DREAM: Diabetic Retinopathy Analysis Using Machine Learning

Sohini Roychowdhury, Dara Koozekanani, Keshab K Parhi

Research output: Contribution to journalArticle

115 Citations (Scopus)

Abstract

This paper presents a computer-aided screening system (DREAM) that analyzes fundus images with varying illumination and fields of view, and generates a severity grade for diabetic retinopathy (DR) using machine learning. Classifiers such as the Gaussian Mixture model (GMM), k-nearest neighbor (kNN), support vector machine (SVM), and AdaBoost are analyzed for classifying retinopathy lesions from nonlesions. GMM and kNN classifiers are found to be the best classifiers for bright and red lesion classification, respectively. A main contribution of this paper is the reduction in the number of features used for lesion classification by feature ranking using Adaboost where 30 top features are selected out of 78. A novel two-step hierarchical classification approach is proposed where the nonlesions or false positives are rejected in the first step. In the second step, the bright lesions are classified as hard exudates and cotton wool spots, and the red lesions are classified as hemorrhages and micro-aneurysms. This lesion classification problem deals with unbalanced datasets and SVM or combination classifiers derived from SVM using the Dempster-Shafer theory are found to incur more classification error than the GMM and kNN classifiers due to the data imbalance. The DR severity grading system is tested on 1200 images from the publicly available MESSIDOR dataset. The DREAM system achieves 100% sensitivity, 53.16% specificity, and 0.904 AUC, compared to the best reported 96% sensitivity, 51% specificity, and 0.875 AUC, for classifying images as with or without DR. The feature reduction further reduces the average computation time for DR severity per image from 59.54 to 3.46 s.

Original languageEnglish (US)
Article number6680633
Pages (from-to)1717-1728
Number of pages12
JournalIEEE Journal of Biomedical and Health Informatics
Volume18
Issue number5
DOIs
StatePublished - Sep 1 2014

Fingerprint

Diabetic Retinopathy
Learning systems
Classifiers
Support vector machines
Adaptive boosting
Area Under Curve
Sensitivity and Specificity
Wool
Exudates and Transudates
Lighting
Cotton
Aneurysm
Screening
Machine Learning
Hemorrhage
Support Vector Machine
Datasets

Keywords

  • Bright lesions
  • classification
  • diabetic retinopathy (DR)
  • fundus image processing
  • red lesions
  • segmentation
  • severity grade

Cite this

DREAM : Diabetic Retinopathy Analysis Using Machine Learning. / Roychowdhury, Sohini; Koozekanani, Dara; Parhi, Keshab K.

In: IEEE Journal of Biomedical and Health Informatics, Vol. 18, No. 5, 6680633, 01.09.2014, p. 1717-1728.

Research output: Contribution to journalArticle

@article{b91b4f7caa1d4786bd0ee852220d9e1e,
title = "DREAM: Diabetic Retinopathy Analysis Using Machine Learning",
abstract = "This paper presents a computer-aided screening system (DREAM) that analyzes fundus images with varying illumination and fields of view, and generates a severity grade for diabetic retinopathy (DR) using machine learning. Classifiers such as the Gaussian Mixture model (GMM), k-nearest neighbor (kNN), support vector machine (SVM), and AdaBoost are analyzed for classifying retinopathy lesions from nonlesions. GMM and kNN classifiers are found to be the best classifiers for bright and red lesion classification, respectively. A main contribution of this paper is the reduction in the number of features used for lesion classification by feature ranking using Adaboost where 30 top features are selected out of 78. A novel two-step hierarchical classification approach is proposed where the nonlesions or false positives are rejected in the first step. In the second step, the bright lesions are classified as hard exudates and cotton wool spots, and the red lesions are classified as hemorrhages and micro-aneurysms. This lesion classification problem deals with unbalanced datasets and SVM or combination classifiers derived from SVM using the Dempster-Shafer theory are found to incur more classification error than the GMM and kNN classifiers due to the data imbalance. The DR severity grading system is tested on 1200 images from the publicly available MESSIDOR dataset. The DREAM system achieves 100{\%} sensitivity, 53.16{\%} specificity, and 0.904 AUC, compared to the best reported 96{\%} sensitivity, 51{\%} specificity, and 0.875 AUC, for classifying images as with or without DR. The feature reduction further reduces the average computation time for DR severity per image from 59.54 to 3.46 s.",
keywords = "Bright lesions, classification, diabetic retinopathy (DR), fundus image processing, red lesions, segmentation, severity grade",
author = "Sohini Roychowdhury and Dara Koozekanani and Parhi, {Keshab K}",
year = "2014",
month = "9",
day = "1",
doi = "10.1109/JBHI.2013.2294635",
language = "English (US)",
volume = "18",
pages = "1717--1728",
journal = "IEEE Journal of Biomedical and Health Informatics",
issn = "2168-2194",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "5",

}

TY - JOUR

T1 - DREAM

T2 - Diabetic Retinopathy Analysis Using Machine Learning

AU - Roychowdhury, Sohini

AU - Koozekanani, Dara

AU - Parhi, Keshab K

PY - 2014/9/1

Y1 - 2014/9/1

N2 - This paper presents a computer-aided screening system (DREAM) that analyzes fundus images with varying illumination and fields of view, and generates a severity grade for diabetic retinopathy (DR) using machine learning. Classifiers such as the Gaussian Mixture model (GMM), k-nearest neighbor (kNN), support vector machine (SVM), and AdaBoost are analyzed for classifying retinopathy lesions from nonlesions. GMM and kNN classifiers are found to be the best classifiers for bright and red lesion classification, respectively. A main contribution of this paper is the reduction in the number of features used for lesion classification by feature ranking using Adaboost where 30 top features are selected out of 78. A novel two-step hierarchical classification approach is proposed where the nonlesions or false positives are rejected in the first step. In the second step, the bright lesions are classified as hard exudates and cotton wool spots, and the red lesions are classified as hemorrhages and micro-aneurysms. This lesion classification problem deals with unbalanced datasets and SVM or combination classifiers derived from SVM using the Dempster-Shafer theory are found to incur more classification error than the GMM and kNN classifiers due to the data imbalance. The DR severity grading system is tested on 1200 images from the publicly available MESSIDOR dataset. The DREAM system achieves 100% sensitivity, 53.16% specificity, and 0.904 AUC, compared to the best reported 96% sensitivity, 51% specificity, and 0.875 AUC, for classifying images as with or without DR. The feature reduction further reduces the average computation time for DR severity per image from 59.54 to 3.46 s.

AB - This paper presents a computer-aided screening system (DREAM) that analyzes fundus images with varying illumination and fields of view, and generates a severity grade for diabetic retinopathy (DR) using machine learning. Classifiers such as the Gaussian Mixture model (GMM), k-nearest neighbor (kNN), support vector machine (SVM), and AdaBoost are analyzed for classifying retinopathy lesions from nonlesions. GMM and kNN classifiers are found to be the best classifiers for bright and red lesion classification, respectively. A main contribution of this paper is the reduction in the number of features used for lesion classification by feature ranking using Adaboost where 30 top features are selected out of 78. A novel two-step hierarchical classification approach is proposed where the nonlesions or false positives are rejected in the first step. In the second step, the bright lesions are classified as hard exudates and cotton wool spots, and the red lesions are classified as hemorrhages and micro-aneurysms. This lesion classification problem deals with unbalanced datasets and SVM or combination classifiers derived from SVM using the Dempster-Shafer theory are found to incur more classification error than the GMM and kNN classifiers due to the data imbalance. The DR severity grading system is tested on 1200 images from the publicly available MESSIDOR dataset. The DREAM system achieves 100% sensitivity, 53.16% specificity, and 0.904 AUC, compared to the best reported 96% sensitivity, 51% specificity, and 0.875 AUC, for classifying images as with or without DR. The feature reduction further reduces the average computation time for DR severity per image from 59.54 to 3.46 s.

KW - Bright lesions

KW - classification

KW - diabetic retinopathy (DR)

KW - fundus image processing

KW - red lesions

KW - segmentation

KW - severity grade

UR - http://www.scopus.com/inward/record.url?scp=84927171189&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84927171189&partnerID=8YFLogxK

U2 - 10.1109/JBHI.2013.2294635

DO - 10.1109/JBHI.2013.2294635

M3 - Article

C2 - 25192577

AN - SCOPUS:84927171189

VL - 18

SP - 1717

EP - 1728

JO - IEEE Journal of Biomedical and Health Informatics

JF - IEEE Journal of Biomedical and Health Informatics

SN - 2168-2194

IS - 5

M1 - 6680633

ER -