DEEP-LEARNING-BASED METHOD FOR IMPROVING COLONOSCOPE ADENOMATOUS POLYP DETECTION RATE

摘要

Disclosed is a deep-learning-based method for improving a colonoscope adenomatous polyp detection rate. The method comprises the following steps: dividing a video stream transmitted from a camera of a colonoscope of an operating table into two parts, with one part of the video stream being transmitted to an operating platform of a doctor and the other part of the video stream being pre-processed and then being transmitted to a polyp detection model embedded in a colonoscope operating system for identification; the polyp detection model detecting whether there is a polyp in each frame of image and the occurrence probability of the polyp; returning a detection result from the polyp detection model to the operating platform of the doctor and displaying same; and if there is a polyp in the video stream, framing the polyp and giving a prompt. By virtue of an artificial intelligence deep neural network, a polyp occurring within the range of a camera of a colonoscope during an operation can be automatically detected, thereby improving the identification rate of a polyp during a colonoscopy and thus indirectly improving the detection rate of an adenomatous polyp.