Abstract

Colonoscopy is one of the most effective means of detecting intestinal polyps and colon cancer. With the development of machine vision, doctors often use automation to help diagnose intestinal polyps when using colonoscopy. The high rate of missed polyps during traditional intestinal polyp detection can lead to an increased recurrence rate and cancer of polyps. Therefore, we propose a multi-scale fusion network with boundary judgment function. The network can accurately segment smaller polyps and improve the blurred boundaries of segmented images. The network achieves the purpose of deepening the boundary information by using the attention mechanism by judging the global information generated by the encoder at different levels and the prediction mask map obtained by multi-scale fusion. Experimental results show that the network can accurately identify the boundaries of diminutive polyps and improve the overall accuracy of polyp segmentation. Compared with the latest network models, we achieved first place in several metrics in the mainstream datasets(e.t., Kvasir, CVC-ClinicDB, CVC-300, and ETIS LaribPolypDB). Codes available: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/Wdy1997/J–Net.git</uri>

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call