On application of gaussian kernel to retinal blood tracing

Mohd Asyraf Zulkifley, Ain Nazari, Siti Khadijah, Adhi Harmoko Saputro

Research output: Contribution to journalArticle

Abstract

Fast disease screening system is becoming very important nowadays as it is useful in minimizing the pain suffered by the patient. Eye is one of the important organs used to screen various diseases such as diabetes, pterygium, glaucoma, stroke and etc. Two popular modalities to capture the visualization information of the eyes are photographed and fundus images. Photographed image captures the exterior appearance of the eyes, while fundus image captures the interior surface of the eye. This study focuses on the fundus image, in which the objective is to extract the retinal blood vessels. Condition of the vessels is a good cue for identifying microaneurysm where a sudden jump in vessels trajectory indicate a blockage in the blood flow. Hence, an edge operator is used to extract the possible lines. Two removal steps, which are outer ring and optic disc removal are performed, so that only the vessels are detected. A filling operation is performed by using gaussian kernel such that areas surrounding existing edges are analyzed to label it as blood vessel or not. The proposed technique is then tested on two databases; DRIVE and STARE. Four performance metrics are calculated based on accuracy, sensitivity, error and specificity where the best results are 0.880, 0.546, 0.120 and 0.912 respectively, obtained based on DRIVE database. The proposed system also works well in STARE database with accuracy of 0.860 and error of 0.140. The system can be further improved by using better detection scheme because the number of false negative is relatively high.

Original languageEnglish
Pages (from-to)19-24
Number of pages6
JournalJournal of Theoretical and Applied Information Technology
Volume77
Issue number1
Publication statusPublished - 10 Jul 2015

Fingerprint

Gaussian Kernel
Tracing
Blood
Blood vessels
Vessel
Blood Vessels
Medical problems
Information Visualization
Labels
Diabetes
Optics
Screening
Pain
Performance Metrics
Visualization
Blood Flow
Stroke
Trajectories
Modality
Specificity

Keywords

  • Blood vessel detection
  • Fundus image
  • Gaussian kernel
  • Optic disc removal
  • Outer ring removal

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

On application of gaussian kernel to retinal blood tracing. / Zulkifley, Mohd Asyraf; Nazari, Ain; Khadijah, Siti; Saputro, Adhi Harmoko.

In: Journal of Theoretical and Applied Information Technology, Vol. 77, No. 1, 10.07.2015, p. 19-24.

Research output: Contribution to journalArticle

Zulkifley, Mohd Asyraf ; Nazari, Ain ; Khadijah, Siti ; Saputro, Adhi Harmoko. / On application of gaussian kernel to retinal blood tracing. In: Journal of Theoretical and Applied Information Technology. 2015 ; Vol. 77, No. 1. pp. 19-24.
@article{9585c1b13c84478885e76bde811956c1,
title = "On application of gaussian kernel to retinal blood tracing",
abstract = "Fast disease screening system is becoming very important nowadays as it is useful in minimizing the pain suffered by the patient. Eye is one of the important organs used to screen various diseases such as diabetes, pterygium, glaucoma, stroke and etc. Two popular modalities to capture the visualization information of the eyes are photographed and fundus images. Photographed image captures the exterior appearance of the eyes, while fundus image captures the interior surface of the eye. This study focuses on the fundus image, in which the objective is to extract the retinal blood vessels. Condition of the vessels is a good cue for identifying microaneurysm where a sudden jump in vessels trajectory indicate a blockage in the blood flow. Hence, an edge operator is used to extract the possible lines. Two removal steps, which are outer ring and optic disc removal are performed, so that only the vessels are detected. A filling operation is performed by using gaussian kernel such that areas surrounding existing edges are analyzed to label it as blood vessel or not. The proposed technique is then tested on two databases; DRIVE and STARE. Four performance metrics are calculated based on accuracy, sensitivity, error and specificity where the best results are 0.880, 0.546, 0.120 and 0.912 respectively, obtained based on DRIVE database. The proposed system also works well in STARE database with accuracy of 0.860 and error of 0.140. The system can be further improved by using better detection scheme because the number of false negative is relatively high.",
keywords = "Blood vessel detection, Fundus image, Gaussian kernel, Optic disc removal, Outer ring removal",
author = "Zulkifley, {Mohd Asyraf} and Ain Nazari and Siti Khadijah and Saputro, {Adhi Harmoko}",
year = "2015",
month = "7",
day = "10",
language = "English",
volume = "77",
pages = "19--24",
journal = "Journal of Theoretical and Applied Information Technology",
issn = "1992-8645",
publisher = "Asian Research Publishing Network (ARPN)",
number = "1",

}

TY - JOUR

T1 - On application of gaussian kernel to retinal blood tracing

AU - Zulkifley, Mohd Asyraf

AU - Nazari, Ain

AU - Khadijah, Siti

AU - Saputro, Adhi Harmoko

PY - 2015/7/10

Y1 - 2015/7/10

N2 - Fast disease screening system is becoming very important nowadays as it is useful in minimizing the pain suffered by the patient. Eye is one of the important organs used to screen various diseases such as diabetes, pterygium, glaucoma, stroke and etc. Two popular modalities to capture the visualization information of the eyes are photographed and fundus images. Photographed image captures the exterior appearance of the eyes, while fundus image captures the interior surface of the eye. This study focuses on the fundus image, in which the objective is to extract the retinal blood vessels. Condition of the vessels is a good cue for identifying microaneurysm where a sudden jump in vessels trajectory indicate a blockage in the blood flow. Hence, an edge operator is used to extract the possible lines. Two removal steps, which are outer ring and optic disc removal are performed, so that only the vessels are detected. A filling operation is performed by using gaussian kernel such that areas surrounding existing edges are analyzed to label it as blood vessel or not. The proposed technique is then tested on two databases; DRIVE and STARE. Four performance metrics are calculated based on accuracy, sensitivity, error and specificity where the best results are 0.880, 0.546, 0.120 and 0.912 respectively, obtained based on DRIVE database. The proposed system also works well in STARE database with accuracy of 0.860 and error of 0.140. The system can be further improved by using better detection scheme because the number of false negative is relatively high.

AB - Fast disease screening system is becoming very important nowadays as it is useful in minimizing the pain suffered by the patient. Eye is one of the important organs used to screen various diseases such as diabetes, pterygium, glaucoma, stroke and etc. Two popular modalities to capture the visualization information of the eyes are photographed and fundus images. Photographed image captures the exterior appearance of the eyes, while fundus image captures the interior surface of the eye. This study focuses on the fundus image, in which the objective is to extract the retinal blood vessels. Condition of the vessels is a good cue for identifying microaneurysm where a sudden jump in vessels trajectory indicate a blockage in the blood flow. Hence, an edge operator is used to extract the possible lines. Two removal steps, which are outer ring and optic disc removal are performed, so that only the vessels are detected. A filling operation is performed by using gaussian kernel such that areas surrounding existing edges are analyzed to label it as blood vessel or not. The proposed technique is then tested on two databases; DRIVE and STARE. Four performance metrics are calculated based on accuracy, sensitivity, error and specificity where the best results are 0.880, 0.546, 0.120 and 0.912 respectively, obtained based on DRIVE database. The proposed system also works well in STARE database with accuracy of 0.860 and error of 0.140. The system can be further improved by using better detection scheme because the number of false negative is relatively high.

KW - Blood vessel detection

KW - Fundus image

KW - Gaussian kernel

KW - Optic disc removal

KW - Outer ring removal

UR - http://www.scopus.com/inward/record.url?scp=84936980933&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84936980933&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84936980933

VL - 77

SP - 19

EP - 24

JO - Journal of Theoretical and Applied Information Technology

JF - Journal of Theoretical and Applied Information Technology

SN - 1992-8645

IS - 1

ER -