| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Department of Radiology, Hôpital Michallon,
Centre Hospitalier Universitaire, Grenoble BP 217, F-38043 Grenoble Cedex 09,
France.
2
Department of Statistics, Hôpital Michallon,
Centre Hospitalier Universitaire, Grenoble, Cedex 09, France.
3
Department of Respiratory Disease, Hôpital
Michallon, Centre Hospitalier Universitaire, Grenoble Cedex 09, France.
Received July 21, 1999;
accepted after revision October 19, 1999.
Address correspondence to G. R. Feretti.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. We used bronchoscopy and helical CT to examine 238 bronchial sections in 28 patients with 32 bronchial abnormalities and in five patients with normal bronchoscopy results. Postprocessing consisted of CT bronchography based on surface rendering. Images were interpreted independently by two observers (a radiologist and a pneumonologist) who were not informed of the bronchoscopy results. After initial interpretation of axial CT scans, the observers analyzed the axial CT scans with CT bronchograms. Results were evaluated for gain in diagnostic accuracy and in confidence.
RESULTS. Mean sensitivity for detection of abnormal bronchial sections was 89% (range, 87-90%) for axial CT and 92% (range, 90-94%) for axial CT with CT bronchography (not significant). Mean specificity of both approaches exceeded 99%. A correct diagnosis of the nature of the bronchial abnormalities was proposed for 68% of the cases in which axial CT was used alone and in 76% in which both axial CT and CT bronchography were used (not significant). The addition of CT bronchography significantly increased the confidence of the pneumonologist in the diagnoses.
CONCLUSION. Axial CT remains the technique of choice to detect and characterize benign abnormalities of the airways. CT bronchography provides little diagnostic gain but increases the confidence of chest physicians in the interpretation of CT scans for the assessment of benign abnormalities and typical carcinoids of the central airways.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Conventional CT is highly accurate (sensitivity, 90-94%) for depicting focal abnormalities in the central airways [3,4,5,6]. In excluding diseases of the central bronchial tree, the specificity of CT is 92-99% [3, 4]. However, conventional CT is imprecise in characterizing bronchial abnormalities depicted on bronchoscopy [4].
Helical CT allows volume acquisition of the airway during a single breath-hold, thereby eliminating respiratory motion [7]. Overlapping thin slices can be reconstructed from the raw data set, and high-quality image reconstructions can be created [8, 9]. With recent advances in computer techniques, CT bronchography (or virtual bronchoscopy) can be derived from helical CT data [10]. CT bronchography gives a bronchoscopist's view of the inner surface of major bronchi and allows the observer to "fly" through the airways [10]. Visualization of CT data in this way may improve the diagnostic value of CT [10], assist in planning for bronchoscopy [11], and provide a noninvasive method of follow-up for patients with certain bronchial abnormalities [12]. However, CT bronchographic postprocessing of image data is time-consuming and requires expensive workstations. Many reports in the literature have emphasized the similarity between real and virtual images; however, few series have assessed the diagnostic gain when CT bronchography is used with axial CT compared with axial CT alone [13].
The purpose of this study was to evaluate the gain attributable to the use of CT bronchography as a complementary imaging technique for the detection and the diagnosis of benign abnormalities of the central airways. We also evaluated the impact of the introduction of CT bronchography on an experienced radiologist and an experienced pneumonologist.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Fiberoptic bronchoscopy was performed by an experienced pneumonologist using a BF P40 fibroscope (Olympus Optical, Tokyo, Japan). Local anesthesia was used for the procedure. Bronchoscopic descriptions of the abnormalities of the airways were always available for review and were used as standards of reference (Table 1). Lesions were classified endoscopically. Six were endobronchial; of these, four were related to typical bronchial carcinoid, one to mucus plugging, and one to foreign body granuloma. Seven were inflammatory; of these, six were stenoses related to granulomatous reaction after prolonged tracheal intubation and one resulted from bronchial tuberculosis. Three stenoses resulted from extrinsic compression of the trachea by a thyroid goiter. Nine anatomic variations were present: six instances of right tracheal bronchus (four were apical segmental bronchi that arose from the right main bronchus and two were apical segmental bronchi that arose from the inferior trachea), one case of levoisomerism, and two cases of tracheal diverticulum. Other diagnoses included sinus track between a cavitated pulmonary mass and the tracheobronchial tree (n = 2), tracheal necrosis after ingestion of acid (n = 1) or intubation (n = 1), rupture of the trachea associated with fracture of the cervical spine (n = 1), tracheal ulceration (n = 1), and pneumonectomy (n = 1). The time interval between CT and bronchoscopy was no more than 15 days and was less than 24 hr in acute situations (i.e., in patients presenting with hemoptysis, inhalation, or thoracic trauma). In 20 patients, CT was performed first.
|
The final diagnosis was established by pathologic analysis of material obtained from surgery in 12 patients and by evaluation of either clinical history or bronchoscopy results in 20 patients.
Helical CT Technique
All helical CT scans were obtained with a Hi-Speed Advantage CT scanner
(General Electric Medical Systems, Milwaukee, WI) at 120 kVp and 200-250 mA
with 3-mm collimation and 1:1-1.5:1 pitch. Length of the scanned volume was
60-120 mm and acquisition time was 20-40 sec so that the acquisition could be
completed during a single breath-hold and inspiration after hyperventilation
for 1 min. In 12 patients, helical CT with 3-mm sections was the only
acquisition that was performed; this fact explains the height of the scanned
volume (120 mm). In 21 patients, helical CT examinations consisted of a focal
acquisition through the region of anatomic interest, which was selected on a
conventional CT scan of the entire thorax obtained during the same session. IV
contrast material was administered to 14 patients.
Overlapping transverse CT scans were retrospectively reconstructed at 1.5-mm intervals. A 180° linear interpolation algorithm was used. Axial CT images were photographed at standard lung windows (level, -400 H; width, 1600 H) and mediastinal windows (level, 30 H; width, 350 H).
CT Bronchography Reconstruction
Axial CT images were transferred to an Advantage Windows workstation
(General Electric Medical Systems) consisting of a Sparc-20 computer (Sun
Microsystems, Mountain View, CA) with 128-MB of main memory and 260-MB of swap
space. Endobronchial views were produced with the use of Navigator 1.3
(General Electric Medical Systems). Rendered voxels were selected on the basis
of the ray-casting algorithm by tracing rays from the point figuring the tip
of the endoscope. Image segmentation was based on thresholding. All voxels
above the threshold level were considered to be within the bronchial lumen. On
the basis of previous work by Zeiberg et al.
[14], we used an upper
threshold of -400 to -550 H to reconstruct the central airways. Image display
used a surface-rendering algorithm and produced perspective images with 512
x 512 pixels on a 16-bit gray scale. Each bronchoscopic image simulated
a coned-down view with a cone angle adjusted to 45°.
A radiologist produced a series of endobronchial views at 1-cm intervals at different levels of the central airways: trachea, right main bronchus, right upper lobe bronchus, intermediate bronchus, middle lobe bronchus, right lower lobe bronchus, left main bronchus, left upper lobe bronchus, and left lower lobe bronchus. Postprocessing time ranged from 20 to 40 min. CT bronchography was recorded on hard-copy films that were available to the observers for interpretation.
Image Analysis
The images were analyzed separately and independently in random order by an
experienced pneumonologist and fibroscopist (observer 1) who had little
training in interpreting CT bronchography and did not perform the
bronchoscopies, and by an experienced radiologist (observer 2) who has been
interpreting CT bronchography since 1995. Both observers were not informed of
the findings on bronchoscopy or the clinical history of the patient; they were
informed that they were looking for benign abnormalities of the airway or
typical carcinoid tumors. The observers analyzed two sets of images. In the
first session, they reviewed the axial CT scans. After 1 month, the observers
interpreted the axial CT scans and the CT bronchograms together, again in
random order.
For purposes of analysis, the central airway of each patient was divided into nine separate sections as follows: trachea, right main bronchus, right upper lobe bronchus, intermediate bronchus, middle lobe bronchus, right lower lobe bronchus, left main bronchus, left upper lobe bronchus, and left lower lobe bronchus. The observers were asked to rate the findings of each bronchial section as normal or abnormal. If the findings were abnormal, the observers were asked to propose the most likely diagnosis. For each abnormal section, the observers rated the CT scans alone and the CT scans with the CT bronchograms for the visibility of the abnormality and their level of confidence in the diagnosis according to a five-point scale (1 = poor; 2 = moderate; 3 = uncertain; 4 = good; and 5 = excellent).
After interpreting the axial CT scans with the CT bronchograms, the observers answered a questionnaire about the added value of CT bronchography in that particular case. For each answer, the observers used a four-point scale (1 = no; 2 = minimal; 3 = considerable; and 4 = enormous).
Statistical Analysis
For each bronchial section, imaging findings were compared with the results
of bronchoscopy on the basis of the presence or absence of an abnormality.
Results were classified as true-positive, true-negative, false-positive, and
false-negative. A true-positive was defined as an abnormal section noted both
on fiberoptic bronchoscopy and on CT. A true-negative was a normal section
noted on both types of imaging. A false-positive was a bronchial section noted
as abnormal on CT but normal on fiberoptic bronchoscopy. A false-negative was
a section noted as abnormal on fiberoptic bronchoscopy but considered normal
on CT. Sensitivity and specificity were calculated for each observer for each
set of images.
For each abnormal site, we evaluated the performance of the observers in terms of specific diagnosis. The sensitivity of each observer was then calculated relative to each set of images.
Results for each observer for CT alone and for CT with CT bronchography were compared by means of a chi-square test. Paired mean scores were compared with analysis of variance for repeated measures. A p value of less than 0.05 was considered significant. The 95% confidence limits were calculated from the binomial distribution.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Results of CT
A total of 284 bronchial sections were identified with axial CT and CT
bronchography, compared with 238 with bronchoscopy. Forty-six bronchial
sections visualized on CT were not visualized on bronchoscopy because they
were located distally to severe tracheal (n = 4) or bronchial
(n = 9) stenosis or obstruction that did not permit distal
progression of the bronchoscope. All bronchial sections visualized with
bronchoscopy were also visualized with CT.
Detection of Abnormal Bronchial Sections
On axial CT scans, observer 1 correctly identified 27 (87%) of 31 abnormal
sections and 237 (99%) of 238 normal sections
(Table 2). Observer 2 correctly
identified 28 (90%) of 31 abnormal sections and 236 (99%) of 238 normal
sections. Of the four abnormal bronchial sections missed by observer 1, three
were right tracheal bronchi (Fig.
1A,1B,1C,1D,1E,1F)
and one was a levoisomerism. Observer 2 missed three right tracheal bronchi.
Observer 1 had one false-positive lesion identification, and observer 2 had
two false-positive lesion identifications. In all three cases, the observers
described bronchial stenoses of less than 30%.
|
|
|
|
|
|
|
With the combined analysis of axial CT scans and CT bronchography, 28 (90%) and 29 (94%) of 31 abnormal sections were correctly identified by observer 1 and observer 2, respectively. With the addition of CT bronchography to axial CT, observer 1 failed to identify the same abnormal bronchial sections that he identified with axial CT, except for one right tracheal bronchus that was correctly recognized. With the addition of CT bronchography, observer 2 identified one of the three tracheal bronchi that he failed to identify with axial CT. With the addition of CT bronchography, both observers had no false-positive interpretations
The average sensitivity was 89% (95% confidence interval [CI], 81-97%) for axial CT scans and 92% (95% CI, 85-99%) for axial CT scans with CT bronchograms. This difference was not statistically significant (p > 0.5). Average specificity with both techniques was similar (>99%).
Characterization of Bronchial Abnormalities
Bronchoscopy revealed 32 specific diagnoses in 28 patients. Observer 1
proposed a correct diagnosis in 21 cases (65%) using axial CT scans and in 23
cases (72%) using axial CT and CT bronchography. Observer 2 proposed a correct
diagnosis in 24 (75%) and 27 (84%) cases, respectively. When using axial CT,
the two observers correctly diagnosed 45 (70%; 95% CI, 59-81%) of 64 lesions;
they gave a correct diagnosis in 50 (78%; 95% CI, 68-88%) cases when using
axial CT and CT bronchography. The gain in diagnostic accuracy was not
statistically significant.
With axial CT, observer 1 misinterpreted 11 abnormalities. Of these, seven were not recognized at all. These included four right tracheal bronchi (Fig. 1A,1B,1C,1D,1E,1F), one levoisomerism, one necrosis of the trachea (Fig. 2A,2B,2C,2D,2E), and one inflammatory stenosis of the trachea. In four cases, the diagnosis was incorrect, although the observer recognized an abnormal bronchial section. A severe stenosis of the left main bronchus related to tuberculosis and associated with an atelectasis of the left lung was misinterpreted as a parenchymal necrosis with a sinus track into the left main bronchus, a carcinoid was misinterpreted as a foreign body granuloma (Fig. 3A,3B), a necrosis of lung parenchyma with a sinus track into the trachea was misinterpreted as tracheal necrosis, and necrosis of the trachea was misinterpreted as a tracheal foreign body.
|
|
|
|
|
|
|
With axial CT and CT bronchography, observer 1 misinterpreted nine lesions: one necrosis of lung parenchyma with a sinus track into the trachea (tracheal stenosis), one foreign body granuloma with severe stenosis of the left main bronchus (endobronchial tumor), three right tracheal bronchi (not seen), two necroses of the trachea (not recognized), one tracheal ulceration (not seen), and one tracheal diverticulum (tracheal rupture).
On axial CT, observer 2 misinterpreted one foreign body granuloma with a severe stenosis of the left main bronchus (bronchial stenosis), one levoisomerism (not seen), three right tracheal bronchi (not seen), two tracheal stenoses (not seen), and one tracheal necrosis (not seen).
On axial CT and CT bronchography, observer 2 misinterpreted five lesions: one mucus plugging (bronchial stenosis), one foreign body granuloma with severe stenosis of the left main bronchus (bronchial stenosis), two right tracheal bronchi (not seen), and one tracheal stenosis (tracheal fistula).
Value of CT Bronchography
Addition of CT bronchography to axial CT increased neither the rating of
the radiologist (observer 2) concerning the visibility of bronchial
abnormalities nor the confidence in diagnoses but significantly increased the
ratings of the pneumonologist (observer 1) (p < 0.001). Confidence
in the diagnoses was significantly higher in cases for which the observers
made the correct diagnosis. For axial CT, the confidence of observer 1 in the
diagnoses averaged 3.0 in cases that were diagnosed incorrectly and 3.39 in
cases that were diagnosed correctly; observer 2 gave an average score of 4.0
in cases that were diagnosed incorrectly and 4.35 in cases that were diagnosed
correctly. After the addition of CT bronchography, observer 1 gave confidence
scores averaging 3.29 in cases with an incorrect diagnosis and 4.54 in cases
with a correct diagnosis, whereas observer 2 gave scores averaging 4.0 and
4.62, respectively. The two observers found that CT bronchography modified
their perception of the abnormalities in 53% of the cases and were of little
or no interest in 46%. However, the opinions of the observers were
statistically significant; the radiologist found CT bronchography of little or
no interest in 68% of cases and of interest in only 32%, whereas the
pneumonologist found CT bronchography of little or no interest in 24% and of
interest in 76% of cases (p < 0.001). The pneumonologist was more
enthusiastic than the radiologist concerning the increase in his level of
confidence in interpreting CT with CT bronchography. CT bronchography
increased the pneumonologist's level of confidence in 77% of cases
(dramatically, 10%; considerably, 67%; a little, 21%; not at all, 0%) compared
with 21% (dramatically, 0%; considerably, 21%; a little, 16%; not at all, 62%)
(p < 0.001) for the radiologist. The pneumonologist would have
been satisfied with CT bronchography as a replacement for conventional
bronchoscopy in 18 cases (56%), compared with five cases (16%) for the
radiologist. Finally, CT bronchography was regarded differently by the two
observers: the pneumonologist found it of no interest in 0% of cases, of
little interest in 19%, of interest in 73%, and of great interest in 8% of
cases, compared with 32%, 35%, 32%, and 0% of cases, respectively, for the
radiologist.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Many reports have emphasized the similarity of real and virtual views of the central airways [10, 12, 15, 16], but assessments of the potential of CT bronchography for improving detection and characterization of bronchial abnormalities have been discussed in only a few reports [13]. McAdams et al. [13] compared the accuracy of CT bronchography with that of axial CT in assessing bronchial anastomotic complications in 17 patients who had undergone lung transplantation. These researchers concluded that CT bronchography was slightly more accurate than axial CT for diagnosis of anastomotic stenoses. It has been suggested that CT bronchography may replace conventional bronchoscopy in some instances [10]; therefore, we studied patients with benign abnormalities of the central airways because morphologic testing to look for or exclude bronchial diseases could be limited to CT alone in such patients. We also included four patients with typical carcinoid tumors because, although these tumors are malignant neoplasms, carcinoid tumors behave in an indolent manner and have an excellent prognosis after surgical resection [18]. Moreover, we wanted to assess the impact of CT bronchography on the diagnostic choices and level of confidence of two experienced physicians with different training (i.e., a pneumonologist and a radiologist).
Our study focused on the proximal airways, which can be explored with bronchoscopy, and did not take into account bronchial abnormalities that are eventually revealed on CT but are located distal to an airway stenosis or airway obstruction that precludes passage of the bronchoscope. However, axial CT and CT bronchography make possible the display of the tracheobronchial tree below a severe stenosis or an obstruction. In this study, axial CT and CT bronchography increased the number of displayed bronchial sections by 46 (284 versus 238 sections displayed at bronchoscopy, or 19%), but no abnormalities were found in any of the 46 sections on CT.
CT bronchography did not significantly increase the average sensitivity of observers for the detection of abnormal bronchial sections (92% with axial CT and CT bronchography versus 89% with axial CT alone). Both observers detected all major abnormalities of the trachea and bronchi on axial CT: of a total of seven missed abnormalities, six were tracheal bronchi and one was a levoisomerism. These misinterpretations may be the consequence of a lack of mental reformation of tubular structures segmented on multiple cross-sectional scans. As has often been reported in the literature [19], none of these abnormalities was associated with a disorder. The addition of CT bronchography permitted identification of two tracheal bronchi. In our study, the average specificity of axial CT performed to exclude abnormal bronchial sections was excellent (>99%) and was similar to the results of published studies [3, 4, 20].
In the present study, the addition of CT bronchography did not significantly affect diagnostic accuracy. The radiologist's diagnosis was correct for 24 (75%) of 32 abnormalities when reviewing axial CT scans and for 27 (84%) of the abnormalities when CT bronchography was added. Addition of CT bronchography did not significantly improve the performance of the pneumonologist, who made the correct diagnosis in 23 cases (72%) when CT bronchography was added, compared with 21 cases (65%) with axial CT alone. We expected greater improvement on the part of the pneumonologist because bronchial abnormalities were presented in a familiar format.
Some limitations of CT bronchography may explain this lack of improvement. First, CT bronchography does not show the mucosal surface, its color, and its texture. Second, all endobronchial abnormalities are displayed in the same color because of the thresholding technique; therefore, endobronchial secretions, foreign body granulomas, inflammatory stenoses, and benign tumors are displayed as endobronchial processes obstructing the lumen, and all have the same appearance. The combined use of CT bronchography and axial CT may resolve some diagnostic problems when axial CT indicates the low density of bronchial secretions within the lumen [12].
Of particular interest are the differences between the radiologist and the pneumonologist regarding the visibility of the abnormalities and confidence in their diagnoses. No statistically significant difference was found in the rating of the radiologist for these two questions. The radiologist found that abnormalities were well indicated on axial CT scans (mean score, 4.6) and that little improvement resulted from CT bronchography (mean score, 4.7, not significant). In the same way, the radiologist's confidence in his diagnoses increased only moderately with the addition of CT bronchography (mean scores, 4.5 versus 4.3, not significant). In contrast, the pneumonologist felt more confident when axial CT scans were presented with CT bronchograms because the mean score increased from 3.3 to 4.2 (p < 0.001). The pneumonologist found that abnormalities were better depicted with the combination of CT bronchography and axial CT than with axial CT alone. When CT bronchograms were interpreted with axial CT scans, the pneumonologist was even more confident when the diagnosis he proposed was correct (mean score, 4.35 versus. 3.39) than when his diagnosis was incorrect (mean score, 3.29 versus 3.0). These differences might be attributable to the differences between the training of the two physicians [21]. The radiologist is accustomed to interpreting axial images and mentally reformatting the three-dimensional anatomy from the exterior of a hollow viscus [9]. In contrast, the pneumonologist is not trained to interpret axial images but is familiar with images of the endobronchial anatomy in three-dimensional perspective. All the answers to the questionnaire confirmed that CT bronchography was of greater interest to the pneumonologist than to the radiologist. Although the pneumonologist felt less confident in his interpretation of CT scans, he would have replaced conventional bronchoscopy with CT bronchography in 55% of cases, compared with 16% for the radiologist. In our opinion, this information should encourage radiologists to produce CT bronchographic synthetic views of given bronchial abnormalities to facilitate communication with pneumonologists.
The technique of CT bronchography uses surface rendering. Segmentation based on thresholding takes advantage of the natural contrast between the airway and the surrounding tissues. Therefore, the level of thresholding is critical in displaying accurate simulations. An upper threshold between -500 and -300 H is adequate for accurate segmentation of the central airways [14]. The other method of producing CT bronchography (i.e., volume rendering) allows preservation of the entire CT data set. Thus, endoluminal and extraluminal anatomy can be reconstructed on the same image [11, 16], an advantage that may be useful for directing transbronchial needle aspiration [11]. Volume rendering requires more powerful and expensive computers and is slower than surface rendering.
One of the limitations of this study is the small number of cases. However, benign diseases of the main bronchi and trachea are less frequent than malignant tumors; this fact explains why we have included only 28 patients in 2 1/2 years. Another limitation is that we did not compare the value of CT bronchography with those of the multiplanar reconstructions for the diagnosis and evaluation of bronchial disease.
In conclusion, we found that little diagnostic gain was realized with the addition of CT bronchography to axial CT in the evaluation of benign bronchial abnormalities. Our results are similar to those of McAdams et al. [13]. Therefore, considering the time needed for CT bronchography, our results do not justify the use of this technique for detecting bronchial abnormalities. CT bronchography does not create new information, but it presents information in a format that may make anatomic abnormalities more obvious than they are on axial images. These simulations simplify communication between radiologists and referring pneumonologists, who may not have received training in mental reconstruction of complex three-dimensional anatomy from axial views.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  P. M. Boiselle and A. Ernst Recent Advances in Central Airway Imaging* Chest, May 1, 2002; 121(5): 1651 - 1660. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Hoppe, B. Walder, M. Sonnenschein, P. Vock, and H.-P. Dinkel Multidetector CT Virtual Bronchoscopy to Grade Tracheobronchial Stenosis Am. J. Roentgenol., May 1, 2002; 178(5): 1195 - 1200. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. M. Summers, N. R. Aggarwal, M. C. Sneller, M. J. Cowan, B. J. Wood, C. A. Langford, and J. H. Shelhamer CT Virtual Bronchoscopy of the Central Airways in Patients With Wegener's Granulomatosis Chest, January 1, 2002; 121(1): 242 - 250. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G.R. Ferretti, I. Bricault, and M. Coulomb Virtual tools for imaging of the thorax Eur. Respir. J., August 1, 2001; 18(2): 381 - 392. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
