Development and Validation of an Artificial Intelligence-assisted Strategy Selection System for Colonoscopy Cleaning

Colorectal Adenoma Clinical Trial

Official title: Development and Validation of an Artificial Intelligence-assisted Strategy Selection System for Colonoscopy Cleaning

Status Recruiting

Clinical Trial Summary

Patients with poor inadequate bowel preparation need to undergo secondary colonoscopy. but the evaluation of intestinal cleanliness is judged by doctors subjectively. there are no objective and effective criteria to guide the evaluation. We use the deep learning technique to develop the EndoAngel with real-time intestinal cleanliness assessment. It can derive a decision curve for bowel cleanliness based on the relationship between the percentage of bowel segments with a Boston score of 1 and the adenoma detection rate. It can help doctors to identify patients who need a second colonoscopy, and provide a new way for artificial intelligence in improving the detection rate of colonoscopic adenomas.

Clinical Trial Description

Colorectal disease as a common human disease seriously affects the health of human life. With the aging of the population, the change of diet structure and the aggravation of environmental pollution, the incidence of colorectal diseases, such as colon cancer, Colon Polyp and inflammatory bowel disease, has gradually increased. Colonoscopy is the simplest and most widely available screening procedure for colorectal cancer(CRC) prevention and early detection. Colonoscopy can clearly observe the small changes in the terminal ileum and the colorectal, such as erosion, ulcers, bleeding, congestion, edema, polyps, early cancer, and so on. Colonoscopy can biopsy the lesion site for pathological examination, to histologically qualitative the characterization of mucosal lesions, such as inflammation, polyp nature, the degree of differentiation of cancer, and so on. It is helpful to understand the severity of the lesion and guide the formulation of the correct treatment plan or judgment of treatment effect. Colonoscopy can also be the minimally invasive endoscopic treatment of colorectal polyps, early cancer, bleeding, foreign bodies and other diseases. Because the quality of bowel preparation affects the colonoscopy's ability to detect adenomas and polyps, adequate bowel preparation is necessary to ensure optimal use of colonoscopy in CRC prevention. Almost all clinical guidelines recommend adequate bowel preparation before colonoscopy. However, up to one third of colonoscopies have been found to show inadequate bowel preparation, which is estimated to increase the cost of colonoscopies by 12% to 22%. And there are 20% of patients' bowel is not adequately prepared. When the patient's bowel preparation is inadequate, the difficulty of flushing may lead to missed detection of adenomas. so doctors need to accurately identify such patients and tell them to have a second colonoscopy after a full bowel cleanse. However, the evaluation of intestinal cleanliness is decided by doctors subjectively, and there is no objective and effective scoring standard to guide the patients to accept the second colonoscopy. Deep learning is an important breakthrough in the field of artificial intelligence in the past decade. It has great potential in extracting tiny features in image analysis and image classification. In 2017, the journal Nature published a paper showing that using artificial intelligence to diagnose skin diseases can reach the level of experts. Subsequently, in the field of digestive endoscopy, more and more studies began to apply artificial intelligence to assist doctors to find polyps and improve the detection rate of polyps and adenomas.Urban, G. team used artificial intelligence to identify polyps with 95% sensitivity. Misawa, M team used artificial intelligence to identify polyps with 90% sensitivity. The purpose of our research group is to develop the EndoAngel with real-time intestinal cleanliness assessment. It can derive a decision curve for bowel cleanliness based on the relationship between the percentage of bowel segments with a Boston score of 0-1 and the detection rate of adenomas. It can help endoscopists to identify patients who need a second colonoscopy, to avoid the missed adenomas and the high cost of cleaning time caused by the wrong decision-making. At the same time, artificial intelligence is in the preliminary research stage in the field of digestive endoscopy, our research results are expected to provide new ideas in improving the detection rate of colonoscopic adenomas. The study Process is: Subjects who met all inclusion criteria and did not meet all exclusion criteria were included in the study before colonoscopy. During the colonoscopy, the endoscopists need to remain in the same without withdrawal while flushing the bowel. The biopsied patients are followed up for one week. the non-biopsied patients are followed up at the end of their colonoscopy , and the results are sent to an independent data analysis team for review. We will collect the patients' video and exclude the clips of irrigation, biopsy, and observation of polyp. Then the EndoAngel evaluates the Boston Bowel Preparation Scale of the ascending colon, transverse colon and descending colon, and calculates the proportion of 1 Score. ;

Related Conditions & MeSH terms

• Adenoma
• Colorectal Adenoma

• NCT number: NCT04444908
• Study type: Observational
• Source: Renmin Hospital of Wuhan University
• Contact: Honggang Yu, Doctor
• Phone: 13871281899
• Email: yuhonggang@whu.edu.cn
• Status: Recruiting
• Start date: May 11, 2020
• Completion date: November 16, 2020