Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2016 Jun 16;7(7):2597-606.
doi: 10.1364/BOE.7.002597. eCollection 2016 Jul 1.

Automated Fine Structure Image Analysis Method for Discrimination of Diabetic Retinopathy Stage Using Conjunctival Microvasculature Images

Affiliations
Free PMC article

Automated Fine Structure Image Analysis Method for Discrimination of Diabetic Retinopathy Stage Using Conjunctival Microvasculature Images

Maziyar M Khansari et al. Biomed Opt Express. .
Free PMC article

Abstract

The conjunctiva is a densely vascularized mucus membrane covering the sclera of the eye with a unique advantage of accessibility for direct visualization and non-invasive imaging. The purpose of this study is to apply an automated quantitative method for discrimination of different stages of diabetic retinopathy (DR) using conjunctival microvasculature images. Fine structural analysis of conjunctival microvasculature images was performed by ordinary least square regression and Fisher linear discriminant analysis. Conjunctival images between groups of non-diabetic and diabetic subjects at different stages of DR were discriminated. The automated method's discriminate rates were higher than those determined by human observers. The method allowed sensitive and rapid discrimination by assessment of conjunctival microvasculature images and can be potentially useful for DR screening and monitoring.

Keywords: (080.2720) Mathematical methods (general); (100.0100) Image processing; (100.2960) Image analysis; (170.1610) Clinical applications; (170.4470) Ophthalmology.

Figures

Fig. 1
Fig. 1
An example of a cropped mosaic image of the conjunctiva of a diabetic (PDR) subject. Regions with light illumination artifacts and blur are visualized. Two 3.1 mm × 3.1 mm regions of interest (ROIs) that were selected for analysis are outlined by squares.
Fig. 2
Fig. 2
Probability densities of z - projections, and L1, L2, L3, and L4 values between non-diabetic subjects (NC, squares) and diabetic subjects (NDR, NPDR, and PDR, triangles). a) NC group 1 and NDR group 2, b) NC group 1 and NPDR group 3, c) NC group 1 and PDR group 3. Misclassified cases in group 1 have negative L1 values and misclassified cases in groups 2,3, and 4 have positive L2, L3 and L4 values, respectively. The larger L1 values and the smaller L2, L3 and L4 values denote more likely true positive and true negative cases, respectively.
Fig. 3
Fig. 3
Probability densities of z - projections, and L2, L3, and L4 values between diabetic groups. a) NDR group 2 (squares) and NPDR group 3 (triangles). Misclassified cases in groups 2 and 3 have negative L2 values and positive L3 values, respectively. b) NDR group 2 (squares) and PDR group 4 (triangles). Misclassified cases in groups 2 and 4 have negative L2 values and positive L4 values, respectively. c) NPDR group 3 (squares) and PDR group 4 (triangles). Misclassified cases in groups 3 and 4 have negative L3 values and positive L4 values, respectively. The larger L2 values denote more likely true positive cases and the smaller L3 and L4 values denote more likely true negative cases, except for comparison of groups 3 and 4, in which the larger L3 values denote more likely true positive cases.
Fig. 4
Fig. 4
Probability densities of z - projections, and L1a, L1b values between non-diabetic groups. Images in NC subjects were randomly stratified into 2 groups of equal size, group 1a (squares) and group 1b (triangles). Misclassified cases in groups 1a and 1b have negative L1a values and positive L1b values, respectively. The larger L1a values and the smaller L1b values denote more likely true positive and true negative cases, respectively.

Similar articles

See all similar articles

Cited by 6 articles

See all "Cited by" articles

LinkOut - more resources

Feedback