Red Lesion Detection in Retinal Fundus Images Using Frangi-based Filters

Page view(s)
22
Checked on Feb 08, 2024
Red Lesion Detection in Retinal Fundus Images Using Frangi-based Filters
Title:
Red Lesion Detection in Retinal Fundus Images Using Frangi-based Filters
Journal Title:
2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)
Publication Date:
25 August 2015
Citation:
Srivastava, R.; Wong, D.W.K.; Lixin Duan; Jiang Liu; Tien Yin Wong, "Red lesion detection in retinal fundus images using Frangi-based filters," in Engineering in Medicine and Biology Society (EMBC), 2015, 37th Annual International Conference of the IEEE , pp.5663-5666, 25-29 Aug. 2015.
Abstract:
This paper presents a method to detect red lesions related to Diabetic Retinopathy (DR), namely Microaneurysms and Hemorrhages from retinal fundus images with robustness to the presence of blood vessels. Filters based on Frangi filters are used for the first time for this task. Green channel of the input image was decomposed into smaller sub images and proposed filters were applied to each sub image after initial preprocessing. Features were extracted from the filter response and used to train a Support Vector Machine classifier to predict whether a test image had lesions or not. Experiments were performed on a dataset of 143 retinal fundus and the proposed method achieved areas under the ROC curve equal to 0.97 and 0.87 for Microaneurysms and Hemorrhages respectively. Results show the effectiveness of the proposed method for detecting red lesions. This method can help significantly in automated detection of DR with fewer false positives.
License type:
PublisherCopyrights
Funding Info:
Description:
© 2015 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
ISSN:
1094-687X
1558-4615
ISBN:
978-1-4244-9271-8
978-1-4244-9270-1
Files uploaded:
File Size Format Action
There are no attached files.