Colorectal adenoma is a common disease and frequently-occurring disease in gastroenterology. With the continuous progress of colonoscopy equipment and the gradual improvement of endoscopic accessories, especially the development of chromo-endoscopy and magnifying endoscopy. The observation of the surface structure and capillary morphology of colorectal adenomas can realize optical biopsy. Currently, most clinical endoscopic diagnosis of colorectal diseases is biopsy under colonoscopy, and further treatment options are determined based on the pathological results of the biopsy. The problem is that the pathological diagnosis of some preoperative biopsy is not completely consistent with the pathological diagnosis of postoperative large specimens. Previous studies have found that the pathological diagnosis accuracy rate of preoperative biopsy is only 66-75%, so there is a certain degree of subjectivity in relying solely on colonoscopy white light biopsy. Based on the previous work, the research team has initially established an intelligent recognition model for colorectal adenoma classification (low-grade intraepithelial neoplasia, high-grade intraepithelial neoplasia), and formed a colorectal adenoma of a certain size with annotated endoscopic image data set. Using the YOLO-V4 algorithm, under the Darknet framework, to train an artificial intelligence (AI) system which specifically for adenoma recognition and diagnosis, its accuracy rate has reached more than 90%. This study intends to increase the sample size based on the previous work, and further improve the accuracy of the classification and diagnosis of the AI system, so as to guide the endoscopist to perform targeted biopsy and improve the accuracy of preoperative biopsy.

Inclusion Criteria:

• Age between 30-75;
• Those who have no mental abnormality and can conduct questionnaire surveys;
• BBPS ≥ 6;
• Colorectal advanced adenoma, and admitted for complete resection with EMR and ESD;
• Provide the relevant information required by this study and sign the informed consent.

Exclusion Criteria:

• Those who cannot provide the relevant information required by this research;
• Patients with inflammatory bowel disease;
• Those with a history of liver cirrhosis, uncontrolled hypertension, history of myocardial infarction, cardiac insufficiency, renal insufficiency, respiratory failure, diabetic ketosis and electrolyte imbalance and other serious diseases;
• Those who cannot stop antiplatelet drugs or anticoagulant drugs;
• Those who have not completed full colonoscopy;
• Pregnant women.

This study has been reviewed by the hospital ethics committee. The enrolled subjects were found to have advanced colorectal adenomas during colonoscopy, and already had the pathological results of AI-assisted guided biopsy. The patients were hospitalized for complete resection with EMR and ESD and were willing to participate in this study.

After completing the pre-experiment.