Multi-center Validation of a Deep Learning Based Bowel Preparation Evaluation System

Inadequate bowel preparation is insufficient for identification of polyps greater than 5 mm. However, bowel preparation assessment involved subjectivity and uncertainty. We constructed a deep learning based system to calculate the proportion of Boston Bowel Prep Scale (BBPS) score of 0-1 during withdrawal phase and performed a prospective observational study to validate the threshold of the adequate proportion.The multi-center study is aimed to verify the extrapolation and robustness of the scoring threshold based on artificial intelligence intestinal cleanliness evaluation system explored in the early stage, and propose a more accurate and quantifiable threshold for evaluating the eligibility of intestinal preparation.