Last modified: 2018-07-07
Abstract
Diabetic retinopathy is a disease caused by complication of diabetes, which causes damage to the retina and can result in blindness. It could be divided into two clinical groups: proliferative diabetic retinopathy and non-proliferative diabetic retinopathy, the earliest clinical symptom found in diabetic retinopathy. One of the early sign of this disease is the appearance of microaneurysm which is a swelling or bulge in blood vessels seen as small, red dots on the retina. Early detection of diabetic retinopathy is very importance. An automated system to detect red small dots in retinal fundus image is expected to avoid further damage to the retina. Before detecting red small dots, blood vessel need to be removed because it has similar intensity value with red small dots. Previous researches on detection of red small dots in retinal fundus images are perform using various approaches. This paper proposed the combination of Tyler Coye and morphological supremum of openings algorithm to enhance blood vessel segmentation for red small dots detection.
Red Small Dots segmentation consist of three parts, the first part is the process to segment the dark area of retinal fundus image after eliminating bright regions including optic disc and exudates. The second part is the process to segment the blood vessel, the third part is to obtain the red small dots segmentation by subtract the result of the first and the second part. In dark area segmentation part, retinal fundus image resizes to 480 x 640 pixels then convert to green channel. Find the image regional minima and multiply with the green channel. Apply morphological reconstruction with the marker from the result of multiply image regional minima with green channel and the green channel image as the mask then enhance the image. Threshold the result to get the dark area of the image.
The second part is Blood Vessel Segmentation. In this part we use the combination of Tyler Coye algorithm and morphological supremum of openings. Coye applies Principal Component Analysis (PCA) with Lab color model for converting the retinal color image into grayscale. The next step is enhance the contrast of an image by applying Contrast-Limited Adaptive Histogram Equalization (CLAHE) then eliminate the background by subtracting the average filtered image. The blood vessel is detect using supremum of openings with linear structuring element at different orientation. Then apply reconstruction by dilation to the image with the output of subtracting the background with the average filtered image as the marker and the result of supremum of openings as the mask. Finally, the binary of the blood vessel is obtained by apply the threshold operation. The last part is subtract the binary image of the blood vessel from the output of dark area segmentation to get the red small dots segmentation.
The proposed method trained using the retinal database from DIARETDB1. The segmentation result is evaluated by measure the accuracy, sensitivity, and specificity and compare with the segmentation result using Tyler Coye algorithm only. Based on the experimental result, it can be concluded that the proposed method gives more optimal red small dots segmentation with the sensitivity of 62.45% compared to the Tyler Coye algorithm with sensitivity 36.59%.