Grading Airway Stenosis Down to the Segmental Level Using Virtual Bronchoscopy: Statistical Analysis
The sensitivity, specificity, accuracy, positive predictive value, and negative predictive value of virtual bronchoscopy were calculated from 2 X 2 contingency tables, with confidence intervals derived from binomial distribution. A Fisher exact test was used to test for significance in field entries smaller than five where the x2 test was not applicable.
Assessments by virtual bronchoscopy were defined as overestimations or underestimations according to whether the particular stenosis was assigned a lower or higher grade on review of the findings of flexible bronchoscopy. The Spearman rank order correlation (r) was calculated to measure the strength of correlation between the results of virtual bronchoscopy and flexible bronchoscopy. A p value of < 0.05 was considered to be significant.
Of the 20 examined patients, 17 had bronchial carcinoma (non-small cell lung cancer, 14 patients; small cell lung cancer, 3 patients) and had at least one tracheobronchial stenosis. Three patients did not have bronchial carcinoma and did not have airway stenosis. A total of 176 central airway regions and 302 segmental airway regions were evaluated. Four central airway regions and 51 segmental airway regions were excluded because complete luminal obstruction of the proximal airways made flexible bronchoscopic assessment impossible. In six patients, the right lower lobar medial basal segmental bronchus (B VII) was absent due to anatomic variation. In one patient, the right upper lobar apical segmental bronchus (B I) and right upper lobar posterior segmental bronchus (B II) had a common bronchial stem due to anatomic variation. The quality of CT data sets was good in all patients. None of the patients exhibited major breathing artifacts. Moderate cardiac motion artifacts were observed in all patients. They appeared as regular ring structures in the region of the tracheal bifurcation and lingula bronchi.