway wall thickness by high-resolution CT (HRCT) in patients with asthma has been shown to correlate with the severity of asthma (22). It has been suggested that CT has potential power to evaluate airways in patients with obstructive pulmonary disease (28). However, to our knowledge, there are no published CT studies of airway dimensions and comparison

(Received in original form July 26, 1999 and in revised form March 28, 2000) Supported by the Research Council of Respiratory Insufficiency, Ministry of Public Health, Japan, and the Ministry of Education, Science, Sports and Culture, Japan. Correspondence and requests for reprints should be addressed to Michiaki Mishima, M.D., Ph.D., Department of Physical Therapeutics (Respiration), Kyoto University Hospital, 53 Shogoin, Kawahara-cho, Sakyo-ku, Kyoto, 606-8397, Japan. E-mail: [email protected] Am J Respir Crit Care Med Vol 162. pp 1102–1108, 2000 Internet address: www.atsjournals.org

with pulmonary function tests (PFTs) in smokers with or without airflow obstruction. We hypothesized that both the measurement of emphysema and airway dimensions using CT in smokers would correlate with PFTs. We also hypothesized that the measurements of airway dimensions by CT would provide additive value to the measurement of emphysema in the prediction of lung function abnormalities. In the present study, we used HRCT to quantify emphysema and thin-section helical (spiral) CT to quantify airway dimensions for comparison with measurements of PFTs in 114 smokers.

METHODS Subjects A total of 114 smokers (94 patients with COPD from Kyoto University Hospital and 20 asymptomatic volunteers) were recruited. A diagnosis of COPD was made according to the criteria of the American Thoracic Society (1). No patients with large bullae or other lung diseases were included in the study. All the patients were clinically stable at the time of the study, and inhalation of ␤-receptor agonist or anticholinergic drugs, or both, were withheld for at least 12 h before the study. None of the asymptomatic volunteers had a history of respiratory illness and none had respiratory symptoms at the time of the study. All subjects had PFTs and CT scans on the same day. This study was approved by the ethics committee of Kyoto University and informed consent was obtained from each participant.

PFTs PFTs were performed with a Chestac-65V (Chest MI Corp., Tokyo, Japan). Forced vital capacity (FVC), forced expiratory volume in one second (FEV1), residual volume (RV) using the helium dilution method, peak expiratory flow rate (PEFR), total lung capacity (TLC), carbon monoxide diffusing capacity (DLCO), and alveolar volume (VA) using the single-breath method were measured. FVC, FEV1 and PEFR were expressed as percentage of predicted values (FVC%P, FEV1%P, and PEFR%P, respectively) according to the prediction equations of the Japanese Society of Chest Diseases (29). The ratio of FEV1 to FVC (FEV1/FVC) and the ratio of RV to TLC (RV/TLC) were expressed as percentages. The ratio of DLCO to VA (DLCO/VA) was expressed as the actual measurement values.

CT Scans We used HRCT to quantify LAA and thin-section helical (spiral) CT to quantify airway dimensions. Both LAA and airway analyses could be done on a helical scan but because there are no available previous data for LAA using helical CT, we used both HR and helical CT in the present study. Both scans were performed in a supine position using the same CT scanner (X-Vigor; Toshiba, Tokyo, Japan). No contrast media were used. CT scans were obtained after deep inspiration. This minimized the influence of variable hyperinflation, allowed us to compare the results under the same conditions, and optimized breathholding time. The HRCT was performed using 2-mm collimation, scan time 1.0 s, 120 kVp, and 200 mA. The images were reconstructed on a 32-cm

1103

Nakano, Muro, Sakai, et al.: CT Measure of Airways and Emphysema field of view (FOV) using a lung algorithm (FC83). Three HRCT scans were used for determination of LAA; a cranial (upper) section was obtained 1 cm above the superior margin of the aortic arch, a middle section was taken at 1 cm below the carina, and a caudal (lower) section was taken approximately 3 cm above the top of the diaphragm (8, 14–17). The helical scan was performed using 120 kVp, 50 mA, 3-mm collimation, and pitch 1.0. Images were reconstructed using the FC10 algorithm at 2-mm spacings. A targeted reconstruction of the right lung was performed using a subject-specific FOV (153 to 214 mm). We used thin-section helical CT for the analysis of airways to accurately select the same anatomic location (origin of the apical bronchus of the right upper lobe) and to reduce the subject’s radiation dose. Each HRCT or helical CT image was composed of a 512 ⫻ 512 matrix of numeric data (CT numbers) in Hounsfield units (HU). These CT data were transferred to a PowerPC personal computer via a magneto-optical disk, and were analyzed using custom software written in C programming language (Symantec C⫹⫹, Symantec Corp., CA).

Analysis of LAA The three HRCT scans were used for the analysis of LAA. Using a method previously reported (8, 14–17), the percentage of low attenuation pixels (LAA%) was calculated automatically. The cutoff level between normal lung density and LAA was defined as ⫺960 HU (8, 14–18). It has been reported that it is useful to divide the patients with COPD into two groups; one with LAA% less than 30 and the other with LAA% more than 30 (15). Thus, we divided the subjects into three groups; COPD patients with LAA% less than 30%, COPD patients with LAA% more than 30%, and the asymptomatic smokers. We designated these three groups as COPD ⭐ 30, COPD ⬎ 30, and Asymptomatics, respectively.

Analysis of Airway The trunk of the apical bronchus of the right upper lobe is usually sliced in cross section and easily identified on CT scans. This airway was chosen because it was large enough to be accurately measured and anatomically reproducible between subjects (see VALIDATION OF THE AIRWAY ANALYSIS). Images containing the apical bronchus to the right upper lobe were selected by a consensus reading of three pulmonologists (Y.N., S.M., and H.S.). With the bronchus identified, the following parameters were measured automatically on the computer (Figure 1); luminal area (Ai), short radius (SR), and long radius (LR) of the lumen, and airway wall thickness (T). Details are described in the APPENDIX. In brief, the following procedures were performed: (1)

The lumen of the bronchus was identified using a threshold ⫺500 (27). The area of the lumen was considered as Ai. (2) SR and LR were defined as the shortest and longest distances from the centroid point of the lumen to the edge. (3) From the centroid point of the lumen, 128 rays fanning out over 360⬚ were examined to determine T along the rays using the “full-width at half-maximum” principle (25, 26). (4) Those rays which projected onto the adjacent vessel were excluded. T was calculated from the nonexcluded rays. Because each subject had one to five images (2.6 ⫾ 1.0, mean ⫾ SD) through the apical sectional bronchus, airway and wall thickness measurements were averaged to yield final Ai and T. Assuming that, in a cross-sectional plane, the airway lumen is a true circle and T is constant throughout the wall, the total diameter (D) of the bronchus was calculated as D ⫽ 2 ( Ai ) ⁄ π ⫹ 2T. We calculated the outer area of the bronchus (Ao) from D; Ao ⫽ π (D/2)2. Airway wall area (WA) was calculated as Ao ⫺ Ai. We also used two previously reported indices to compare airway dimensions: the ratio of airway wall thickness to total diameter (the T/D ratio) (20–22) and the percentage wall area [WA% ⫽ (WA/Ao) ⫻ 100] (19, 22). To test the validity of using the airway dimensions of a single airway as a measure of airway thickening, we chose 20 subjects randomly from the total of 114 smokers and measured airway dimensions on all the airways that were sectioned in cross-section in the helical CT scans. The average WA%, for the airways other than the apical segmental bronchus, was calculated for each individual and compared with the WA% of the apical segmental bronchus.

Validation of the Airway Analysis The airway analysis software and algorithm were tested using a phantom. The phantom was made from a polystyrene foam block and eight plastic tubes, which represent lung parenchyma and airways respectively. The actual tube dimensions were measured using an optical micrometer caliper to the closest 0.01 mm. The plastic tubes ranged from 0.55 mm to 2.25 mm in wall thickness and from 8.0 mm2 to 335.1 mm2 in luminal area. Helical scans of the phantom were obtained and reconstructed on a 200-mm FOV. The pixel size at this FOV is 0.39 ⫻ 0.39 mm.

Statistical Method All statistical analyses were done using Stat View software (SAS Institute Inc., Cary, NC). Results were expressed as mean ⫾ SD. Univariate (linear) regression analysis and stepwise multiple regression analysis were used to evaluate the relationship between the CT parameters and PFTs. Mann-Whitney U test and analysis of variance (ANOVA) with Bonferroni/Dunn correction were used to compare groups. Adjustments for multiple comparisons (Bonferroni adjust-

Figure 1. A representative case showing the process of airway analysis. Using a helical CT image containing the apical bronchus to the upper lobe (A), the algorithm defines the luminal area (B) and wall thickness (C). Note that the identification of the airway wall thickness was successfully performed even when the pulmonary blood vessel ran parallel to the bronchus.

1104

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE

ments) were performed in both the univariate and the multivariate correlations. A value of p less than 0.05 was considered significant.

The maximal CT density of the plastic tubes was 49 ⫾ 102 HU, which was virtually identical with the 50 ⫾ 49 HU in the 10 randomly selected images of the right apical bronchus. There were significant positive relationships between the actual and CT-measured wall thickness (r ⫽ 0.975, p ⬍ 0.0001) and between the actual and CT-measured luminal area (r ⫽ 1.00, p ⬍ 0.0001). Figure 2 shows the error of CT measurements expressed as follows: Error (%) ⫽ (actual measurement ⫺ CT measurement)/ actual measurement ⫻ 100 The results showed that our system and algorithm are accurate within the range of 1.0 mm to 2.3 mm in thickness and within the range of 12 mm2 to 335 mm2 in luminal area. Studies of 114 Smokers

All the subjects were male and were either current (53/114) or ex-smokers. Anthropometric and pulmonary function data are shown in Table 1. The results of CT measurements are shown in Table 2. Because body size varied considerably, T, Ai, and

2000

TABLE 1 ANTHROPOMETRIC AND PULMONARY FUNCTION DATA IN 114 SMOKERS

RESULTS Validation Study of the Airway Phantoms

VOL 162

Age, yr Smoking index, pack-years Height, cm Weight, kg Body surface area, m2 FVC, %predicted FEV1, %predicted FEV1/FVC, % PEFR, %predicted RV/TLC, % DLCO/VA, ml/min/mm Hg/L

Mean ⫾ SD

Range

68 ⫾ 9 71 ⫾ 43 162 ⫾ 7 57 ⫾ 8 1.61 ⫾ 0.13 74 ⫾ 20 48 ⫾ 28 44 ⫾ 17 67 ⫾ 28 46 ⫾ 11 3.81 ⫾ 1.23

32–88 5–268 145–180 40–80 1.29–1.92 35–127 8–124 11–86 25–155 24–70 1.51–7.94

Ao were normalized using body surface area (BSA) and BSA . BSA was chosen because of the similarity of the units (mm2 for airway walls and m2 for BSA). However, similar results were obtained when we normalized airway measurements with height. The mean SR/LR ratio (airway roundness ratio) was 0.86 ⫾ 0.11. None of the 20 asymptomatic smokers had chronic bronchitis (by definition) and 59% of the obstructed patients had chronic bronchitis. However, there were no significant differences between the patients with and without chronic bronchitis in terms of T, T/ BSA , Ai, Ai/BSA, Ao, Ao/BSA, WA%, and T/D ratio. There were also no significant relationships between smoking index and airway dimensions. Table 3 shows the correlation coefficients (r values) for univariate regression analyses of FEV1%P and CT parameters. As expected, we observed an inverse relationship between LAA% and FEV1%P. The data for airway measurements suggest that the airway wall is thicker and the luminal area smaller in patients who have more severe airflow obstruction. However, there was no significant relationship between Ao and FEV1%P either before or after the correction for BSA. There were no correlations between LAA% and WA% or between LAA% and T/D ratio. Therefore we adopted WA% as a primary measurement of airway abnormality and LAA% as an index of emphysema. An average of 2.6 additional airways per case (range of 1 to 6) was measured in the subset of 20 subjects. The mean luminal area of the additional airways was 9.6 mm2 (range 2.4 to 24.7 mm2). The relationship between the WA% in the apical segmental bronchus and the average WA% in the additional airways was highly significant (r ⫽ 0.860, p ⬍ 0.0001, Figure 3).

TABLE 2 CT MEASUREMENTS OF AIRWAYS AND EMPHYSEMA IN 114 SMOKERS

Figure 2. Percent error for luminal area (A) and wall thickness (B) in the phantom study. The error of the smallest plastic tube (0.55 mm in thickness and 8.0 mm2 in actual area) was 27% in area and ⫺81% in thickness. However, the errors of other phantom tubes were within the range of ⫺2% to 4% in area and ⫺11% to 9% in thickness.

LAA% T, mm T/ BSA , mm/m Ai, mm2 Ai/BSA, mm2/m2 Ao, mm2 Ao/BSA, mm2/m2 T/D ratio WA%, %

Mean ⫾ SD

Range

31.7 ⫾ 15.1 1.5 ⫾ 0.2 1.2 ⫾ 0.2 17.3 ⫾ 7.6 10.8 ⫾ 4.7 47.0 ⫾ 12.9 29.4 ⫾ 8.3 0.204 ⫾ 0.034 64.5 ⫾ 7.9

4.2–70.6 1.2–2.3 0.9–1.9 3.8–39 2.3–24.3 18.1–84.1 10.8–52.4 0.144–0.289 49.4–82.2

Definition of abbreviations: Ai ⫽ area of the inner lumen; Ao ⫽ area of the bronchus; BSA ⫽ body surface area; D ⫽ diameter of the bronchus; LAA% ⫽ percentage of the low attenuation area to the corresponding lung area; T ⫽ thickness of the bronchial wall; WA% ⫽ percentage of the wall area.

1105

Nakano, Muro, Sakai, et al.: CT Measure of Airways and Emphysema TABLE 3 CORRELATION COEFFICIENTS (r VALUES) OF UNIVARIATE REGRESSION ANALYSIS FOR FEV1 (% PREDICTED) WITH CT MEASURED PARAMETERS IN 114 SMOKERS

LAA% T, mm T/ BSA , mm/m Ai, mm2 Ai/BSA, mm2/m2 Ao, mm2 Ao/BSA, mm2/m2 T/D ratio WA%, %

r Values

p Value*

⫺0.529 ⫺0.199 ⫺0.298 0.273 0.192 0.195 0.065 ⫺0.333 ⫺0.338

⬍ 0.001 NS ⬍ 0.01 ⬍ 0.05 NS NS NS ⬍ 0.001 ⬍ 0.001

Definition of abbreviation: NS ⫽ not significant. For other abbreviations, see Table 2. * p Values were adjusted for multiple comparisons.

There were 37 subjects in the COPD ⭐ 30 group, 57 in the COPD ⬎ 30 group, and 20 in the Asymptomatics. The mean ⫾ SD of WA% was 69 ⫾ 8 in COPD ⭐ 30, 63 ⫾ 7 in COPD ⬎ 30, and 59 ⫾ 5 in Asymptomatics. There were significant differences between COPD ⭐ 30 and COPD ⬎ 30 (p ⬍ 0.0001), between COPD ⭐ 30 and Asymptomatics (p ⬍ 0.0001), and between COPD ⬎ 30 and Asymptomatics (p ⬍ 0.05). Figures 4A and 4B show the relationships for FEV1%P versus LAA% and FEV1%P versus WA%. COPD ⭐ 30, COPD ⬎ 30, and Asymptomatics are identified by separate symbols. There were significant negative relationships for FEV1%P and LAA% and for FEV1%P and WA%. These data show that emphysema and airway wall thickening increased as FEV1%P decreased. Table 4 shows univariate and stepwise multiple regression analyses of LAA% and WA% versus measurements of lung function. In the univariate regression analysis LAA% correlated significantly with FEV1%P, FEV1/FVC, PEFR%P, RV/ TLC, and DLCO/VA but not FVC%P, whereas WA% correlated significantly with FVC%P, FEV1%P, PEFR%P, and RV/TLC but not FEV1/FVC or DLCO/VA. It is also apparent from univariate regression analyses that LAA% correlates more closely with FEV1%P and FEV1/FVC, whereas WA% correlates more closely with PEFR%P and RV/TLC. To evaluate whether the measurements of airway dimensions by CT

Figure 4. Relationship for FEV1 (percentage of predicted) versus LAA% (A) and versus WA% (B) in 114 smokers. FEV1 correlated negatively with both LAA% (LAA% ⫽ ⫺0.29 ⫻ FEV1 ⫹ 45.7, r ⫽ ⫺0.529, p ⬍ 0.001) and WA% (WA% ⫽ ⫺0.095 ⫻ FEV1 ⫹ 69.1, r ⫽ ⫺0.338, p ⬍ 0.001). Solid diamonds: COPD patients with the LAA% more than 30% (COPD ⬎ 30); open circles: COPD patients with the LAA% less than 30% (COPD ⭐ 30), open triangles: asymptomatic smokers (Asymptomatics).

would provide additive value to the measurements of LAA in the estimate of lung function tests, we compared the results of univariate and stepwise multiple regression analysis (Table 4). The results showed that FVC%P, FEV1%P, FEV1/FVC, PEFR%P, and RV/TLC were predicted better by a combination of LAA% and WA% than by LAA% or WA% alone.

TABLE 4 CORRELATION COEFFICIENTS (r VALUES) OF UNIVARIATE AND STEPWISE MULTIPLE REGRESSION ANALYSES FOR PFTs USING LAA% AND WA% Univariate Regression Analysis

FVC, % predicted FEV1, % predicted FEV1/FVC, % PEFR, % predicted RV/TLC, % DLCO/VA, ml/min/mm Hg/L

Figure 3. Relationship between WA% in the right apical bronchus and average WA% in the additional airways. There was a significant relationship between these two measurements (average WA% in the additional airways ⫽ 0.57 ⫻ WA% in the right apical bronchus ⫹ 36.7, r ⫽ 0.860, p ⬍ 0.0001).

LAA%

WA%

Multiple Regression Analysis LAA% and WA%

⫺0.159† ⫺0.529* ⫺0.650* ⫺0.395* 0.378* ⫺0.683*

⫺0.437* ⫺0.338* ⫺0.192† ⫺0.487* 0.422* 0.030†

0.482* 0.659* 0.700* 0.660* 0.597* NA

Definition of abbreviations: WA% ⫽ percentage of the wall area; LAA% ⫽ percentage of the low attenuation area to the corresponding lung area; NA ⫽ stepwise multiple regression analysis showed no additional predictive value of including WA%. * p ⬍ 0.001; p values were adjusted for multiple comparisons. † Not significant.

1106


However, WA% added no predictive value to LAA% for DLCO/VA.

DISCUSSION The measurement of in vivo airway dimensions using CT has been reported in animal models (24, 25) and in human subjects (19–23, 27). In most studies HRCT images and a visual assessment method were used (19–24). The visual assessment method is time-consuming (22) and intraobserver or interobserver error is not trivial. We used a computer-based method to measure airway dimensions. This technique is faster and more reproducible. In the present study, we used thin-section helical CT images to assess the airway dimensions because the contiguous images allowed more accurate identification of the right apical segmental bronchus. It has been reported that HRCT can accurately assess the dimensions of hollow tubes of 1 to 5 mm in diameter with a thickness from 0.5 to 2 mm (24, 30). We used the “full-width half-maximum” technique (25, 26) to estimate the wall thickness (see APPENDIX) and used a single threshold (⫺500 HU) (27) to measure the luminal area. It has been reported that this method is suitable for airways with a large inside diameter and thick walls but can produce an underestimate of luminal area and an overestimate of wall thickness in small airways and airways with thin walls (26). Therefore, we chose to measure a relatively large and thickwalled segmental bronchus. The validation study of the airway phantoms using the thin-section helical CT technique showed that our system and algorithm are accurate within the range of 1.0 to 2.3 mm for wall thickness and 12 to 335 mm2 for luminal area. However, our method overestimates the thickness when it is less than 1.0 mm and underestimates the luminal area when it is less than 12 mm2. Because the ranges of the airway measurements in this study were from 1.2 to 2.3 mm for thickness and from 3.8 to 39.0 mm2 for luminal area, underestimation of the luminal area may affect the result. However, when we limited the analysis to only those subjects whose luminal area was greater than 12 mm2 (86/114 subjects), the results of the analysis were unchanged. We measured a single specific airway in the present study because measuring airway wall dimensions without respect to standardization of anatomic location could present problems; specifically wall area as a percent of wall and lumen area varies by anatomic location in the tracheobronchial tree (31). Further, our aim was to test whether a relatively simple measurement that could be added to conventional CT scanning would be of predictive value for airflow obstruction. To test the generalizibility of the measurements on this specific airway, we randomly chose 20 subjects from the total group and measured WA% on all other airways that were sectioned in cross-section in the helical slices to compare the average WA% in airways other than the apical segmental bronchus with the apical segmental bronchus. The highly significant relationship between WA% in the apical segmental bronchus and the additional airways (Figure 3) indicates that the dimensions of this airway are representative of other airways in the same subject. However, the intercept was not zero and the slope was not unity. The reason that the intercept is positive is likely due to the fact that smaller airways have a relatively greater airway wall area (31). We cannot explain why the slope is less than one. One possibility is that airway walls can only thicken to a certain extent. There appears to be a maximal WA% of approximately 80. If this is the case, this could explain the slope. Our finding of thickened cartilaginous airways in smokers is consistent with the results of other investigators. Haraguchi

VOL 162

2000

and colleagues (32) showed that the bronchi of patients with COPD had more degenerated cartilage and perichondrial fibrosis than a control group. They also showed that the degree of perichondrial fibrosis correlated with the thickness of the epithelial basement membrane. Tiddens and colleagues (33) evaluated airway wall dimensions in cartilaginous airways of patients with COPD. They found that the wall area of central airways was increased in obstructed patients (FEV1/FVC ⫽ 40%) in comparison to nonobstructed patients (FEV1/FVC ⫽ 80%). They also showed that the increase in cartilaginous airway wall area correlated significantly with a semiquantitative measure of peripheral airway inflammation. It has also been shown that the small airways of patients with mild COPD have thicker walls and narrower lumens than control subjects (31, 34). Because we made only a single airway measurement in most subjects, we cannot comment on the heterogeneity of changes in airway dimensions. However, such heterogeneity would decrease the chances of finding a relationship between physiologic measurements and morphometric measurements on a single airway. The finding of a significant correlation between the airway dimensions and measurements of airflow obstruction suggests that large airway dimensions may reflect small airway pathology. Bosken and coworkers (34) found that smoking history was unrelated to the morphometric measures of peripheral airway dimensions. In the present study, there were not significant relationships between smoking index and airway dimensions. Our data show that airway wall area and thickness were greater and airway luminal area was smaller as FEV1%P decreased. In contrast, there was no significant relationship between Ao and FEV1%P. These results suggest that as the airway thickens, it encroaches upon the lumen rather than expanding outward into the lung parenchyma. Alternatively, the airway wall thickening could be accompanied by a degree of airway smooth muscle shortening which could have the same effect. Airflow obstruction in smokers is caused by loss of lung recoil and airway narrowing. In any one subject it is difficult to define the relative contributions of these mechanisms to the airflow obstruction. Emphysema, and its CT surrogate LAA%, correlates with loss of recoil but to date there have been no objective measurements of airway narrowing by CT in smokers. Our results show that measures of large airway thickening correlate with abnormalities of lung function except FEV1/FVC and DLCO/VA. Wall thickening correlated especially well with FVC%P and RV/TLC, suggesting that airway wall thickening is associated with gas trapping. However, because RV was measured using a helium dilution method, which may lead to the underestimation of RV in severely obstructed individuals, careful interpretation is required. The correlation of wall area and peak expiratory flow suggests that wall thickening influences maximal flow early in forced expiration when peak flow occurs. The correlation of WA% and FEV1%P also suggests that wall thickening and airway narrowing of large airways influences airflow obstruction. One possibility is that large airway wall thickening is simply an indicator of more peripheral airway inflammation and narrowing because this is the site of the major obstruction in smokers (35). However, the finding that WA% correlates most closely with PEFR%P, a measure known to be influenced by large airway caliber, suggests a direct effect of large airway remodeling on expiratory flow. The results of the stepwise multiple regression analysis suggest that WA% and LAA% are measuring independent aspects of pulmonary pathophysiology in smokers: the airway and parenchymal components, respectively. There were signif-

1107

Nakano, Muro, Sakai, et al.: CT Measure of Airways and Emphysema

icant differences in WA% between COPD ⭐ 30 and COPD ⬎ 30, between COPD ⭐ 30 and Asymptomatics, and between COPD ⬎ 30 and Asymptomatics. These results suggest that at comparable levels of FEV1, the patients with more emphysema have less severe airway thickening than those with less severe emphysema. However, even those with more severe emphysema had thicker airways than the asymptomatic smokers. In summary, the novel finding in this study is that measurements of airway dimensions in a single airway using a protocol that can be easily added to CT assessment of emphysema yield additive information about the mechanism of airflow obstruction in smokers. Acknowledgment: The authors thank Drs. M. Okazawa, A. Niimi, H. O. Coxson, and J. R. Mayo for useful suggestions. They also thank Mr. R. Tanaka, Ms. M. Morimoto, Mr. H. Akazawa, and Mr. N. Narai for their technical assistance with CT.

References 1. ATS Statement. 1995. Standards for the diagnosis and care of patients with chronic obstructive pulmonary disease. Am. J. Respir. Crit. Care Med. 152(Suppl.):S77–S83. 2. Bergin, C., N. Müller, D. M. Nichols, G. Lillington, J. C. Hogg, B. Mullen, M. R. Grymaloski, S. Osborne, and P. D. Paré. 1986. The diagnosis of emphysema: a computed tomographic–pathologic correlation. Am. Rev. Respir. Dis. 133:541–546. 3. Müller, N. L., C. A. Staples, R. R. Miller, and R. T. Abboud. 1988. “Density mask”: an objective method to quantitate emphysema using computed tomography. Chest 94:782–787. 4. Gevenois, P. A., V. de Maertelaer, P. De Vuyst, J. Zanen, and J.-C. Yernault. 1995. Comparison of computed density and macroscopic morphometry in pulmonary emphysema. Am. J. Respir. Crit. Care Med. 152:653–657. 5. Gevenois, P. A., P. De Vuyst, V. de Maertelaer, J. Zanen, D. Jacobovitz, M. G. Cosio, and J.-C. Yernault. 1996. Comparison of computed density and microscopic morphometry in pulmonary emphysema. Am. J. Respir. Crit. Care Med. 154:187–192. 6. Gevenois, P. A., P. De Vuyst, M. Sy, P. Scillia, L. Chaminade, V. de Maertelaer, J. Zanen, and J.-C. Yernault. 1996. Pulmonary emphysema: quantitative CT during expiration. Radiology 199:825–829. 7. Kuwano, K., K. Matsuba, T. Ikeda, J. Murakami, A. Araki, H. Nishitani, T. Ishida, K. Yasumoto, and N. Shigematsu. 1990. The diagnosis of mild emphysema: correlation of computed tomography and pathology scores. Am. Rev. Respir. Dis. 141:169–178. 8. Sakai, N., M. Mishima, K. Nishimura, H. Itoh, and K. Kuno. 1994. An automated method to assess the distribution of low attenuation areas on chest CT scans in chronic pulmonary emphysema patients. Chest 106:1319–1325. 9. Lamers, R. J., G. R. Thelissen, A. G. Kessels, E. F. Wouters, and J. M. van Engelshoven. 1994. Chronic obstructive pulmonary disease: evaluation with spirometrically controlled CT lung densitometry. Radiology 193:109–113. 10. Gurney, J. W., K. K. Jones, R. A. Robbins, G. L. Gossman, K. J. Nelson, D. Daughton, J. R. Spurzem, and S. I. Rennard. 1992. Regional distribution of emphysema: correlation of high-resolution CT with pulmonary function tests in unselected smokers. Radiology 183:457–463. 11. Kinsella, M., N. L. Müller, R. T. Abboud, N. J. Morrison, and A. DyBuncio. 1990. Quantitation of emphysema by computed tomography using a “density mask” program and correlation with pulmonary function tests. Chest 97:315–321. 12. Coxson, H. O., R. M. Rogers, K. P. Whittall, Y. D’Yachkova, P. D. Paré, F. C. Sciurba, and J. C. Hogg. 1999. A quantification of the lung surface area in emphysema using computed tomography. Am. J. Respir. Crit. Care Med. 159:851–856. 13. Nishimura, K., K. Murata, M. Yamagishi, H. Itoh, A. Ikeda, M. Tsukino, H. Koyama, N. Sakai, M. Mishima, and T. Izumi. 1998. Comparison of different computed tomography scanning methods for quantifying emphysema. J. Thoracic Imaging 13:193–198. 14. Nakano, Y., H. Sakai, S. Muro, H. Hirai, Y. Oku, K. Nishimura, and M. Mishima. 1999. Comparison of low attenuation areas on CT between inner and outer segments of the lung in COPD patients: incidence and contribution to lung function. Thorax 54:384–389. 15. Mishima, M., T. Hirai, H. Itoh, Y. Nakano, H. Sakai, S. Muro, K. Nishimura, Y. Oku, K. Chin, M. Ohi, T. Nakamura, J. H. T. Bates, A. M.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29. 30.

31.

32.

33.

34.

35.

Alencar, and B. Suki. 1999. Complexity of terminal airspace geometry assessed by lung CT in normal subjects and patients with chronic obstructive pulmonary disease. Proc. Natl. Acad. Sci. U.S.A. 96:8829–8834. Mishima, M., Y. Oku, K. Kawakami, N. Sakai, M. Fukui, T. Hirai, K. Chin, M. Ohi, K. Nishimura, H. Itoh, M. Tanemura, and K. Kuno. 1997. Quantitative assessment of the spatial distribution of low attenuation areas on X-ray CT using texture analysis in patients with chronic pulmonary emphysema. Front. Med. Biol. Eng. 8:19–34. Mishima, M., H. Itoh, H. Sakai, Y. Nakano, S. Muro, T. Hirai, Y. Takubo, K. Chin, M. Ohi, K. Nishimura, K. Yamaguchi, and T. Nakamura. 1999. Optimized scanning conditions of HRCT in the follow-up of pulmonary emphysema. J. Comput. Assist. Tomogr. 23:380–384. Mishima, M., T. Hirai, Z. Jin, Y. Oku, N. Sakai, Y. Nakano, H. Sakai, K. Chin, M. Ohi, K. Kawakami, K. Shimada, H. Itoh, K. Yamaguchi, T. Sawa, and K. Kuno. 1997. Standardization of low attenuation area versus total lung area in chest X-ray CT as an indicator of chronic pulmonary emphysema. Front. Med. Biol. Eng. 8:79–86. Okazawa, M., N. Müller, A. E. McNamara, S. Child, L. Verburgt, and P. D. Paré. 1996. Human airway narrowing measured using high resolution computed tomography. Am. J. Respir. Crit. Care Med. 154:1557–1562. Boulet, L.-P., H. Turcotte, G. Carrier, M. Boutet, and M. Laviolette. 1995. Increased maximal airway response to methacholine during seasonal allergic rhinitis in nonasthmatic subjects: relationships with airway wall thickness and inflammation. Eur. Respir. J. 8:913–921. Boulet, L.-P., M. Bélanger, and G. Carrier. 1995. Airway responsiveness and bronchial-wall thickness in asthma with or without fixed airflow obstruction. Am. J. Respir. Crit. Care Med. 152:865–871. Awadh, N., N. L. Müller, C. S. Park, R. T. Abboud, and J. M. FitzGerald. 1998. Airway wall thickness in patients with near fatal asthma and control groups: assessment with high resolution computed tomographic scanning. Thorax 53:248–253. Senéterre, E., F. Paganin, J. M. Bruel, F. B. Michel, and J. Bousquet. 1994. Measurement of the internal size of bronchi using high resolution computed tomography (HRCT). Eur. Respir. J. 7:596–600. McNamara, A. E., N. L. Müller, M. Okazawa, J. Arntorp, B. R. Wiggs, and P. D. Paré. 1992. Airway narrowing in excised canine lungs measured by high-resolution computed tomography. J. Appl. Physiol. 73:307– 316. Amirav, I., S. S. Kramer, M. M. Grunstein, and E. A. Hoffman. 1993. Assessment of methacholine-induced airway constriction by ultrafast highresolution computed tomography. J. Appl. Physiol. 75:2239–2250. Reinhardt, J. M., N. D. D’Souza, and E. A. Hoffman. 1997. Accurate measurement of intrathoracic airways. IEEE Trans. Med. Imaging 16:820– 827. McNitt-Gray, M. F., J. G. Goldin, T. D. Johnson, D. P. Tashkin, and D. R. Aberle. 1997. Development and testing of image-processing methods for the quantitative assessment of airway hyperresponsiveness from high-resolution CT images. J. Comput. Assist. Tomogr. 21:939–947. King, G. G., N. L. Müller, and P. D. Paré. 1999. Evaluation of airways in obstructive pulmonary disease using high-resolution computed tomography. Am. J. Respir. Crit. Care Med. 159:992–1004. Japanese Society of Chest Diseases. 1993. Standards of pulmonary function tests for Japanese. Jpn. J. Thorac. Dis. 31:appendix. Webb, W. R., G. Gamsu, S. D. Wall, C. E. Cann, and E. Proctor. 1984. CT of a bronchial phantom: factors affecting appearance and size measurements. Invest. Radiol. 19:394–398. Kuwano, K., C. H. Bosken, P. D. Paré, T. R. Bai, B. R. Wiggs, and J. C. Hogg. 1993. Small airways dimensions in asthma and in chronic obstructive pulmonary disease. Am. Rev. Respir. Dis. 148:1220–1225. Haraguchi, M., S. Shimura, and K. Shirato. 1999. Morphometric analysis of bronchial cartilage in chronic obstructive pulmonary disease and bronchial asthma. Am. J. Respir. Crit. Care Med. 159:1005–1013. Tiddens, H. A. W. M., P. D. Paré, J. C. Hogg, W. C. J. Hop, R. Lambert, and J. C. de Jongste. 1995. Cartilaginous airway dimensions and airflow obstruction in human lungs. Am. J. Respir. Crit. Care Med. 152:260–266. Bosken, C. H., B. R. Wiggs, P. D. Paré, and J. C. Hogg. 1990. Small airway dimensions in smokers with obstruction to airflow. Am. Rev. Respir. Dis. 142:563–570. Hogg, J. C., P. T. Macklem, and W. M. Thurlbeck. 1968. Site and nature of airway obstruction in chronic obstructive lung disease. N. Engl. J. Med. 278:1355–1360.

APPENDIX The step-by-step algorithm for airway measurement is as follows:

1108


1. Identification of the inner lumen of the bronchus. (i) The airway selected for measuring is identified and one pixel in the 512 ⫻ 512 matrix inside the lumen of this bronchus is labeled as a “seed pixel.” (ii) All the pixels, which are contiguous with the “seed pixel” and whose CT numbers are below ⫺800 HU, are labeled and defined as the temporary lumen. This lumen is smaller than the actual lumen because the threshold of ⫺800 HU is very low. (iii) The contiguous one pixel thick layer surrounding the temporary lumen is then examined. If, in this layer, more than 30% of pixels have a CT number below a threshold value of ⫺500 HU (27), then the pixels whose CT numbers are below ⫺500 HU are added to the temporary lumen. This step is necessary to avoid the failure of recognition of luminal area, which may occur in the presence of “bridging” pixels between the lumen and lung parenchyma. Preliminary study of 30 cases revealed that the minimum proportion to avoid this “bridging” phenomenon was 30%. (iv) The process described in the preceding step is repeated until the percentage of pixels in this layer that have CT numbers below ⫺500 HU become less than 30% of the total. (v) The algorithm defines all the labeled pixels as the final lumen (Figure 1B). 2. Measurements of the area (Ai), centroid, short radius (SR), and long radius (LR) of the lumen. (i) The single pixel area is calculated from the FOV. The single pixel area (mm2) ⫽ [FOV (mm)/512] ⫻ [FOV (mm)/512] (ii) The area of the lumen (Ai) is calculated by multiplying the pixel area. Ai ⫽ (number of the pixels labeled as the final lumen) ⫻ (pixel area) (iii) All the coordinates of the pixels inside the lumen are averaged and the results are considered to be the coordinates of the centroid. (iv) From this centroidal pixel, circles with various radii are drawn.

VOL 162

2000

(v) The radius of the biggest circle that is completely inside the airway lumen is considered to be the SR of the bronchial lumen. (vi) The radius of the smallest circle that is completely outside the airway lumen is considered to be the LR. (vii) The airway roundness is assessed by SR/LR ratio. 3. Measurement of the airway wall thickness (T). (i) A ray is cast outward from the centroid to the airway wall and parenchyma. (ii) The CT numbers along the ray are examined to determine the inner and outer boundaries of the airway wall using the “full-width at half-maximum” principle (25, 26). The profile of the CT numbers along this ray has a local minimum (LM#1) in the lumen, a maximum (Max) in the wall, and a local minimum (LM#2) in the parenchyma. The middle value (“half-maximum”) between the LM#1 and Max is calculated and considered as the inner boundary. Similarly, the middle value between Max and LM#2 is calculated and considered as the outer boundary. In this process the boundaries are estimated to the size of one-tenth of the pixel using the linear regression method. The distance between the inner and outer boundaries is defined as the thickness of airway wall along this ray. (iii) This process is repeated along 128 different direction rays and calculates 128 values of thickness. (iv) Since the apical bronchus runs parallel to a pulmonary blood vessel whose CT numbers are too high to be separated from the airway wall, the wall thickness cannot be adequately measured in this region of apposition. Therefore, among the 128 values, the thicknesses longer than 2.5 mm are excluded. Preliminary study of 30 cases confirmed that using a thickness value longer than 2.5 mm identified blood vessels by visual observation. (v) After this exclusion, the mean value and the SD of the remainder thicknesses are calculated. Finally, only the values within the range of mean ⫾ 2SD are used to define thickness (T) (Figure 1C).