Grading Airway Stenosis Down to the Segmental Level Using Virtual Bronchoscopy: CT Scanning

Patients were only included if they had undergone both CT scanning and flexible bronchoscopy. The mean interval between virtual bronchoscopy and flexible bronchoscopy was 5.4 days (range, 0 to 17 days). Eventually, 17 of the 20 patients had lung cancer as the final diagnosis. Histologic diagnoses included non-small cell lung cancer (14 diagnoses) and small cell lung cancer (3 diagnoses). Because both procedures were performed in the context of patient care and were evaluated in retrospect, no institutional review board approval or patient informed consent had to be obtained, according to the guidelines of our institution.

CT scan examinations were performed using a multirow detector CT scanner (Asteion; Toshiba; Tokyo, Japan) with 4 X 1 mm collimation, pitch 1.375, 120 keV, 0.75-s rotation time, and 100 to 180 mA. Acquisition time was 25 to 35 s to enable complete acquisition during a single breathhold. The chest was scanned during inspiration in a caudocranial direction after a power injection of 80 mL (flow rate, 2 mL/s; scan delay, 30 s) of iopromide IV contrast medium containing 300 mg/mL iodine (Ultravist 300; Berlex Laboratories; Montville, NJ). The reconstruction interval and slice thickness were each 1 mm.
Image Processing
Axial CT scan images were transferred to a workstation (Advantage for Windows 4.0; General Electric Medical Systems; Milwaukee, WI) running on appropriate hardware (Ultra Sparc 60; Sun Microsystems; Mountain View, CA) featuring two 450-Mhz central processing units (Sun Ultra Sparc II) and 2 gigabytes of random access memory. Appropriate software (Navigator, version 2.03; General Electric Medical Systems) was used for the reconstruction of virtual bronchoscopic images. Image display used a surface-rendering algorithm and produced perspective grayscale images with a matrix of 512 X 512. Image segmentation was based on thresholding. All voxels with a density below the threshold level were considered to be within the bronchial lumen. An upper threshold between —400 and —550 Hounsfield units (HU) was used for endoluminal rendering of the central airways. For virtual bronchoscopic reconstruction of the peripheral airways, an upper threshold between — 500 and — 800 HU was used, depending on the caliber of the individual airways. To avoid the overestimation of airway stenosis, the upper threshold value was approached until the tracheobronchial wall was completely rendered without defects. Each bronchoscopic image simulated a coned-down view with an adjusted cone angle of between 50° and 80°, depending on the bronchial diameter and location within the tracheobronchial tree.