Automatic optical diagnosis of small colorectal lesions by laser-induced autofluorescence

Background and study aims: Endoscopic optical diagnosis can potentially replace histopathological evaluation of small colorectal lesions. The aim of this study was to evaluate diagnostic performance of WavSTAT, a novel system for automatic optical diagnosis based on laser-induced autofluorescence spectroscopy. Patients and methods: Consecutive patients who were scheduled for colonoscopy were included in the study. Each detected lesion with a size of <= 9mm was differentiated using high resolution endoscopy (HRE) by the endoscopist, who then reported this as a low or high confidence call. Thereafter, all lesions were analyzed using WavSTAT. Histopathology was used as the reference standard. The primary outcome measures were the accuracy of WavSTAT to differentiate between adenomatous and nonadenomatous lesions, and the accuracy of an algorithm combining HRE (lesions differentiated with high confidence) and WavSTAT (all remaining lesions). The secondary outcome measure was the accuracy of on-site recommended surveillance intervals. Results: At total of 87 patients with 207 small colorectal lesions were evaluated. Accuracy and negative predictive value of WavSTAT were 74.4% and 73.5 %, respectively. The corresponding figures for the algorithm were 79.2% and 73.9 %, respectively. Accuracy of on-site recommended surveillance interval was 73.7% for WavSTAT alone and 77.2% for the algorithm of HRE and WavSTAT. Conclusions: Both accuracy of WavSTAT alone and the algorithm combining HRE with WavSTAT proved to be insufficient for the in vivo differentiation of small colorectal lesions, and do not fulfill American Society for Gastrointestinal Endoscopy performance thresholds for assessment of diminutive lesions. Future studies should assess whether combining WavSTAT with more advanced imaging techniques could result in a higher accuracy.