Light-Induced Auto uorescence Spectroscopy for Detection of Nasopharyngeal Carcinoma in Vivo

Auto uorescence spectral signals were measured in vivo from 85 nasopharyngeal carcinoma lesions and 131 normal tissue sites of 59 subjects during routine nasal endoscopy. Diagnostic algorithms based on principal component analysis and the ratio of the spectral signals between multiple-wavelength bands were developed for classifying the auto uorescence spectra. Performances of the algorithms were evaluated using the cross-validation method. The principal component analysis based algorithms using information from the entire  uorescence spectrum can differentiate nasopharyngeal carcinoma lesions from normal tissue with 95% sensitivity and 93% speciŽ city. With 94% sensitivity, the speciŽ cities of multiple-wavelength ratio algorithms are about 83%. The results demonstrate that light-induced auto uorescence endoscopy with principal component analysis algorithms can provide accurate diagnostic information for the detection of nasopharyngeal carcinoma in vivo and may be potentially used in clinical practice combined with the routine white-light endoscopy procedure.