Semi-automated quantification of hard exudates in colour fundus photographs diagnosed with diabetic retinopathy

Abstract

Background: Hard exudates (HEs) are the classical sign of diabetic retinopathy (DR) which is one of the leading causes of blindness, especially in developing countries. Accordingly, disease screening involves examining HEs qualitatively using fundus camera. However, for monitoring the treatment response, quantification of HEs becomes crucial and hence clinicians now seek to measure the area of HEs in the digital colour fundus (CF) photographs. Against this backdrop, we proposed an algorithm to quantify HEs using CF images and compare with previously reported technique using ImageJ.

Methods: CF photographs of 30 eyes (20 patients) with diabetic macular edema were obtained. A robust semi-automated algorithm was developed to quantify area covered by HEs. In particular, the proposed algorithm, a two pronged methodology, involved performing top-hat filtering, second order statistical filtering, and thresholding of the colour fundus images. Subsequently, two masked observers performed HEs measurements using previously reported ImageJ-based protocol and compared with those obtained through proposed method. Intra and inter-observer grading was performed for determining percentage area of HEs identified by the individual algorithm.

Results: Of the 30 subjects, 21 were males and 9 were females with a mean age of the 50.25 ± 7.80 years (range 33-66 years). The correlation between the two measurements of semi-automated and ImageJ were 0.99 and 0.99 respectively. Previously reported method detected only 0-30% of the HEs area in 9 images, 30-60% in 12 images and 60-90% in remaining images, and more than 90% in none. In contrast, proposed method, detected 60-90% of the HEs area in 13 images and 90-100% in remaining 17 images.

Conclusion: Proposed method semi-automated algorithm achieved acceptable accuracy, qualitatively and quantitatively, on a heterogeneous dataset. Further, quantitative analysis performed based on intra- and inter-observer grading showed that proposed methodology detects HEs more accurately than previously reported ImageJ-based technique. In particular, we proposed algorithm detect faint HEs also as opposed to the earlier method.

Keywords: Automated quantification; Colour fundus photography; Diabetic retinopathy; Disease management; Hard exudates; ImageJ; Macular edema.

Fig. 1

Sample colour fundus photograph diagnosed with diabetic retinopathy which depicts hard exudates alongside cotton wool spots. In this particular CF image, gray scale intensity distribution for HEs ranges from 85 to 200, for cotton wool spots from 80 to 130, for optic disc from 90 to 200 and for vessel reflections from 80 to 110

Fig. 2

Schematic of proposed methodology

Fig. 3

Graphical illustration of proposed algorithm: (a) Sample CF image; (b)--(e) images obtained after performing RGB to gray scale conversion, top-hat filtering, adaptive histogram equalization and thresholding in step-1, respectively; (f)--(i) images obtained after performing green plane extraction, top-hat filtering, adaptive histogram equalization and thresholding in step-2, respectively; (j) image after combing result of step-1 and step-2; (k) removing outliers using rectangular selection box and (l) image after removing outliers

Fig. 4

Qualitative comparison between ImageJ-based and proposed methods: Left-- Representative CF photographs of diabetic macular edema with HEs; Middle-- Detected HEs (indicated by green colour) by Sasaki’s ImageJ methodology; and Right-- Detected HEs (indicated by blue colour) by proposed algorithm

Fig. 5

Area comparison: (a) HEs area obtained by ImageJ-based and proposed method, and (b) Corresponding difference

Fig. 6

Bland-Altman plots: (a) Intra-observer repeatability of Grader A for images analyzed using ImageJ-based method, (b) Intra-observer repeatability of Grader A for images analyzed using proposed method, (c) Intra-observer repeatability of Grader B for images analyzed using ImageJ-based method, (d) Bland-Altman plot indicating repeatability of Grader B for images analyzed using proposed method, (e) Inter-observer repeatability between graders A and B for images analyzed using ImageJ-based method, and (f) Inter-observer repeatability between graders A and B for images analyzed using proposed method

Fig. 7

Statistical analysis: Distribution of percentage area of HEs obtained by (a) observer grading performed on images analysed using ImageJ-based ImageJ methodology, (b) observer grading performed on images analysed using proposed methodology, and (c) Average grading performed on images analysed using proposed methodology