Establishment And Preliminary Verification Of An Artificial Intelligence-assisted Application System For Intestinal Preparation

BackgroundAs a common malignant tumor of the digestive tract,the incidence of colorectal cancer（CRC）in China ranks second in the world in 2020.The key to the prevention and treatment of CRC lies in early detection and early treatment.Colonoscopy has become the gold standard for colorectal lesion screening,diagnosis and treatment because it can look directly at colorectal lesions and early intervention with minimally invasive therapy to prevent the formation and development of CRC.The quality of bowel preparation is one of the quality criteria for evaluating colonoscopy,and adequate bowel preparation is essential for the success of colonoscopy,contributing to good mucosal visualization and lesion detection.One of the risk factors for inadequate bowel preparation is overweight.In this project,starting from improving the quality of intestinal preparation,United Software Company developed an intestinal preparation APP for artificial intelligence-assisted diagnosis systems and used it to improve the intestinal cleanliness of overweight patients undergoing colonoscopy in clinical practice.ObjectiveEstablish an AI-assisted intestinal preparation quality assessment system and verify whether it can improve the quality of intestinal preparation in overweight patients.Methods1 Build a modelMulti-center enrolled 1500 participants,of which 80% are modeled as a modeling group and 20% are used for correction,based on convolutional neural network algorithms,to build a highly sensitive and highly specific intestinal preparation quality assessment artificial intelligence model.2 Verification Model122 overweight patients who were scheduled to undergo colonoscopy in our hospital from 2021.01 to 2021.05 were randomly divided into two groups,the control group used oral education + paper instructions for intestinal preparation education,and the experimental group used oral education + paper instructions,and then used the smart phone APP for intensive intestinal preparation education.The intestinal readiness of the 2 groups was evaluated according to the Boston Bowel Readiness Assessment Scale（BBPS）and the intestinal readiness adequacy rate of the 2 groups was compared.ResultsThe sensitivity of the model is calculated to be 0.81,the specificity is 0.97,the accuracy is 0.95,and the accuracy rate is 0.82.The intestinal preparation adequacy rate of the right half colon and the total colon was higher in the experimental group than in the control group（96.72% vs.86.89%,c2=3.92,P=0.048;96.72% vs.85.25%,c2=4.9,P=0.027）,and the difference was statistically significant（P < 0.05）.The intestinal adequacy rates of transverse colon and left colon in the experimental group were higher than those in the control group,but the difference was not statistically significant（P>0.05）.ConclusionThe application system of artificial intelligence assisted intestinal preparation can guide patients to intestinal preparation in a timely,simple,fast and personalized,and can significantly improve the intestinal preparation adequacy rate of overweight patients.