Unsupervised Retinal Blood Vessel Segmentation Technique using pdAPSO and Difference Image Methods for Detection of Diabetic Retinopathy

Emmanuel Gbenga Dada & Stephen Bassi Joseph


Retinal vessel segmentation is a practice that has the potential of enhancing accuracy in the diagnosis and timely prevention of illnesses that are related to blood vessels. Acute damage to the retinal vessel has been identified to be the main cause of blindness and impaired vision. A timely detection and control of these illnesses can greatly decrease the number of loss of sight cases. However, the manual protocol for such detection is laborious and although autonomous methods have been recommended, the accuracy of these methods is often unreliable. We propose the utilization of the Primal-Dual Asynchronous Particle Swarm Optimisation (pdAPSO) and differential image methods in addressing the drawbacks associated with segmentation of retinal vessels in this study. The fusion of pdAPSO and differential image (which focuses on the median filter) produced a significant enhancement in the segmentation of huge and miniscule retinal vessels. In addition, the method also decreased erroneous detection near the edge of the retinal (that is not sensitive to light). The results are favourable for the median filter when compared to mean filter and Gaussian filter. The accuracy rate of 0.9559 (with a specificity of sensitivity rate of 0.9855), and a sensitivity rate of 0.7218 were obtained when tested using the Digital Retinal Images for Vessel Extraction database. The above result is a pointer that our approach will help in detecting and diagnosing the damage done to the retinal and thereby preventing loss of sight.

Full Text:



  • There are currently no refbacks.