AUTHOR=Su Houcheng , Lin Bin , Huang Xiaoshuang , Li Jiao , Jiang Kailin , Duan Xuliang 

TITLE=MBFFNet: Multi-Branch Feature Fusion Network for Colonoscopy

JOURNAL=Frontiers in Bioengineering and Biotechnology

VOLUME=Volume 9 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/bioengineering-and-biotechnology/articles/10.3389/fbioe.2021.696251

DOI=10.3389/fbioe.2021.696251

ISSN=2296-4185

ABSTRACT=Colonoscopy is one of the main methods for the detection of rectal polyps, rectal cancer and other diseases at present. With the rapid development of computer vision, deep learning-based semantic segmentation methods can be applied to the detection of medical lesions. However, it is a challenging for current methods to detect polyps with high accuracy and read-time performance. In order to solve this problem, we proposed a MBFFNet (Multi - branch feature fusion network, MBFFNet), which is an accurate real-time segmentation method for detecting colonoscopy. Specifically, first, we use U-Net as the basis of our model architecture, and adopt stepwise sampling with channel multiplication to integrate features, which decreases the number of FLOPs caused by stacking channels in U-Net. Second, in order to improve the accuracy of our model, we extract features from multiple layers and resize the feature maps to the same size in different ways, such as up-sampling and pooling, to supplement the lost information in the multiplication based up-sampling. Based on mIOU, Dice loss with CE, we conduct experiments on both the CPU and GPU environments to verify the effectiveness of our model. The experimental results show that our MBFFNet is superior to the selected baselines in accuracy, model size and FLOPS. In addition, the experiments on other types of medical tasks show that MBFFNet has a good generalization ability in medical image segmentation.