Vessel segmentation in eye fundus images using ensemble learning and curve fitting

Authors:
Elco Oost;Yuki Akatsuka;Akinobu Shimizu;Hidefumi Kobatake;Daisuke Furukawa;Akihiro Katayama
Affiliations:
Dept. of Electrical and Electronic Engineering, Tokyo University of Agriculture and Technology;Dept. of Electrical and Electronic Engineering, Tokyo University of Agriculture and Technology;Dept. of Electrical and Electronic Engineering, Tokyo University of Agriculture and Technology;Dept. of Electrical and Electronic Engineering, Tokyo University of Agriculture and Technology;Medical Informatics and Computer Vision Division, Canon Inc.;Medical Informatics and Computer Vision Division, Canon Inc.
Venue:
ISBI'10 Proceedings of the 2010 IEEE international conference on Biomedical imaging: from nano to Macro
Year:
2010

Citing 1
Cited 1

A decision-theoretic generalization of on-line learning and an application to boosting

Journal of Computer and System Sciences - Special issue: 26th annual ACM symposium on the theory of computing & STOC'94, May 23–25, 1994, and second annual Europe an conference on computational learning theory (EuroCOLT'95), March 13–15, 1995

Automated delineation of lung tumors from CT images using a single click ensemble segmentation approach

Pattern Recognition

Quantified Score

Hi-index	0.00

Visualization

Abstract

A novel segmentation algorithm for the detection of retinal vessels in funduscopic images is proposed, in which the benefits of both supervised and unsupervised methods are exploited. Ensemble learning based segmentation (ELBS) is employed for the segmentation of large and medium sized vessels, after which a local curve fitting technique is used for the detection of the thin retinal vessels. The general ELBS algorithm is modified to boost performance by the incorporation of specific knowledge of false negative segmentation result areas. Curve fitting is based on a two-hypotheses polynomial regression and is capable of automatically removing outliers from a point cloud. Evaluation on the DRIVE database compared the presented method favorably to previously published algorithms. Sensitivity and specificity were 0.8854 and 0.9363.