|
CT virtual bronchoscopy (CTVB) is a new technique for performing simulated bronc h oscopy using data obtained from spiral CT scan of the airway, in which the viewp oint is placed within the airway lumen to produce an endoscopic display.［1- 3］Being a noninvasive technique, it is useful assess airway lesions. In this article, we discuss the choice of spiral CT scan protocol, reconstruction meth ods, and preliminary clinical application.
METHODS
Patients
In this study, 52 patients were 32 men and 20 women (age range 21-75 years; mean age 52 years). Of the 52 patients, 46 had lung cancer of central type, 4 posto perative lung cancer, 1 tracheal adenoidcysticcarcinoma, and 1 bronchial divert iculum. All patients underwent fiberoptic bronchoscopy, and the patients with l ung cancers were confirmed pathologically.
CT scanning
CTVB images were created using spiral CT data acquired with a GE HiSpeed CT/i CT scanner, with 3 mm or 5 mm collimation, 1∶1 or 1∶1.5 pitch, 120 kV, 300 mA, and a single breath hold acquisition. The scan ranged from 3.0 cm above t he carina to the level of the right inferior pulmonary vein. If the extent of t he disease was greater or the scan time exceeded 30 seconds, the examination was divided into two segments with overlapping one slice (3 mm or 5 mm), with 5-1 0 seconds for breathing between the two scans. In order to shorten scan time, s ub-second scan (0.8 s) was used sometime. If pitch was changed from 1∶1 to1∶1.5, the scan time could be reduced 1/3.
Images processing
The images were reconstructed in the axial plane at 1.0 mm or 2.0 mm interva ls (66% or 60% overlap) using a standard or bone reconstruction algorithm, FOV 3 0 cm. The CT data were transferred to a computer workstation (Advantage Window s 3.1, GE Medical Systems). Endobronchial views were produced with 3D soft ware system (Navigator Smooth). Navigator "wandering" through the tracheobronchia l tree was accomplished interactively by using the workstation monitor and compu ter mouse. The workstation screen was divided into four quadrants, containing a n endoluminal view, a transaxial view, a multiplanar reconstruction (MPR) corona l view, and an MPR sagittal view. Navigator was guided by using the MPR images as well as the endoluminal view or displayed as movies. Perspective angle was 3 0 0 -60 0 , and the angle size was dependent on the distance of cursor to th e obs erved point. The CTVB images had the ability to rotate in all directions within the 3D space. During cursor movement in the tracheobronchial endolumen, its l ocation was identified in each quadrant in real-time. The cursor's pos ition on the endoscopic view was correlated with its 3D coordinate on the MPR vi ews in real-time as the cursor moved. VB associated with MPR could determine the exact position of a referenced anatomical structure such as the carina and b ronchial orifices in addition to the pathologic processes of endoluminal masses and extra-luminal invasion. To demonstrate the length of tracheobronchial sten osis and external compression, we used Min IP and SSD, which were better for de picting tracheobronchial trees and their 3D images.
Statistical analysis
CT virtual bronchoscopy was compared with fiberoptic bronchoscopy for identif ying the tracheobronchial masses; the sensitivity and accuracy were calculated. Statistical analysis was made by the x 2 test; P<0.05 was considered a s statistical significant difference.
RESULTS
Colored CTVB images showed vividly the tracheobronchial lumen, cartilage rings, carina, and bronchial orifices, which mimic fiberoptic endoscope display. 100% segmental bronchus and more than 80% subsegmental bronchus were visualized. The visualization rate was related to the size of FOV and operator's skill, and re ducing the size of field of view (FOV) increased the vasualization rate of subse gmental b ronchi. In 46 lung cancers of central type confirmed pathologically, fiberoptic bronc hoscopy showed the masses in 45 cases and CT virtual bronchoscopy showed the mas ses in 42, with a sensitivity of 93.3% and an accuracy of 93.5%. The dif fere nce in sensitivity was not statistically significant (x 2 =1.33, 0.10<P<0 .25). The tumors characterized by masses or nodules ( Fig. 1 ) caused bronchi al stenosis (n=35) or occlusion (n=7) ( Figs. 2 and 3 ). Fiberoptic endoscopy sh o wed the bronchial rings near the masses were blurred, smooth, or absent ( Fig. 4 ), whereas CTVB was unable to demonstrate necrosis and hemorrhage in two tumor s. In our 3 patients with lung cancers confirmed by fiberoptic endoscopy but no t CT VB, the tumors located mainly at the extralumens and resulted in slight thickeni ng of the bronchial wall. CTVB could pass through the stenotic trachea or bro nchi and reveal occlusive bronchi from the distal end if fiberoptic endoscopy failed. Combined SSD and MPR image identified accurately the size, location an d extent of the masses ( Figs. 5 and 6 ). Residual bronchial stump after operation (n=4) showed smooth blind ends without recurrence of tumors confirmed by fiberoptic endoscopy ( Fig. 7 ). Diverticulum in the right bronchial mid dle segment (n=1) exhibited a local concavity on CTVB and local protrusion on SS D ( Fig. 8 ). The tracheal sagittal dimension enlarged with a "scabbard" ap pearance in one patient, and the right wall of the trachea protruded into the lu men in another. CTVB failed to identify endoluminal tumor or extrinsic impressi on, but SSD confirmed the condition from extraluminal view.
DISCUSSION
CT scan protocol and reconstruction
Virtual bronchoscopy image may be accomplished in three steps: selection of scan parameters; preprocessing of source images; and reconstruction of VB image. ［1-3］
To compare the parameters relevant to image quality, we used a variety of sectio n thickness (10 mm, 5 mm, 3 mm, and 1 mm), pitch (1∶1 and 1∶1.5) and reconstruction interval (1 mm, 2 mm, 1 mm, and 0.1 mm; with 90%, 66%, 60%, and 90% overlap, respectively) for scanning and reconstruction of images (120 k V, 300 mA, standard interpolator and bone algorithm). The most optimal par ameters were section thickness 3 mm, pitch 1∶1, scanning duration <30 seconds, and >60% overlap of reconstruction sections. Standard algorithm was better than bone algorithm. The protocol was basically consistent with that of Hopper et a l.［4］Subsecond scan was used with thinner sections to cover the same scanning distance and shorten breathhold time. However, good images were obtained by us ing 5 mm section thickness, 1∶1.5 pitch, and 2 mm interval sections (ove rlap 60%). We also used 1 mm section thickness, pitch 1∶1, 66% overlap, but t he images were not good and jaggy appearance was conspicuous, especially in those patients who could not hold their breath for scanning.
The threshold setting on CTVB image affected the size of endoluminal lesions for overestimation or underestimation. When the threshold was negative, the size of the lesions increased apparently. Geometric distortion was seen in the super oinferior direction.［5］To choose threshold, we adjusted window width to zero on axial image for a optimal appearance of the lesions. The CT value for this wi ndow level was about -500 Hu. Moreover, to choose a median value of air and le sions was another method. Its value was also about -500 Hu.［2］Observing the postoperative bronchial stump or occlusive bronchi was necessary to increase th reshold up to the appearance of lesions.
Clinical application of CTVB
With the rapid development of computer hardware and software, CTVB has provided fast images similar to those shown by fiberoptic bronchoscopy.［6］CTVB can cl early demonstrate the trachea, carina, left and right major bronchi, lobar and s egmental bronchi, down to the fourth order of the bronchial orifices within 3-5 min. CTVB is noninvasive and is capable of passing through bronchial stenosi s or observing occlusive bronchi from the distal to proximal end of the bronchi. Simultaneous VB and MPR images are useful to evaluate of the thickness of the tr acheobronchial wall and the length of the stricture. It is also used to detect the vascular structures of the lesion (after contrast enhancement) as well as tumor extension. VB associated with SSD and MPR can precisely demonstrate the anatomic relationship between endoluminal and extraluminal lesions. Our study showed that CTVB is highly sensitive to bronchial masses, but it fails to observe mucosal abnormalities and to obtain histologic samples.
REFERENCES
1. Summers RM, Feng DH, Holland SM, et al. Virtual bronchoscopy: segmentation method for real-time display. Radiology 1996;200:857-862.
2. Summers RM, Shaw DJ, Shelhamer JH. CT virtual bronchoscopy of simulated endobronchial lesions: effect of scanning, reconstruction and display settings and potential pitfalls. AJR Am J Roentgenol 1998；170:947-950.
3. Higgins WE, Ramaswamy K, Swift RD, et al. Virtual bronchoscopy for three-d imensional pulmonary image assessment: state of the art and future needs. Radi ographics 1998；18：761-778.
4. Hopper KD, Lyriboz TA, Mahraj RPM, et al. CT bronchoscopy optimization of i maging parameters. Radiology 1998；209:872-877.
5. Ferretti GR, Vining DJ, Knoplioch J, et al. Tracheobronchial tree: three-d imensional spiral CT with bronchoscopic perspective. J Comput Assist Tomogr 1996；20:777-781.
6. Vining DJ, Liu K, Choplin RH, et al. Virtual bronchoscopy: relationship s of virtual reality endobronchial simulations to actual bronchoscopic findings . Chest 1996；109:549-553.
|
PDF(-)
Rapid Response |