Share Email Print

Proceedings Paper

New algorithm for detecting smaller retinal blood vessels in fundus images
Author(s): Robert LeAnder; Praveen I. Bidari; Tauseef A. Mohammed; Moumita Das; Scott E. Umbaugh
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

About 4.1 million Americans suffer from diabetic retinopathy. To help automatically diagnose various stages of the disease, a new blood-vessel-segmentation algorithm based on spatial high-pass filtering was developed to automatically segment blood vessels, including the smaller ones, with low noise. Methods: Image database: Forty, 584 x 565-pixel images were collected from the DRIVE image database. Preprocessing: Green-band extraction was used to obtain better contrast, which facilitated better visualization of retinal blood vessels. A spatial highpass filter of mask-size 11 was applied. A histogram stretch was performed to enhance contrast. A median filter was applied to mitigate noise. At this point, the gray-scale image was converted to a binary image using a binary thresholding operation. Then, a NOT operation was performed by gray-level value inversion between 0 and 255. Postprocessing: The resulting image was AND-ed with its corresponding ring mask to remove the outer-ring (lens-edge) artifact. At this point, the above algorithm steps had extracted most of the major and minor vessels, with some intersections and bifurcations missing. Vessel segments were reintegrated using the Hough transform. Results: After applying the Hough transform, both the average peak SNR and the RMS error improved by 10%. Pratt's Figure of Merit (PFM) was decreased by 6%. Those averages were better than [1] by 10-30%. Conclusions: The new algorithm successfully preserved the details of smaller blood vessels and should prove successful as a segmentation step for automatically identifying diseases that affect retinal blood vessels.

Paper Details

Date Published: 9 March 2010
PDF: 7 pages
Proc. SPIE 7624, Medical Imaging 2010: Computer-Aided Diagnosis, 76243B (9 March 2010); doi: 10.1117/12.844996
Show Author Affiliations
Robert LeAnder, Southern Illinois Univ. Edwardsville (United States)
Praveen I. Bidari, Southern Illinois Univ. Edwardsville (United States)
Tauseef A. Mohammed, Southern Illinois Univ. Edwardsville (United States)
Moumita Das, Southern Illinois Univ. Edwardsville (United States)
Scott E. Umbaugh, Southern Illinois Univ. Edwardsville (United States)

Published in SPIE Proceedings Vol. 7624:
Medical Imaging 2010: Computer-Aided Diagnosis
Nico Karssemeijer; Ronald M. Summers, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?