An Automatic Detection and Identification of fundus images for Diabetic Retinopathy

Image Processing

Diabetic retinopathy refers to an eye disease which affects a human eye and leads to vision loss. This disease can be diagnosed only at the later stages, so an early detection is necessary. Early detection of DR (Diabetic Retinopathy) is thus must and can be done with the help of Image Processing. The fundus image from the dataset can be taken from the patients’ database and using the proper algorithm and pre-processing, the early detection can be made possible. For this certain parameters can be taken for the symptoms of DR as Exudates, micro-aneurysms, lesions, capillary, the fovea region, etc. With the help of these parameters, especially with the identification of neovascularization, it can also be checked that whether the DR is a proliferative DR or a Non Proliferative DR, and furthermore the severity of the DR can also be measured as Mild NonProliferative DR, i.e. Mild NPDR, Moderate NPDR or a PDR. For the detection of the Diabetic Retinopathy, the sequence of process need to be followed. For this the image must be gone through the pre-processing followed by the feature extraction. Here using the DCT (Discreet Cosine Transform) technique the feature like the Colour, Shape and Texture can be extracted. This is done on some particular region and then this is to be sent for the classification process done on the SVM (Support Vector Machine). This gives the output whether the person is having the diabetic retinopathy or not, and if it is there, then the severity is also classified.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.