Souahlia, Abdelkenm, Belatreche, Ammar, Benyettou, Abdelkader and Curran, Kevin (2017) Blood vessel segmentation in retinal images using echo state networks. In: 2017 Ninth International Conference on Advanced Computational Intelligence (ICACI). IEEE, pp. 91-98.
Full text not available from this repository.Abstract
We propose a novel supervised technique for blood vessel segmentation in retinal images based on echo state networks. Retinal vessel segmentation is widely used for numerous clinical purposes such as the detection of various cardiovascular and ophthalmologic diseases. A large number of retinal vessel segmentation methods have been reported, yet achieving accurate and efficient vessel segmentation still remains a challenge. Recently, reservoir computing has drawn much attention as a new computing framework based on recurrent neural networks. The Echo State Network (ESN), which uses neural nodes as the computing elements of the recurrent network, represents one of the efficient learning models of reservoir computing. This paper investigates the viability of echo state networks for blood vessel segmentation in retinal images. Initial image features are projected onto the echo state network reservoir which maps them, through its internal nodes activations, into a new set of features to be classified into vessel or non-vessel by the echo state network readout which consists, in the proposed approach, of a multi-layer perceptron. Experimental results on the publicly available DRIVE dataset, commonly used in retinal vessel segmentation research, demonstrate the ability of the proposed method in achieving promising performance results in terms of both segmentation accuracy and efficiency.
Item Type: | Book Section |
---|---|
Uncontrolled Keywords: | echo state network, retinal images, vessel segmentation, pixel classification, feature extraction |
Subjects: | G400 Computer Science |
Department: | Faculties > Engineering and Environment > Computer and Information Sciences |
Depositing User: | Ay Okpokam |
Date Deposited: | 29 Aug 2017 16:19 |
Last Modified: | 12 Oct 2019 20:46 |
URI: | http://nrl.northumbria.ac.uk/id/eprint/31684 |
Downloads
Downloads per month over past year