Retinal vessel image segmentation algorithm based on encoder-decoder structure

The accurate segmentation of retinal vessel image is significant for the early diagnosis of some diseases. A retinal vessel image segmentation algorithm based on an encoder-decoder structure is proposed. In the encoding section, the Inception module is used, which uses convolution kernels of differe...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Multimedia tools and applications Jg. 81; H. 23; S. 33361 - 33373
Hauptverfasser: Zhai, ZhengLi, Feng, Shu, Yao, Luyao, Li, Penghui
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York Springer US 01.09.2022
Springer Nature B.V
Schlagworte:
Algorithms
Blood vessels
Coders
Computer Communication Networks
Computer Science
Data Structures and Information Theory
Decoding
Encoders-Decoders
Feature extraction
Image segmentation
Medical imaging
Modules
Multimedia Information Systems
Special Purpose and Application-Based Systems
Inception module
Image segmentation
Feature fusion
Retinal blood vessels
Pyramid pooling modules
ISSN:1380-7501, 1573-7721
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The accurate segmentation of retinal vessel image is significant for the early diagnosis of some diseases. A retinal vessel image segmentation algorithm based on an encoder-decoder structure is proposed. In the encoding section, the Inception module is used, which uses convolution kernels of different scales to achieve features extract to obtain images multi-scale information. So as to enable the model to perceive blood vessels of various shapes and improve the accuracy of segmentation of small blood vessels, multiple pyramid pooling modules are adopted in the decoding process to aggregate more contextual information, and multi-scale and multi-local area feature fusion is used to improve segmentation effect. In addition, the feature fusion method is applied in the upsampling process to fuse low-order semantic features to obtain more low-level detailed information, thereby further promote the segmentation effect. The experimental results on DRIVE and STAER fundus image datasets show that the algorithm has higher sensitivity, accuracy and AUC value compared with other algorithms, and the segmentation effect is better.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-022-13176-5