Abstract
OBJECTIVE: The aim of this study is to evaluate the performance of a confocal fluorescence microlaparoscope for in vivo detection of ovarian cancer. METHODS/MATERIALS: Seventy-one patients scheduled for open or laparoscopic oophorectomy were consented for the imaging study. High-resolution confocal microlaparoscopic images of the epithelial surface of the ovary were acquired in vivo or ex vivo after tissue staining using acridine orange. Standard histologic evaluation of extracted tissue samples was performed and used as the gold standard of disease diagnosis. Trained human observers from different specialties viewed the microlaparoscopic images, rating each image on a 6-point scale ranging from "definitely not cancer" to "definitely cancer." Receiver operating characteristic curves were generated using these scores and the gold standard histopathologic diagnosis. Area under the receiver operating characteristic curve (AUC) was calculated as a performance metric. RESULTS: Forty-five of the consented patients were used in the final evaluation study. From these 45 patients, 63 tissue locations or samples were identified and imaged with the confocal microlaparoscope. Twenty of the samples were high-grade cancers, and the remaining 43 samples were normal or noncancerous. Twenty-three of the samples were imaged in vivo, and the remaining 40 samples were imaged ex vivo. The average AUC score and standard error (SE) for detection of cancer in all images were 0.88 and 0.02, respectively. An independent-samples t test was conducted to compare AUC scores for in vivo and ex vivo conditions. No statistically significant difference in the AUC score for in vivo (AUC, 0.850; SE, 0.049) and ex vivo (AUC, 0.888; SE, 0.027) conditions was observed, t(6) = 1.318, P = 0.2355. CONCLUSIONS: Area under the receiver operating characteristic curve scores indicate that high-resolution in vivo images obtained by the confocal laparoscope can distinguish between normal and malignant ovarian surface epithelium. In addition, in vivo performance is similar to that which can be obtained from ex vivo tissue.