11-05-2013, 04:09 PM
AUTOMATIC RETINA EXUDATES SEGMENTATION WITHOUT A MANUALLY LABELLED TRAINING SET
AUTOMATIC RETINA.docx (Size: 2.93 KB / Downloads: 17)
ABSTRACT
Retinal image analysis is commonly used for diagnosis and monitoring of diseases. In fundus photographs bright lesions representing hard and soft exudates are the earliest signs of Diabetic Retinopathy. Diabetic Retinopathy is a progressive eye disease that currently affects 250 million people worldwide. In severe situations this Diabetic Retinopathy(DR) leads to Diabetic Macular Edema(DME). This Diabetic Macular Edema can cause visionless and blindness. This occurs mainly by swelling of retina, Opthalmologists can infer the presence of the fluid that causes retina- thickening in diabetic patients by presence of accompanying lipid deposits called exudates.For detection of those exudates we are using this project.
Manual detection of exudates by opthalmologists is laborious as they have to spend great amount of time in the analysis and diagnosis of retinal photographs. Automatic screening techniques for exudate segmentation have great significance in saving cost, time and labour. Image processing techniques for exudate segmentation can help in extracting the location size and severity grade of exudates in retinal images.Basically the approaches to exudates segmentation are divided into Thresholding method, Region growing method, Morphology method, Classification method.But in our project the approach which we used falls under the category of thresholding method which do not require supervised learning step. By this we can reduce the time and also we can prevent common issues with human segmentation inconsistencies. In addition this supervised learning step required large amount of data. That's why we introduce a new way to normalize the fundus image and directly compose our method with an implementation of Morphology and Threshold based technique.