Learning Curve of Virtual Chromoendoscopy for the Prediction of Hyperplastic and Adenomatous Colorectal Lesions: a Prospective 2-center Study

Background: Computed virtual chromoendoscopy (CVC) enables high-definition imaging of mucosal lesions with improved tissue contrast. Previous studies have shown that CVC yields an improved detection rate of colorectal lesions. However, the learning curve for interpretation of CVC images is unknown.

Objective: To examine the learning curve of correctly identifying hyperplastic and adenomatous colorectal lesions by using CVC.

Design: Prospective, 2-center study.

Patients: Consecutive patients undergoing screening colonoscopy were included. CVC images were analyzed by using corresponding polypectomies as the reference standard followed by a prospective, double-blind review of i-scan images.

Methods: A training set containing 20 images with known histology was reviewed to standardize image interpretation, followed by a blind review of 110 unknown images. Overall, 4 endoscopists from 2 different endoscopy centers evaluated the images, which were obtained by 1 endoscopist using high-definition endoscopy with CVC.

Results: Patients were included in a prospective fashion. Seventy-seven of 110 colorectal lesions were adenomas and 33 were hyperplastic lesions. Mean diameter of colonic polyps was 4 mm (range, 2-20 mm). Overall accuracy for the group was 73.9% for lesions 1 to 22, 79.6% for lesions 23 to 44, 84.1% for lesions 45 to 66, 87.5% for lesions 67 to 88, and 94.3% for lesions 89 to 110. Accuracy of i-scan for prediction of polyp histology was not dependent on polyp size (≤5 mm, 6-10 mm, or > 10 mm). The ability to obtain high-quality images was stable over time, and high-quality images were constantly produced.

Limitation: Post-hoc assessment.

Conclusion: Accurate interpretation of CVC images for prediction of hyperplastic and adenomatous colorectal lesions follows a learning curve but can be learned rapidly.