[VINIF2020] Development of a Real-time AI-assisted System to Detect Colon Polyps and Identify Lesions at High Risk of Malignancy During Endoscopy

Objectives

The purpose of the project is to conduct research and development of a real-time AI-assisted system to detect colon polyps and identify lesions at high-risk of malignancy during endoscopy. We will develop a platform for colonoscopy data creation and create big datasets for Vietnam-specific problems of colonoscopy with rich semantic annotation. We then propose novel algorithms that achieve state-of-the-art results in detection of colon polyps and in classifying high-risk lesions for colon polyps in endoscopic images. The outperformance of our system will be achieved based on the new algorithms trained on the newly collected large scale endoscopic datasets. The algorithms will be embedded on a dedicated device to build a real-time endoscopic image diagnosis system, as an AI “doctor”, to assist endoscopists during colonoscopy. The system can be applied in clinical practice to increase colon polyp detection rate, especially for adenomas and high-risk lesions, as well as to serve the increasing number of patients while maintaining healthcare quality. The AI-assisted system can be utilized for training endoscopists in Vietnam and some other developing countries with limited resources.