ABSTRACT
Aim
With an increasing percentage of retinal pathology because of diabetes, eye screening for diabetic retinopathy (DR) is in demand throughout the world. Haemorrhages (HEs) are one of the common signs for the red lesion detection for early diagnosis of this progressive degenerative disease of the retina. The detection of HEs in early stage prevents further progression of the eye disease and reduces the risk of blindness. Methods: The proposed method is based on morphological segmentation and geometrical feature techniques for HEs extraction. This approach uses preprocessing, removal of other retinal image details, determining connected components analysis and applying a specific shape feature set which results in improving the recognition of HEs. Results: The proposed algorithm demonstrated 95.47% accuracy for a Diaretdb1 database at image level detection. Moreover, the proposed method achieved better performance results for HEs extraction when individual images were analysed and compared with the true HEs count according to the multiple expert evaluations. Conclusion: The results obtained prove the potential use of the proposed algorithm in terms of true HEs count to facilitate the DR grading criteria related to HEs.
Disclosure statement
No potential conflict of interest was reported by the authors.
Additional information
Notes on contributors
Shilpa Joshi
Shilpa Joshi does research in Biomedical Engineering and on Diabetic Retinopathy.
P. T. Karule
Dr. P. T. Karule currently works as Professor and Head in Department of Electronics Engineering, Yashwantrao Chavan college of Engineering. Nagpur.