Edge-Boosted U-Net for 2D Medical Image Segmentation

Renjie Zhao,Guitao Cao,Weiting Chen

doi:10.1109/access.2019.2953727

Abstract

Automatic medical image segmentation has always been a heated study in computer-assisted diagnosis (CAD). It is a quite challenging task due to the diversity and complexity of medical images. In this paper, we propose an Edge-boosted U-Net (EU-Net) to address the problem of medical image segmentation. The architecture is basically a U-shape network, combined with three main parts: Edge Aggregation Path, Feature Fusion Block, and Feature Attention Block. The Edge Aggregation Path is to extract multilevel edge-relevant information. The Feature Fusion Block is designed to fuse features from different paths. And the Feature Attention Block is embedded in the network to generate more informative feature maps. The collaboration of these three parts effectively boosts the performance of the whole network. We verified the importance of each part by conducting several experiments. Meanwhile, we compared the proposed method with other state-of-the-art methods on three different modalities of public medical image datasets. Our method achieves the superiority with IoU and dice coefficient respectively on all the datasets. Notably, it attains 2% accuracy improvement over other methods on the challenging datasets.

Highlights

Automatic medical image segmentation is a critical and fundamental task in medical image analysis, surgical planning and treatments
1) EVALUATION ON FEATURE ATTENTION BLOCK(FAB) In the first ablation study, in order to analyze the effectiveness of the proposed Feature Attention Block, we conducted the
Single: A network like U-Net Sum-Fusion: A network with main decoder path and Edge Aggregation Path (EAP) Cat-Fusion: A network with main decoder path and EAP Feature Fusion Block (FFB): Our whole network We compared these networks in terms of IoU and Dice coefficient

Summary

Introduction

Automatic medical image segmentation is a critical and fundamental task in medical image analysis, surgical planning and treatments. An unknown malignant tumor can be found by a doctor more and quickly under the help of automatic image segmentation. An accurate segmentation mask tends to facilitate doctors’ reliable diagnosis. With the rapid development of image processing techniques, various methods are deployed to address the need of image segmentation, such as cell segmentation [1]–[3], brain segmentation [4]–[6], liver segmentation [7]–[9] and so on. Facing the fundamental task of medical image segmentation, researchers have generally proposed two kinds of methods up to now.

Methods

Results

Conclusion