Detection of colorectal polyps based on deep learning

Colorectal polyps are usually a sign of early precancerous lesions of colorectal cancer. Colorectal polyp screening is of great significance for the early detection and prevention of colorectal cancer. CT images can provide an important basis for the detection of colorectal polyps, but the polyps in CT images are small, and the traditional CT image detection methods rely heavily on the segmentation results of the inner and outer walls of the colon and rectum, so the detection performance is not good. In order to solve this problem, a deep learning method based on expert knowledge and multi-scale feature network is proposed. drawing lessons from the film reading experience of radiologists, the size knowledge of polyps is obtained by clustering, and the hierarchical network is adopted at the same time. Through the fusion of deep and shallow feature information, multi-scale features are used to better detect small targets. The final experimental results show that the proposed method has a mean average accuracy of 98.8% on the open data set. It shows that the proposed multi-scale feature fusion module has better performance in detecting small medical targets such as colorectal polyps.