AI-based algorithm for clinical decision support system in colonoscopy

AIM: to estimate the implementation of original method that uses artificial intelligence (AI) detect colorectal neoplasms. MATERIALS AND METHODS: we selected 1070 colonoscopy videos from our archive with 5 types lesions: hyperplastic polyp, serrated adenoma, adenoma low-grade dysplasia, high-grade dysplasia and invasive cancer. Then 9838 informative frames were selected, including 6543 Lesions annotated obtain data set was finally used for training a convolution al neural network (YOLOv5). RESULTS: trained algorithm is able neoplasms an accuracy 83.2% sensitivity 77.2% on test sample dataset. The most common errors revealed analyzed. CONCLUSION: obtained provided AI-based can in video stream recording. Further development technology probably will provide creation clinical decision support system colonoscopy.