CT scans were performed using a multidetector-row spiral CT scanner with four detector arrays (SOMATOME plus Volume Zoom; Siemens, Berlin, Germany). CT scans were acquired with the following parameters: 140 kVp, 150 mA, 4 detector × 1 mm collimation, and helical pitch 6 or 7. In this clinical study, the entire lung of each patient was scanned with patients in the supine position, holding their breaths at deep inspiration. All CT row data sets were reconstructed to isotropic voxel data using both soft-tissue and bone algorithms. The length of the 1-voxel side was 0.625 mm. Reconstructed data were transferred to the workstation, and then reconstructed into three-dimensional chest images (AZE Ltd., Tokyo, Japan). First, a three-dimensional bronchial skeleton was automatically reconstructed using a certain threshold level in Hounsefield units (HU). The threshold level was determined on an individual basis (−950 to −980 HU) to obtain airway images as distal as possible. If some portions of lung parenchyma remained with the skeleton, they were manually removed to prevent an analysis error. Finally, we could obtain an accurate bronchial skeleton (Figure 1A) and could identify any bronchus in the source images of axial, sagittal, and coronal slices (Figure 1B). The selected bronchial pathway was automatically converted to a curved MPR image as shown in Figure 1C. The bronchial long-axis image appeared as a straight pathway. Of note, we could obtain short-axis images that were exactly perpendicular to the long axis at any site. In this study, a segmental bronchus was defined as the third generation of bronchus (Figure 1D). On the monitor of the workstation, image interpretation was done with a window width of 1,000 and a window level of −700. From the centroid point of the lumen, rays fanning out over 360° were examined to determine airway wall thickness along the rays using the full width at half-maximum principle (16, 32). After this process, if the outline of automatically obtained airway walls was obviously out of contour, it was then corrected. Based on manual plotting at several points, our software used cubic spline interpolation and built a new circle. Finally, we could get values for the airway luminal area (Ai) and the outer area of the bronchus (Ao). The wall area percent was defined as WA % = (Ao − Ai) / Ao × 100 (Figure 2A). As airway dimensions may be influenced by physical characteristics, we also assessed Ai after correcting for body surface area. Assuming that the airway lumen is true circle and airway wall thickness is constant throughout the wall, the inner diameter (Di), the total diameter (Do), and airway wall thickness (T) were calculated as Di = 2√Ai/π, Do = 2√Ao/π and T = (Do − Di)/2. All these analyses were done by one of the authors (M.H.) who was blinded to the subjects' background data.

Figure 1. (A) Accurate bronchial skeleton. (B) Source images of axial, sagittal, and coronal slices. Yellow lines indicate the same series of airway in both A and B. (C) A curved MPR image of the selected airway (B8) from the right lower lobe. (D) Short-axis images obtained from the curved MPR image are precisely perpendicular to the long axis of the airway. The numbers of the bronchial generation are shown (see text for details).

Download Figure | Download Powerpoint

[More] [Minimize]

Figure 2. (A) A representative case shows the measurement of airway dimensions on the short-axis image. Airway wall thickness was determined using the full width at half-maximum principle; we could obtain the outline of airway walls automatically (see text for details). The schema of the short-axis image shows the airway luminal area (Ai) and the airway wall area (WA). The ratio of wall area was calculated as WA / the outer area of the bronchus (Ao) × 100. (B) A photograph of a column-shaped phantom with a 2-mm inner diameter and a 1-mm wall thickness (upper). A curved MPR image of the phantom and a short-axis image (lower). (C) A photograph of the sigmoid-shaped phantom with a 3-mm inner diameter (upper). A curved MPR image obtained from original CT data (middle). Nine short-axis images at various sites (lower).

Download Figure | Download Powerpoint

[More] [Minimize]

We did a validation study using three phantoms to test our newly developed software. The phantoms were made of acrylic resin, and their inner and outer diameters and wall thicknesses were optically accurate. Both phantom 1 and phantom 2 were cylindrical. Phantom 1 had an inner diameter of 2 mm and a wall thickness of 1 mm (Figure 2B). Phantom 2 had an inner diameter of 1.5 mm and a wall thickness of 1 mm. Phantom 3 had a sigmoid shape with an inner diameter of 3 mm and a wall thickness of 1 mm (Figure 2C). Phantom 3 was used to prove the accuracy of the algorithm, which should allow the short axis images to be obtained exactly perpendicular to the long axis at any site. When doing the CT scans, the phantoms were placed into polystyrene foam blocks, representing the lung parenchyma. CT scans were performed under the same conditions as those used in the clinical study.

The subjects in this study had been recruited from 2003 through 2005 as part of the Hokkaido COPD cohort study, the results of which will be the subject of a separate paper. A diagnosis of COPD was made based upon the GOLD guideline (1). Subjects who had an allergic diathesis, an episodic wheeze, or a history of bronchial asthma were excluded from this study. A total of 52 COPD patients in various clinical stages according to the GOLD criteria (stage I, n = 14; stage II, n = 22; stage III, n = 14; stage IV, n = 2) were enrolled. There were only two subjects with stage IV disease because we excluded patients on long-term oxygen therapy from our cohort study. All subjects gave their written informed consent. The subjects' characteristics are shown in Table 1. This study was approved by the Health Authority Research Ethics Committee of Hokkaido University.

Most pulmonary function tests were obtained on the same day as the CT, but some were done within a week before or after the CT. We used the rolling seal type of spirometers CHESTAC-33 (CHEST M.I., Inc., Tokyo, Japan). The results of pulmonary function tests met the requirements of the Japanese Respiratory Society guideline (33), which are similar to those of the ATS. In brief, three acceptable and two reproducible maneuvers were required from up to eight forced expirations. Inhalation of short-acting bronchodilators in the previous 12 h was prohibited; treatment with long-acting β₂-agonists or sustained-release theophyllines in the previous 24 h was also prohibited. A long-acting anticholinergic inhaler was not available in Japan at the time of this study. Acceptable maneuvers were defined as those with peak expiratory flow within 10% of the maximum observed, a rapid start, absence of major flow fluctuations, and adequate expiration time. Reproducible maneuvers agreed within 200 ml of the larger FEV₁. The FEV₁ and forced vital capacity (FVC) values taken to characterize each participant were the maximum results obtained from acceptable maneuvers. FEV₁ and FVC were expressed as percentage of predicted values according to the prediction equations of the Japanese Respiratory Society (33). The diffusing capacity of the lung for carbon monoxide (Dl_CO) based on the single-breath method was also measured in all subjects according to the pulmonary function test guideline of the Japanese Respiratory Society (33). Dl_CO divided by alveolar volume was expressed as percentage of predicted values according to the prediction equations of Burrows (34). Lung volumes—total lung capacity (TLC), functional residual capacity (FRC), and residual volume (RV)—were measured by the helium-closed-circuit method. Lung volumes were expressed as percentage of predicted values according to the prediction equations of Nishida (35).

All statistical analyses were done using Pearson correlation coefficients. Results are expressed as mean ± SD. Linear regression analysis was used to evaluate the relationship between the airway parameters and FEV₁ (% predicted). A value of p less than 0.05 was considered to be significant.

For phantom 1, Ai was 2.8 ± 0.1 mm² and WA was 9.8 ± 0.2 mm², both of which were close to the actual values (3.1 mm² and 9.4 mm², respectively) (Table 2). The coefficients of variation for Ai and WA were 2.8% and 2.2%, respectively; both were well within acceptable limits. On the other hand, for phantom 2, whose inner diameter was 1.5 mm, the coefficient of variation was 14.9% for Ai and 3.5% for WA; the coefficient of variation for Ai was not as good as that obtained with phantom 1, whose inner diameter was 2 mm. These data indicate that the measurement of airway dimensions by our new software is very accurate and reproducible for airways with an inner diameter ⩾ 2 mm. To confirm how accurately we could get precise short-axis images of bronchi that run in various directions in the lung, we obtained images of the sigmoid-shaped phantom (phantom 3). Axial images of phantom 3 obtained at any point were perfect, round shapes (Figure 2C), confirming that the software can be used for any bronchi. Data was obtained from 40 points of the sigmoid-shaped phantom, which had an Ai of 7.1 mm² and a WA of 12.6 mm²; the measured Ai was 6.8 ± 0.8 mm², and the WA was 13.2 ± 0.2 mm². The coefficients of variation for Ai and WA were 3.3% and 2.5%, respectively; both of these values were considered acceptable for human studies.

To measure airway dimensions in patients with COPD, we selected two bronchi, the apical bronchus (B1) of the right upper lobe and the anterior basal bronchus (B8) of the right lower lobe. The mean inner diameter of the sixth generation in both B1 and B8 was calculated to be approximately 2.3 mm. Because the phantom studies demonstrated that the airways with an inner diameter as small as 2.0 mm could be accurately measured, we decided to sample airways from the third to the sixth generations. For B1, Ai of the third to the sixth generation ranged from 19.5 ± 12.1 mm² to 4.3 ± 2.5 mm² (mean ± SD); for B8, Ai ranged from 16.3 ± 7.4 mm² to 4.3 ± 2.4 mm² (Table 3). The measured value of WA and the calculated values of Di and T are also listed in Table 3. The Ai of all generations except for the third generation of B1 had a significant positive correlation with FEV₁ (% predicted). The WA% of all generations except the third and fourth generations of B1 had a significant negative correlation with FEV₁ (% predicted). For both Ai and WA%, the correlation coefficients improved as the airways became smaller. Furthermore, in comparisons of the Ai and WA% of B1 with those of B8, the B8 indices generally had better correlations with FEV₁ (% predicted) than the B1 indices (Table 3). Correcting the Ai by body surface area made little difference in terms of r values, although the corrected r values were slightly lower in comparison with the values before correction. Representative data illustrate the characteristic differences between the third and sixth generations in Figure 3. For correlations of the other parameters of lung function with the airway dimensions see Tables E1 and E2 in the online supplement. FVC (% predicted), FEV₁/FVC, RV (% predicted), and RV/TLC, which are the parameters for airflow limitation, showed similar tendency with Ai or WA% for FEV₁ (% predicted) in both B1 and B8. However, FEV₁ (% predicted) showed generally better correlations with airway dimensions than any other pulmonary function parameter. On the other hand, none of the airway dimension parameters showed a statistically significant correlation with Dl_CO/alveolar volume, TLC (% predicted), or FRC (% predicted).

Figure 3. Representative data illustrate the characteristic difference between the third and sixth generations. Relationship of FEV₁(% predicted), airway luminal area (Ai), and wall area percent (WA %) in 52 patients with COPD. (A, B) Apical bronchus (B1) of the right upper lobe. (C, D) Anterior basal bronchus (B8) of the right lower lobe. Open circles, third generation; filled circles, sixth generation. The correlation coefficients between airflow limitation and airway dimensions are better in the sixth generation than in the third generation of both B1 and B8.

Download Figure | Download Powerpoint

[More] [Minimize]

Even in the same bronchial skeleton, at each bifurcation we had to select one branch among the two or three branches of the same generation and then decide on locating a short-axis image around the center between two successive bifurcations. This process allows one to obtain slightly different axial images, potentially causing unwanted bias. Thus, to assess the variance of the measurements for the same generation of bronchi in the same subjects, we performed the following study. We selected eight subjects with various COPD stages and measured Ai and WA% six times in each selected bronchus (Figure 4). The coefficients of variation for the measurements of Ai from the third to the sixth generation were 9.0 ± 4.3%, 10.5 ± 5.3%, 10.4 ± 5.0%, and 14.9 ± 9.1% in B1, and 6.5 ± 3.0%, 11.2 ± 5.4%, 11.5 ± 7.2%, and 13.0 ± 4.4% in B8. The coefficients of variation of WA% from the third to the sixth generation were 6.1 ± 3.2%, 6.5 ± 3.0%, 6.3 ± 3.0%, and 5.5 ± 4.1% in B1, and 6.2 ± 2.7%, 6.9 ± 4.2%, 7.2 ± 3.4%, and 5.8 ± 1.8% in B8. As with the results obtained from all 52 subjects, we found statistically significant correlations of FEV₁ (% predicted) and the two parameters of the sixth generation of B1 and B8. Additionally, both Ai and WA% in the third generation of the bronchi showed statistically significant correlations with FEV₁ (% predicted) in B8 but not in B1, which was again consistent with the data obtained from all the subjects. These results suggest that one-site measurement around the center of one targeted branch reasonably reflects the dimensions of each branch with negligible bias.

Figure 4. The same generation of bronchi was measured six times in eight patients with COPD. Bars indicate the averages of the individual data. (A, B) Apical bronchus (B1) of the right upper lobe. (C, D) Anterior basal bronchus (B8) of right lower lobe. Open circles, third generation; filled circles, sixth generation; Ai, airway luminal area; WA%, wall area percent.

Download Figure | Download Powerpoint

[More] [Minimize]

In this study, we developed new software using curved MPR for depicting a three-dimensional airway skeleton and measuring luminal areas of airways and wall areas very precisely at any site in the lungs. Using phantoms, we demonstrated that we could obtain accurate short images with an inner diameter as small as 2 mm and a wall thickness of 1 mm, and that the short-axis images obtained were precisely perpendicular to the long axis even with a sigmoid-shaped phantom. We thus did not need to consider the obliquity of targeted airways. We then measured airway dimensions in patients with COPD of various clinical stages. Although in this study we examined only two bronchi, one from B1 of the right upper lobe and the other from B8 of the right lower lobe, we demonstrated that the values of Ai and WA% correlated well with FEV₁ (% predicted). More importantly, the correlation coefficients between both airway parameters and FEV₁ (% predicted) improved as the airway size decreased from the third to the sixth generation both in B1 and B8. These results clearly support the concept that distal (small) airways rather than proximal (large) airways are the more important determinant of airflow limitation in COPD, as has been previously suggested by a number of pathology-function correlation studies in the past (11–14).

To the best of our knowledge, this is the first three-dimensional CT study to demonstrate the relationship between the airflow limitation index and the dimensions of various-sized airways in patients with COPD. It was the software using curved MPR that allowed us to do so. Although the idea of three-dimensional analysis is not entirely new (22, 36–38), it has not been applied to measuring airway dimensions in patients with COPD. In COPD, the quantitative evaluation of the airways using CT has received less attention than the analysis of emphysema. In fact, there have been only a few such studies (27–30). Nakano and colleagues were the first to report that wall thickening at B1 in the right upper lobe had a significant correlation with FEV₁ (% predicted) (27), and they subsequently demonstrated that measuring airway dimensions in large airways could roughly estimate small airway dimensions by comparing the CT findings of large airways with the histologic measurement of small airways in lungs excised from the same subjects (28). The important role of small airway lesions in COPD airflow limitation has been long recognized (11–14). In an earlier study, Hogg and colleagues used a retrograde catheter technique and were the first to report that the resistance in small airways of excised lungs from patients with COPD was greatly increased compared with normal lungs, in which 25% of the total airway resistance is accounted for by the small airways (11). They further demonstrated that the resistance correlated with the histologic findings of narrowing and obliteration of small airways along with mucous plugging. Very recently, Hogg and colleagues again demonstrated that, in surgically excised lungs, the severity of small airway disease paralleled the COPD clinical stage as defined by airflow limitation (14). The obstruction of the small airways consists of airway-wall thickening that occurs through a remodeling process and airway narrowing due to the accumulation of inflammatory cells and exudates in the lumen. The current study confirms that the inner area and, to a lesser extent, the wall area in distal (small) airways, rather than proximal (large) airways, are closely related to the severity of airflow limitation in vivo in patients with COPD. In contrast to Nakano and colleagues (27), we did not find a statistically significant correlation between FEV₁ (% predicted) and WA% in the third generation of B1, though a similar trend was observed in our study. However, the correlation coefficients that we obtained for the fifth and sixth generations were as high as 0.6 to 0.7, which are much better than the correlation coefficient (0.338) that they obtained in the third generation of B1 in their study.

Of note, the airway luminal area, Ai, is not a pure index of airway disease itself, particularly when measured in vivo. Rather, it is influenced by several direct or indirect factors. Certainly, airway-wall thickening contributes to airway narrowing (39, 40). Secretions or exudates within the airway luminal area may further contribute to airway narrowing. In addition, airway caliber is influenced by the pressure balance between the inside and the outside of the airway walls (1, 41). This would particularly hold true for small airways, which lack cartilage in their walls. In our study, we measured airway dimensions while subjects were holding their breath at deep inspiration. Therefore, the elastic recoil pressure of the surrounding tissue may also have influenced airway caliber. For instance, if the airway being measured was surrounded by lung destruction (emphysema), this would contribute to further airway narrowing as a consequence of lowered elastic recoil pressure in addition to the presence of airway disease. This factor may account for the relatively higher correlation coefficients of Ai and FEV₁ (% predicted) than of WA% and FEV₁ (% predicted) in our study. Unfortunately, our current software does not permit us to measure the dynamic change in airway caliber in association with respiration.

It should also be noted that the correlations of airflow limitation were generally better with the airway dimensions of B8 than with those of B1. This may be due to the fact that the volume of the lower lobe associated with B8 is larger than that of the upper lobe associated with B1. Consistent with this finding, it has been reported that the volume of the lower lobes contributes more to general pulmonary function than the volume of the upper lobes in patients with COPD (42).

Finally, we should mention some limitations of this study. First, we assessed only two bronchi, B1 and B8. Variation in airway dimensions may exist even in the lungs of normal subjects (43). On the other hand, King and colleagues recently reported that airway narrowing after a methacholine challenge varied in the large airways of asthmatics and normal subjects but not in the small airways (44), and Kim and colleagues reported that no statistically significant differences were found in the inner diameter of the bronchus divided by its outer diameter among segments, lobes, and lungs in normal subjects (45). Nevertheless, further extensive studies are definitely required to examine possible variations in airway dimensions in COPD. Second, we did not perform another validation study of the new software, which would have ideally compared the airway dimensions obtained by CT measurements with those of excised lung specimens. However, because the data we obtained using the phantoms was so encouraging, we proceeded to the human study. We are currently improving the power of the software so that it can be used with the more advanced multidetector CT scanners.

In conclusion, we developed new software using curved MPR to analyze airway dimensions exactly perpendicular to the long axis of airways at any site in the lungs. Using this software, we have for the first time demonstrated in vivo in patients with COPD that FEV₁ (% predicted) was highly correlated with airway luminal area and, to a lesser extent, with wall thickening from the third to the sixth generation of both B1 and B8, and that the correlation coefficients improved as airway size decreased. In the near future, with further advances in CT technology and in software, this type of approach will become even more attractive. Using this readily accessible and relatively noninvasive technique, we plan to conduct a longitudinal study of the subjects recruited for the Hokkaido COPD cohort study.