DBFU-Net: Double branch fusion U-Net with hard example weighting train strategy to segment retinal vessel

Jianping Huang; Zefang Lin; Yingyin Chen; Xiao Zhang; Wei Zhao; Jie Zhang; Yong Li; Xu He; Meixiao Zhan; Ligong Lu; Xiaofei Jiang; Yongjun Peng

doi:10.7717/peerj-cs.871

DBFU-Net: Double branch fusion U-Net with hard example weighting train strategy to segment retinal vessel

PeerJ Comput Sci. 2022 Feb 18:8:e871. doi: 10.7717/peerj-cs.871. eCollection 2022.

Authors

Jianping Huang^#¹, Zefang Lin^#¹, Yingyin Chen¹, Xiao Zhang¹, Wei Zhao¹, Jie Zhang², Yong Li¹, Xu He¹, Meixiao Zhan¹, Ligong Lu¹, Xiaofei Jiang³, Yongjun Peng²

Affiliations

¹ Zhuhai People's Hospital, Zhuhai Hospital Affiliated with Jinan University, Jinan University, Zhuhai Interventional Medical Center, Zhuhai Precision Medical Center, Zhuhai, China.
² Zhuhai People's Hospital, Zhuhai Hospital Affiliated with Jinan University, Jinan University, Department of Nuclear Medicine, Zhuhai, China.
³ Zhuhai People's Hospital, Zhuhai Hospital Affiliated with Jinan University, Jinan University, Department of cardiology, Zhuhai, China.

^# Contributed equally.

Abstract

Background: Many fundus imaging modalities measure ocular changes. Automatic retinal vessel segmentation (RVS) is a significant fundus image-based method for the diagnosis of ophthalmologic diseases. However, precise vessel segmentation is a challenging task when detecting micro-changes in fundus images, e.g., tiny vessels, vessel edges, vessel lesions and optic disc edges.

Methods: In this paper, we will introduce a novel double branch fusion U-Net model that allows one of the branches to be trained by a weighting scheme that emphasizes harder examples to improve the overall segmentation performance. A new mask, we call a hard example mask, is needed for those examples that include a weighting strategy that is different from other methods. The method we propose extracts the hard example mask by morphology, meaning that the hard example mask does not need any rough segmentation model. To alleviate overfitting, we propose a random channel attention mechanism that is better than the drop-out method or the L2-regularization method in RVS.

Results: We have verified the proposed approach on the DRIVE, STARE and CHASE datasets to quantify the performance metrics. Compared to other existing approaches, using those dataset platforms, the proposed approach has competitive performance metrics. (DRIVE: F1-Score = 0.8289, G-Mean = 0.8995, AUC = 0.9811; STARE: F1-Score = 0.8501, G-Mean = 0.9198, AUC = 0.9892; CHASE: F1-Score = 0.8375, G-Mean = 0.9138, AUC = 0.9879).

Discussion: The segmentation results showed that DBFU-Net with RCA achieves competitive performance in three RVS datasets. Additionally, the proposed morphological-based extraction method for hard examples can reduce the computational cost. Finally, the random channel attention mechanism proposed in this paper has proven to be more effective than other regularization methods in the RVS task.

Keywords: Deep learning; Fundus image; Hard examples; Random channel attention mechanism; Vessel segmentation.

Grants and funding

This study is supported by the National Key Research and Development Program of China (Grant No. 2017YFA0205200), the National Natural Science Foundation of China (Grant No. 81901857), and the Natural Science Foundation of Guangdong Province, China (No. 2018A030313074). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.