Bronchoscopic fibered confocal fluorescence microscopy image characteristics and pathologic correlations

Background: Fibered confocal fluorescence microscopy (FCFM) is a new imaging modality in bronchoscopy. The purpose of this study was to assess FCFM reliability, interpretation, and to make image-pathologic correlations. Methods: Twenty-six patients underwent FCFM. A validation set was used to determine image characteristics and interobserver reliability. Each patient underwent bronchoscopy using a standardized protocol. The images were evaluated by 4 observers based on brightness, fiber thickness, and alveolar cellularity. Image characteristics showing good interobserver agreement were tested to see if they were related to smoking status. Subsequently, 18 consecutive patients underwent FCFM and biopsy to correlate images with pathology. The blinded reviewers were asked to distinguish between controls and patients with pathologically proven disease. Results: Interobserver agreement for image brightness, as measured by intraclass correlation coefficients (ICCs), ranged from 0.48 to 0.92 (P<0.001) and varied by location. ICCs for image brightness were high, ranging from 0.53 to 0.99 (P<0.001). Agreement for fiber thickness was poor for respiratory bronchioles (ICC 0.12, P<0.05) and fair for alveoli (ICC range, 0.37 to 0.42, P<0.001). The intraobserver (ICC range, 0.69 to 0.91, P<0.001) and intrapatient (ICC 0.65 to 0.84, P<0.001) reliability were excellent. Computer image interpretation showed excellent agreement with humans (ICC 0.62 to 0.99, P<0.001). Smoking was inversely associated with respiratory bronchiole brightness (P<0.001). In FCFM-pathologic correlation, FCFM could distinguish normal from diseased tissue; however, specific diseases could not be distinguished from other diseases. Conclusion: FCFM shows a high degree of image reliability and can detect changes in the respiratory bronchioles because of smoking and other diseases, but whether it can discriminate among diseases requires additional study.