The addition of autofluorescence bronchoscopy (AFB) to standard white light bronchoscopy improves the capability of localizing small precancerous lesions. Currently visual identification alone classifies bronchoscopic findings. We analyzed light intensity from AFB images to determine whether abnormal pathology produced a quantitative alteration in autofluorescence.
Abnormal AFB lesions were imaged and biopsied. The lesions typically appeared brown or brownish-red when compared to normal areas which appeared green. All biopsies were placed in a 10% formalin solution and sent for pathology interpretation. The images were converted from a proprietary Xillix® file format to Tag Image File Format (TIFF). Using Universal Imaging’s MetaMorph® software, the biopsied site and one normal region of interest (ROI) were defined on each image by two pulmonologists experienced in AFB who were blinded to the pathology of the lesions. The average pixel intensity for red light and green light for each region was determined and red to green light ratios (RGR) were calculated.
A training set of ROIs was initially defined for a set of seven images in which the pathology was known to be “positive” (preinvasive or invasive cancers). The range of RGR of the abnormal ROIs (0.41 to 0.79) was found to be distinct from the range of RGR of the normal ROIs (0.23 to 0.38). This range of abnormal ROIs was then used in a blinded evaluation of 30 test images. Of the 17 positive images, five had positive pathology (positive predictive value 0.29). Of the 13 images with a negative test (RGR less than 0.41), twelve had negative pathology (negative predictive value 0.92) for an overall accuracy of 57% as compared to 43% with visualization alone.CONCLUSIONS: Quantification of AFB images supplements the bronchoscopist’s interpretation of precancerous lesions of the tracheobronchial tree.
Quantification of red to green light ratios enhances the ability to identify neoplastic and pre-neoplastic lesions during fluorescence bronchoscopy.
D.A. Orlando, None.