International Journal of Advanced Engineering Research and Studies

E-ISSN22498974

Research Article

DETECTION OF EXUDATES ON DIABETIC RETINOPATHY IMAGES BASED ON MORPHOLOGICAL OPERATION AND CONNECTED COMPONENT ANALYSIS
M.PonniBalaa*, S.Mohanapriyab, Dr.S.Vijayachitrac Assistant professor (Sr.G), Department of EIE, Kongu Engineering College, Perundurai-638052 b P.G Student M.E (Control and Instrumentation), Kongu Engineering College, Perundurai-638052 c Professor, Department of EIE, Kongu Engineering College, Perundurai-638052 ABSTRACT
Diabetic Retinopathy is a major cause for blindness, identified early by the formation of exudates in the retina. The conventional method followed by opthalmogists is the regular supervision of the retina. As this method takes time and energy of the opthalmogists, a new feature based classification for the detection of exudates in color fundus image is proposed in this paper. This method reduces the professionals work to examine on every fundus image rather than only on abnormal image. The exudates are separated from the fundus image by thresholding and removal of optic disk using morphological operation and connected component analysis. The features are extracted from processed image and used for classification of images as exudates and non-exudates. KEYWORDS Exudates, Fundus image, Connected component, Morphological operation
a*

Address for Correspondence

I. INTRODUCTION Diabetic Retinopathy (DR) is the eye related disease caused by the diabetes. DR is the most common sight threatening disease when untreated leads to vision loss and in many cases it cannot be reversed [1]. There are two types of DR: Non-Proliferative diabetic retinopathy (NPDR) and Proliferative diabetic retinopathy (PDR).In the first case, the damaged blood vessels leak extra fluid and small amount of blood into the eye. This condition leads to the formation of exudates in the retina. As the disease progresses the amount of exudates also increases. In the second case, the blood vessels in the retina closes and prevents blood flow in the eye. This condition leads to the emerge of new blood vessels in order to supply blood to the blocked area and this condition is called as neovascularization [2].

Fig 1 Normal Retinal Image

Fig 2 Retinal image with Exudates Fig 1 shows the normal retinal image and Fig 2 shows the retinal image with exudates. Exudates are manifested as spatially random yellowish or whitish patches of varying sizes, shapes and locations. These are the visible sign of DR and a major cause of visual loss in Non-Proliferative forms of DR. In this work, the retinal images are classified as exudates and nonexudates using statistical features extracted from the images. The detection of exudates in retinal images was investigated by many researchers. IJAERS/Vol. I/ Issue II/January-March, 2012/86-88

In [3], global and local thresholding values are used to segment exudate lesions. Before thresholding, the images were preprocessed to eliminate photographic non-uniformities, and the contrast of the exudates was then enhanced. Ege et al. [4] located exudates using template matching, region growing, and thresholding techniques. A Bayesian classifier was then used for classification of images into exudates and nonexudates. Sinthanayothin [5] applied a recursive region growing technique using selected threshold values in graylevel images. In this work, the processed retinal image includes only exudates and no other bright lesions were considered. In [6] statistical features are extracted and classified using Bayes Minimum Distant discriminant (MDD) classifier. The classifier is compared with original and brightness enhanced image. The brightness enhanced image gives 100% accuracy where as 66% for normal image. Walter et al. [7] detects exudates as well as optic disc in retinal images. Optic disk is removed by morphological filtering and watershed transformation and for exudates detection morphological operation and thresholding technique is used. Choosing of correct parameters like window size and thresholding is stated as crucial part in this work. Osareh et al [8] segmented images using Fuzzy CMeans (FCM) clustering technique. The features are extracted and ranked using Genetic Algorithm and classified using neural network. Doaa Youssef et al [9] proposed green channel have high contrast and necessary information (exudates). Canny edge detector and Hough transform is used for optic disk elimination and morphological operation for detection of blood vessel tree in this work. II. MATERIALS AND METHODS A. Methodology The objective of this project work is to classify the retinal image into exudates and non-exudates (normal). The retinal image is subjected to the preprocessing steps and is morphologically operated by a disk shape structuring element. The optic disk is eliminated by connected component analysis and features like exudates area, size, color, and

International Journal of Advanced Engineering Research and Studies homogeneity and texture properties are extracted. The flowchart of the proposed method is shown in Fig 3.

E-ISSN22498974

Fig 3 Flowchart of the proposed methodology B. Image acquisition The images for this work are taken from DIARETDB0 database [10]. This database consists of 130 color fundus images of which 20 are normal and 110 contain various signs of the diabetic retinopathy includes red small dots, haemorrhages, hard exudates, soft exudates and neovascularisation. In addition to the fundus images, the database also has information about the medical findings (i.e ground truth). C. Pre-processing Pre-processing is the initial step in all the case of image related diagnosis system. In case of Diabetic Retinopathy [11], the retinal images in the dataset are often noisy and poorly illuminated because of unknown noise and camera settings. Also the color of retina has wide variation from patient to patient. Thus to remove noise and undesired region the images are subjected to preprocessing steps, which include green channel extraction, histogram equalization and contrast enhancement. The exudates appear bright in the green channel compared to red and blue channels in RGB image. Hence green channel is used for further processing by neglecting other two components. Histogram equalization and contrast enhancement are used to increase the contrast between the exudates and the image background. D. Optic disk elimination The enhanced retinal image is converted to binary image by applying proper thresholding value. This binary image is subjected to morphological operations (i.e. opening and closing) [12]. Closing operation is defined as dilation followed by erosion and opening is the reverse of closing operation. Dilation is an operation that grows or thickens objects in a binary image. Erosion shrinks or thins the objects in the binary image. The process of thickening and thinning is controlled by a shape called structuring element. As the optic disk and exudates are circular in shape, a disk shape structuring element is used in this work. The mathematical equations governing morphological operations are: .. Opening. .. Closing. .. Erosion. .. Dilation. Where f is a gray level image and sB is structuring element B of size s [13][14]. IJAERS/Vol. I/ Issue II/January-March, 2012/86-88

Because of the similar attributes of exudates and optic disk, the morphologically operated image has both exudates and optic disk. The optic disk is removed by connected component analysis. Connected component analysis [15] is an image segmentation technique which groups an image pixels into components based on pixel connectivity, i.e. all pixels in a connected component share similar pixel intensity values and are in some way connected with each other. It is defined as a region R (subset of S) is said to be connected under c(s) if for all s, r R there exists a sequence of M pixels, s1, , sM such that s1 c(s), s2 c(s1), , sM c(sM1), r c(sM) i.e. there is a connected path from s to r. Since the optic disk occupies maximum area in the image, by using connected component properties the maximum area is eliminated. E. Feature extraction The processed image after the removal of maximum area (optic disk) has only exudates. This image is used for feature extraction. The statistical features like exudates area, entropy, correlation, energy, contrast, homogeneity, standard deviation, mean, skewness and kurtosis are extracted from the image. From this features, the most effective features are used for classification. III. RESULTS AND DISCUSSION The retinal image database contains images of various diabetic retinopathy signs (red small dots, hemorrhages, hard exudates, soft exudates and neovascularisation). Thus a set of images (15 normal and 15 abnormal) were chosen and analyzed for the classification.
Input RGB Image PREPROCESSED IMAGE 100 100 200 200

300

300

400

400

500

500

600

600

700

700

800

800

900

900

1000

1000

1100

1100

200

400

600

800

1000

1200

1400

200

400

600

800

1000

1200

1400

(a)
IMAGE WITH OPTIC DISK 100 100 200 200 300 300

(b)
OPTIC DISK ELIMINATED IMAGE

400

400

500

500

600

600

700

700

800

800

900

900

1000

1000

1100

1100

200

400

600

800

1000

1200

1400

200

400

600

800

1000

1200

1400

(c) (d) Fig 4 Steps involved in Exudates detection for normal image. (a) Original image (b) Preprocessed image (c) Image with Optic Disk (d) Optic disk removed image The images are processed and features are extracted. The processing steps for two images (normal and abnormal) are shown in the Fig 4 and Fig 5 respectively. The images presented in Fig 4 and Fig 5 is taken from database DIARETDB0. From the processed image the various statistical features are extracted. The important finding of this result is that there are distinct differences in the statistical features such as Exudates area, Entropy,

International Journal of Advanced Engineering Research and Studies Contrast, Homogeneity and Kurtosis for normal and diabetic retinopathy images.
Input RGB Image PREPROCESSED IMAGE 100 100

E-ISSN22498974

7.

200

200

300

300

400

400

500

500

600

600

8.

700

700

800

800

900

900

1000

1000

1100

1100

200

400

600

800

1000

1200

1400

200

400

600

800

1000

1200

1400

9.

(a)
IMAGE WITH OPTIC DISK 100

(b)
OPTIC DISK ELIMINATED IMAGE 100

200

200

10.

300

300

400

400

500

500

600

600

11.

700

700

800

800

900

900

1000

1000

1100

1100
200 400 600 800 1000 1200 1400

200

400

600

800

1000

1200

1400

(c) (d) Fig 5 Steps involved in Exudates detection for abnormal image. (a) Original image (b) Preprocessed image (c) Image with Optic Disk and Exudates (d) Optic disk removed image The comparative difference in the feature value for the normal and Diabetic Retinopathy images is shown in the Table 1. Table 1 Comparative difference between normal and DR images
Category No.of subjects Exudates Area Entropy Contrast Homogeneity Kurtosis Normal 15 <1000 <0.03 1E-004 ~1 Positive Diabetic Retinopathy 15 >1000 >0.03 1E-005 <0.9998 Negative

12.

13. 14.

15.

2000. Proceedings. IEEE Conference on, 2000, vol. 2, pp. 181-186 vol. 2. T. Walter, J. C. Klein, P. Massin, and A. Erginay, A contribution of image processing to the diagnosis of diabetic retinopathy-detection of exudates in color fundus images of the human retina, Medical Imaging, IEEE Transactions on, vol. 21, no. 10, pp. 1236-1243, 2002. A. Osareh, B. Shadgar, and R. Markham, A computational-intelligence-based approach for detection of exudates in diabetic retinopathy images, Information Technology in Biomedicine, IEEE Transactions on, vol. 13, no. 4, pp. 535-545, 2009. D. Youssef, N. Solouma, A. El-dib, M. Mabrouk, and A. B. Youssef, New feature-based detection of blood vessels and exudates in color fundus images, in Image Processing Theory Tools and Applications (IPTA), 2010 2nd International Conference on, 2010, pp. 294-299. T. Kauppi et al., Diaretdb0: Evaluation database and methodology for diabetic retinopathy algorithms, Citeseer. M. Goldbaum et al., Automated diagnosis and image understanding with object extraction, object classification, and inferencing in retinal images, in Image Processing, 1996. Proceedings., International Conference on, 1996, vol. 3, pp. 695-698 vol. 3. R. C. Gonzalez and R. E. Woods, Digital image processing. Prentice Hall Upper Saddle River, NJ:, 2002. P. Soille, Morphological image analysis: principles and applications. Springer-Verlag New York, Inc., 2003. J. Serra, Image analysis and mathematical morphology. London.: Academic Press.[Review by Fensen, EB in: J. Microsc. 131 (1983) 258.] Review article General article, Technique Microscopy Staining, Mathematics, Cell size (PMBD, 185707888), 1982. C. A. Bouman, Connected component analysis, Digital Image Processing, Course Notes, 2007.

IV. CONCLUSION The main idea of this project is to classify retinal images as exudates and non-exudates using statistical features. The retinal image is subjected to various processing and features are extracted for the image. From the calculated feature values, a retinal image can be classified under which category the image falls on (i.e. either exudates or non-exudates) manually. This work can be automated using an artificial intelligence technique (a classifier) in future. REFERENCES
1. M. Rema and R. Pradeepa, Diabetic retinopathy: an Indian perspective, INDIAN JOURNAL OF MEDICAL RESEARCH, vol. 125, no. 3, p. 297, 2007. R. Sivakumar, G. Ravindran, M. Muthayya, S. Lakshminarayanan, and C. U. Velmurughendran, Diabetic retinopathy classification, in TENCON 2003. Conference on Convergent Technologies for AsiaPacific Region, 2003, vol. 1, pp. 205-208 Vol. 1. R. Phillips, J. Forrester, and P. Sharp, Automated detection and quantification of retinal exudates, Graefes archive for clinical and experimental ophthalmology, vol. 231, no. 2, pp. 90-94, 1993. B. M. Ege et al., Screening for diabetic retinopathy using computer based image analysis and statistical classification, Computer Methods and Programs in Biomedicine, vol. 62, no. 3, pp. 165-175, 2000. C. Sinthanayothin, Image analysis for automatic diagnosis of diabetic retinopathy., 1999. H. Wang, W. Hsu, K. G. Goh, and M. L. Lee, An effective approach to detect lesions in color retinal images, in Computer Vision and Pattern Recognition,

2.

3.

4.

5. 6.

IJAERS/Vol. I/ Issue II/January-March, 2012/86-88