Improving Colonoscopy Polyp Detection Rate Using Semi-Supervised Learning

Colorectal cancer is one of the biggest health threats to humans and takes thousands of lives every year. Colonoscopy is the gold standard in clinical practice to inspect the intestinal wall, detect polyps and remove polyps in early stages, preventing polyps from becoming malignant and forming colorectal cancer instances. In recent years, computer-aided polyp detection systems have been widely used in colonoscopies to improve the quality of colonoscopy examination and increase the polyp detection rate. Currently, the most efficient computer-aided systems are built with machine learning methods. However, developing such a computer-aided detection system requires experienced doctors to label a large number of image data from colonoscopy videos, which is extremely time-consuming, laborious and expensive. One possible solution is to adopt a semi-supervised learning, which can build a detection system on a dataset where part of its data is not necessary to be labeled. In this paper, on the basis of state-of-the-art object detection method and semi-supervised learning technique, we design and implement a semi-supervised colonoscopy polyp detection system containing four main steps: running standard supervised training with all labeled data; running inference on unlabeled data to obtain pseudo labels; applying a set of strong augmentation to both unlabeled data and pseudo label; combining labeled data, and unlabeled data with its pseudo labels to retrain the detector. The semi-supervised learning system is evaluated both on public dataset and our original private dataset and proves its effectiveness. Also, the inference speed of the semi-supervised learning system can meet the requirement of real-time operation.