The learning curve of in vivo probe-based confocal laser endomicroscopy for prediction of colorectal neoplasia.

BACKGROUND: Probe-based confocal laser endomicroscopy (pCLE) is an emerging tool for in vivo imaging of the GI tract that requires the endoscopist to interpret microscopic images. The learning curve for interpretation of pCLE images is unknown.
OBJECTIVE: To examine the learning curve of correctly identifying benign and neoplastic colorectal lesions by using pCLE and to evaluate the learning curve of obtaining high-quality images.
DESIGN: Prospective, double-blind review of pCLE images of 76 colorectal lesions by using corresponding polypectomies as the reference standard. A training set of 20 images with known histology was first reviewed to standardize image interpretation, followed by blinded review of 76 unknown images.
SETTING: Eleven endoscopists from 3 different endoscopy centers evaluated the images obtained by 1 endoscopist using the high-definition confocal probe. PATIENTS: Patients undergoing screening and surveillance colonoscopies. INTERVENTION: Intravenous fluorescein pCLE imaging of colorectal lesions followed by polypectomies. MAIN OUTCOME MEASUREMENTS: Accuracy of image interpretation with constructing learning curve for pCLE image interpretation and acquisition.
RESULTS: Of the 76 colorectal lesions, 51 (67%) were neoplastic and 25 (33%) were benign, based on histopathology. Accuracy for the overall group was 63% for lesions 1 to 20, 64% for lesions 21 to 40, 79% for lesions 41 to 60, and 86% for lesions 61 to 76. The ability to obtain high-quality images was stable over the 76 cases. LIMITATIONS: Small sample size and use of offline video sequences.
CONCLUSIONS: Accurate interpretation of pCLE images for predicting neoplastic lesions can be learned rapidly by a wide range of GI specialists. Furthermore, the ability to acquire high-quality pCLE images is also quickly learned.