Comparative Evaluation of Super High-Resolution CT Scan and Virtual Bronchoscopy for the Detection of Tracheobronchial Malignancies: Statistical Analysis
VB images were converted to 3-D endoscopic views using commercial software (Navigator; General Electric) on a workstation (Advantage Windows; General Electric). The VB was derived from the SHR-CT axial images of the thorax with no further radiation exposure necessary. The radiologist (RMS), who was blinded to the findings of the other modalities, first placed the viewpoint in the proximal trachea. Retrograde inspection of the subglottic area of the trachea was performed. Subsequent analysis consisted of sequential antegrade inspection of the trachea, mainstem bronchi, lobar bronchi, and segmental bronchi. Little additional time was needed to perform and interpret the images. For the purpose of this study, the interpretation of the VB reconstruction was kept to < 10 min for all cases. All detected abnormalities in the tracheobronchial tree were recorded in the database. All SHR-CT and VB scans and FBs were interpreted in a blinded manner. The presence or absence of obstructive lesions (ie, bronchial narrowing of > 50%), endolu-minal masses (ie, a mass protruding into the lumen with < 50% occlusion), or mucosal lesions (ie, hemorrhage, erythema, or tissue friability) were recorded. In addition, the anatomic locations of the detected lesions were noted. read
The results of the imaging modalities were compared directly with actual FB findings at the same anatomic sites. A true-positive (TP) result occurred when the finding of the imaging modality equaled that of FB, when FB visualized a lesion. A true-negative (TN) result occurred when the finding of the imaging modality equaled that of FB, when the FB finding was within normal limits. A false-negative (FN) result occurred when an imaging modality failed to detect a lesion that had been documented during FB. A false-positive (FP) result occurred when an imaging modality demonstrated an abnormality but FB had revealed the area to be normal. The overall sensitivity (ie, TP/[TP + FN]) and specificity (ie, TN/[TN + FP]) of each imaging modality was determined. A subanalysis was performed for obstructive lesions, endoluminal masses, and mucosal lesions.