Abstract
5‐aminolevulinic acid (ALA)‐induced protoporphyrin IX (PPIX) fluorescence has shown an outstanding sensitivity for the assessment of oral lesions, but its application was hampered by low specificity due to the high false‐positive rates. The purpose of our study was to explore the feasibility of quantifying PPIX fluorescence images to improve the diagnostic specificity for detecting early oral lesions in vivo. A digitized 5‐ALA‐mediated endoscopic imaging system was utilized to acquire PPIX fluorescence images from in vivo oral tissues. Forty‐nine patients (118 biopsies) with known or suspected premalignant or malignant oral lesions were recruited for ALA‐PPIX fluorescence endoscopic imaging. The red and blue channels of PPIX fluorescence images were digitized and stored for fluorescence quantification. The red‐to‐blue intensity ratios were calculated from the fluorescence images to correlate with histologic findings of the biopsies. The results showed that normal oral mucosa exhibited blue color of the back‐scattered excitation light in the fluorescence images, whereas the suspicious lesions displayed bright reddish fluorescence. Applying the red‐to‐blue intensity ratio (I~R~/I~B~) as a diagnostic algorithm yielded a sensitivity of 92% and 98%, and specificity of 96% and 96%, for separating benign tissue from dysplasia, and cancer tissue, respectively, and a sensitivity and specificity of 98% and 92%, respectively, for differentiating cancer tissue from dysplasia in the oral cavity. Our study demonstrates that quantifying ALA‐PPIX fluorescence endoscopic images associated with the red‐to‐blue intensity ratio as a diagnostic algorithm can provide good differentiation between the different stages of oral premalignancy and malignancy (p<0.0001, unpaired 2‐sided Student's t‐test), and thus has a potential to significantly improve the noninvasive diagnosis and evaluation of early oral neoplasia in vivo. © 2004 Wiley‐Liss, Inc.