Automated detection of retinal health using PHOG and SURF features extracted from fundus images

Joel E.W. Koh, Eddie Y.K. Ng, Sulatha V. Bhandary, Augustinus Laude, U. Rajendra Acharya

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Many health-related problems arise with aging. One of the diseases that is prevalent among the elderly is the loss of sight. Various eye diseases, namely age-related macular degeneration (AMD), diabetic retinopathy (DR), and glaucoma are the prime causes of vision loss as we grow old. Nevertheless, early detection of such eye diseases can impede the progression of this problem. Therefore, the elderly are encouraged to attend regular eye checkups for early detection of eye diseases. However, it is time-consuming and laborious to conduct a mass eye screening session frequently. Hence, we proposed a novel approach to develop an automated retinal health screening system in this work. This paper discusses a retinal screening system to automatically differentiate normal image from abnormal (AMD, DR, and glaucoma) fundus images. The fundus images are subjected to the pyramid histogram of oriented gradients (PHOG) and speeded up robust features (SURF) techniques. Then, the extracted data are subjected to adaptive synthetic sampling to balance the number of data in the two classes (normal and abnormal). Subsequently, we employed the canonical correlation analysis approach to fuse the highly-correlated features extracted from the two (PHOG and SURF) descriptors. We have achieved 96.21% accuracy, 95.00% sensitivity, and 97.42% specificity with ten-fold cross-validation strategy using k-nearest neighbor (kNN) classifier. This novel algorithm has high potential in the diagnosis of normal eyes during the mass eye screening session or in polyclinics quickly and reliably. Hence, the patients having abnormal eyes can be sent to the main hospitals which will reduce the workload for the ophthalmologists. [Figure not available: see fulltext.]

Original languageEnglish
Pages (from-to)1-15
Number of pages15
JournalApplied Intelligence
DOIs
Publication statusPublished - 29-08-2018

Fingerprint

Screening
Health
Electric fuses
Classifiers
Aging of materials
Sampling

All Science Journal Classification (ASJC) codes

  • Artificial Intelligence

Cite this

Koh, Joel E.W. ; Ng, Eddie Y.K. ; Bhandary, Sulatha V. ; Laude, Augustinus ; Acharya, U. Rajendra. / Automated detection of retinal health using PHOG and SURF features extracted from fundus images. In: Applied Intelligence. 2018 ; pp. 1-15.
@article{6a9073b9d1a9442682f0698b921d1b76,
title = "Automated detection of retinal health using PHOG and SURF features extracted from fundus images",
abstract = "Many health-related problems arise with aging. One of the diseases that is prevalent among the elderly is the loss of sight. Various eye diseases, namely age-related macular degeneration (AMD), diabetic retinopathy (DR), and glaucoma are the prime causes of vision loss as we grow old. Nevertheless, early detection of such eye diseases can impede the progression of this problem. Therefore, the elderly are encouraged to attend regular eye checkups for early detection of eye diseases. However, it is time-consuming and laborious to conduct a mass eye screening session frequently. Hence, we proposed a novel approach to develop an automated retinal health screening system in this work. This paper discusses a retinal screening system to automatically differentiate normal image from abnormal (AMD, DR, and glaucoma) fundus images. The fundus images are subjected to the pyramid histogram of oriented gradients (PHOG) and speeded up robust features (SURF) techniques. Then, the extracted data are subjected to adaptive synthetic sampling to balance the number of data in the two classes (normal and abnormal). Subsequently, we employed the canonical correlation analysis approach to fuse the highly-correlated features extracted from the two (PHOG and SURF) descriptors. We have achieved 96.21{\%} accuracy, 95.00{\%} sensitivity, and 97.42{\%} specificity with ten-fold cross-validation strategy using k-nearest neighbor (kNN) classifier. This novel algorithm has high potential in the diagnosis of normal eyes during the mass eye screening session or in polyclinics quickly and reliably. Hence, the patients having abnormal eyes can be sent to the main hospitals which will reduce the workload for the ophthalmologists. [Figure not available: see fulltext.]",
author = "Koh, {Joel E.W.} and Ng, {Eddie Y.K.} and Bhandary, {Sulatha V.} and Augustinus Laude and Acharya, {U. Rajendra}",
year = "2018",
month = "8",
day = "29",
doi = "10.1007/s10489-017-1048-3",
language = "English",
pages = "1--15",
journal = "Applied Intelligence",
issn = "0924-669X",
publisher = "Springer Netherlands",

}

Automated detection of retinal health using PHOG and SURF features extracted from fundus images. / Koh, Joel E.W.; Ng, Eddie Y.K.; Bhandary, Sulatha V.; Laude, Augustinus; Acharya, U. Rajendra.

In: Applied Intelligence, 29.08.2018, p. 1-15.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Automated detection of retinal health using PHOG and SURF features extracted from fundus images

AU - Koh, Joel E.W.

AU - Ng, Eddie Y.K.

AU - Bhandary, Sulatha V.

AU - Laude, Augustinus

AU - Acharya, U. Rajendra

PY - 2018/8/29

Y1 - 2018/8/29

N2 - Many health-related problems arise with aging. One of the diseases that is prevalent among the elderly is the loss of sight. Various eye diseases, namely age-related macular degeneration (AMD), diabetic retinopathy (DR), and glaucoma are the prime causes of vision loss as we grow old. Nevertheless, early detection of such eye diseases can impede the progression of this problem. Therefore, the elderly are encouraged to attend regular eye checkups for early detection of eye diseases. However, it is time-consuming and laborious to conduct a mass eye screening session frequently. Hence, we proposed a novel approach to develop an automated retinal health screening system in this work. This paper discusses a retinal screening system to automatically differentiate normal image from abnormal (AMD, DR, and glaucoma) fundus images. The fundus images are subjected to the pyramid histogram of oriented gradients (PHOG) and speeded up robust features (SURF) techniques. Then, the extracted data are subjected to adaptive synthetic sampling to balance the number of data in the two classes (normal and abnormal). Subsequently, we employed the canonical correlation analysis approach to fuse the highly-correlated features extracted from the two (PHOG and SURF) descriptors. We have achieved 96.21% accuracy, 95.00% sensitivity, and 97.42% specificity with ten-fold cross-validation strategy using k-nearest neighbor (kNN) classifier. This novel algorithm has high potential in the diagnosis of normal eyes during the mass eye screening session or in polyclinics quickly and reliably. Hence, the patients having abnormal eyes can be sent to the main hospitals which will reduce the workload for the ophthalmologists. [Figure not available: see fulltext.]

AB - Many health-related problems arise with aging. One of the diseases that is prevalent among the elderly is the loss of sight. Various eye diseases, namely age-related macular degeneration (AMD), diabetic retinopathy (DR), and glaucoma are the prime causes of vision loss as we grow old. Nevertheless, early detection of such eye diseases can impede the progression of this problem. Therefore, the elderly are encouraged to attend regular eye checkups for early detection of eye diseases. However, it is time-consuming and laborious to conduct a mass eye screening session frequently. Hence, we proposed a novel approach to develop an automated retinal health screening system in this work. This paper discusses a retinal screening system to automatically differentiate normal image from abnormal (AMD, DR, and glaucoma) fundus images. The fundus images are subjected to the pyramid histogram of oriented gradients (PHOG) and speeded up robust features (SURF) techniques. Then, the extracted data are subjected to adaptive synthetic sampling to balance the number of data in the two classes (normal and abnormal). Subsequently, we employed the canonical correlation analysis approach to fuse the highly-correlated features extracted from the two (PHOG and SURF) descriptors. We have achieved 96.21% accuracy, 95.00% sensitivity, and 97.42% specificity with ten-fold cross-validation strategy using k-nearest neighbor (kNN) classifier. This novel algorithm has high potential in the diagnosis of normal eyes during the mass eye screening session or in polyclinics quickly and reliably. Hence, the patients having abnormal eyes can be sent to the main hospitals which will reduce the workload for the ophthalmologists. [Figure not available: see fulltext.]

UR - http://www.scopus.com/inward/record.url?scp=85028544709&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85028544709&partnerID=8YFLogxK

U2 - 10.1007/s10489-017-1048-3

DO - 10.1007/s10489-017-1048-3

M3 - Article

SP - 1

EP - 15

JO - Applied Intelligence

JF - Applied Intelligence

SN - 0924-669X

ER -