Early Detection of Glaucoma in Retinal Images Using Cup to Disc Ratio S.Kavitha1, S.karthikeyan2, Dr.K.Duraiswamy3 1 Asst.Prof.Nandha Engineering College, Erode, 2 Professor, K.S.R.College of Engineering, Tiruchengode. 3 Dean, K.S.R.College of Technology, Tiruchengode. email:email@example.com Abstract- Glaucoma is a disease characterized by elevated Optical Coherence Tomography (OCT) and Heidelberg intraocular pressure (IOP). This increased IOP leads to Retinal Tomography (HRT) are very expensive [1, 2] damage of optic nerve axons at the back of the eye, with compared to digital fundus images. With the help of image eventual deterioration of vision. CDR is a key indicator for processing, the features of the fundus images such as optic the detection of glaucoma. The existing approaches disc and cup could be localized to suspect the glaucoma [4, 5]. determined the CDR using manual threshold analysis The detection of OD position is a pre-requisite for the which is fairly time consuming. This paper proposes two computation of some important diagnostic indexes like methods to extract the disc automatically. The component glaucoma. The optic disk is the region on the retina at which analysis method and region of interest (ROI) based optic nerve axons enter and leave the eye. The optic disc and segmentation is used for the detection of disc. For the cup, cup were located by identifying the area with the highest component analysis method is used. Later the active variation in intensity of adjacent pixels . This paper contour is used to plot the boundary accurately. This presents the automatic detection of optic disc by region of method has been tested on numerous image data sets from interest based segmentation and component labeling. As the Madurai Eye Care Centre, Coimbatore. shape of the optic disc is important to diagnose eye diseases, the exact boundary detection of the optic disc is investigated. “Active Contour” was applied to detect the exact contour of Index terms - Glaucoma, fundus, Region of Interest (ROI), optic disc. The major advantage of this method is their ability Component Analysis. to bridge the discontinuities in the image features being located. The contour was plotted by calculating the centroid of the image feature. Later the optic cup has been detected by I .INTRODUCTION component analysis method. Then the number of white pixels was calculated to find the area of disc and cup from the result Glaucoma is defined as „multi factorial optic neuropathy‟ obtained by the above mentioned methods. The estimation of which is a potentially blinding disease which affects 66.8 boundary was achieved by active contour. With the help of million people worldwide. It is the second leading cause of detected area of disc and cup the CDR was calculated to blindness. Risk assessment of the disease goes a long way in suspect the glaucoma. Finally clinical CDR is compared with diagnosis and management of the disease. Your eye has the proposed three methods. pressure just like your blood and when this intra ocular pressure (IOP) increases to dangerous level, it affects the optic II. METHODOLOGY nerve. Although the raised IOP is a significant risk factor for developing glaucoma, there is no set threshold for IOP that A. Extraction of Optic Disc causes glaucoma. This can result in decreased peripheral vision and eventually leads to blindness. Glaucoma is similar 1. Manual Threshold Analysis: The formulation of the Manual to ocular hypertension, but with accompanying optic nerve threshold for extracting the optic disc includes the removal damage. Vision loss is caused by damage to the optic nerve, of the blood vessels in the retinal images. The which carries image information from the light receptors to the morphological operation such as the dilation, erosion, is brain. Untreated glaucoma leads to permanent damage to the performed. The morphological functions are applied to do optic nerve and result in visual field loss. Early detection and the pre-processing. The dilation and erosion of A and B is prevention is the only way to avoid total loss of vision. defined by, Automated screening for glaucoma associated eye pathology is of benefit for several reasons. The most important of these A∙B = ( A B) ΘB (1) is that pathology can be detected at the asymptomatic stage of disease progression and in amendable to treatment with good Dilation causes objects to grow in size by adding pixels to the outcomes. The ailment is physiologically described as the boundaries of the object in the image. Erosion is done to degeneration of optic nerve cells with associated visual field contrast the boundary of the object. The result of this effects. The suspection of glaucoma makes easier by means of operation has smooth image without any blood vessels. computer screening . The detection of glaucoma through Individual pixels in a grayscale image are marked as „object‟ pixels if their value is greater than some threshold value B. Extraction of Optic Cup (assuming an object to be brighter than the background) and as Component Analysis Method: The segmentation of optic cup „background‟ pixels otherwise. Typically, an object pixel is is considerably more challenging than the optic disc due to given a value of „1‟ while a background pixel is given a value high density of vascular architecture traversing the cup of „0‟. The minimum threshold value of the disc is set in order boundary. The disc and cup could not be easily distinguished to extract the optic disc boundary. The boundary is estimated as the border between the two was unclear. The component by plotting the boundary over the input image as in figure().In analysis method will localize the optic cup efficiently, even most conventional level set method, undesired effects also though the image is low contrast. In this paper, the green known as shocks can develop during the evolution of the level component from the original fundus image is extracted in set function, resulting in inaccurately detected contours. To order to detect the cup . The morphological operation like avoid this, periodic re-initialization of the level set function is closing and opening are performed to get the area of cup more necessary but may not be an optimum solution. accurately . The close operation would fill the gap of the cup and also smoothen its boundary. The open operation 2. Color Component Analysis: In the component analysis would remove any small stray bright spots that are present in method, RGB components are analyzed and it was found that the image . The component analysis method detects the the optic disc was more easily discriminated in RED image area of the cup more accurately than the manual threshold . In order to measure disc more accurately, the blood analysis. Later the result of the morphological image is vessels in the image had to be removed. This is achieved by converted to binary image. The exact optic cup was obtained performing the closing and opening operation. Closing is from the binary image. By calculating the number of white similar to dilation and it tends to enlarge the boundaries of pixels in the binary image obtained, the area of cup was foreground regions in an image and shrink background color measured. holes in such regions. Opening is similar erosion and it tends to remove some of the foreground pixels from the edge of region of foreground pixels . First CLOSE operation is III. DETERMINATION OF CDR done followed by the OPEN . The boundary of the optic disc is detected by converting the opening image into the There are many features for identifying the vision impaired binary image in which white pixels closely approximate the disease glaucoma. The first and important feature is the cup to edge of the optic disc. This method accurately determines the disc ratio, which specifies the change in the cup area. Due to boundary of the optic disc only when the image has the glaucoma the cup area will increase slowly by intra ocular contrast disc. If the disc boundary is blurred then it fails to pressure (IOP) and results in dramatic visual loss. Increase in extract the boundary exactly. cup area is analyzed by examining the CDR value. The CDR was calculated by taking the ratio between the area of optic 3.ROI based segmentation: The optic disc is extracted by cup and disc. If the CDR > 0.3 indicates the suspection of finding the region of interest based on colors. In order to glaucoma. If the CDR< 0.3 , it is considered as normal image. calculate the ROI, in the original image the mathematical morphology like dilation, erosion is done to smoothen the image. After performing the morphological functions the small holes gets filled and object boundary is smoothen. The ROI for the input image is then examined; the result of this step gives the binary image of the optic disc. To localize the boundary exactly the component labeling is used. In this image areas are label by using the neighborhood connecting pixels. All the connected pixels with the same input value are assigned the same identification label. Figure (1) shows the component labeling algorithm. (a) (b) Fig 2.suspection of glaucoma (a) CDR<0.3 (b) CDR>0.3 (a) (b) Fig.1. Result from the component labeling (a) Before labeling (b) After labeling. IV. EXPERIMENTAL RESULTS component analysis for the cup. In third and last row the The first row presents the input image and the next row cup was same, because in both component analysis method shows the results of manual threshold analysis. The third was used. Table 1 compares the CDR value obtained from row presents simulation results of component analysis for manual threshold analysis, component analysis and finally cup and disc. The last row gives the result obtained from with ROI based segmentation for disc and component ROI based segmentation method for disc and analysis method for cup. Fig 3.First row: Input images ; Second row: Result from manual threshold; Third row: Result from component analysis; Fourth row: Result from ROI method. The table relates the clinical CDR with proposed CDR value. It can be obviously seen that the ROI and component analysis ROI and Component Analysis method provides better result when compared with the other two methods. Then graph is plotted for the clinical CDR with 0.8 the detected CDR for the three methods. 0.7 Cup to Disc Ratio 0.6 Table 1: Comparison of clinical CDR value with several methods 0.5 cdr.roi MANUAL COMPONENT ROI & 0.4 CLINICAL THRESHOLD ANALYSIS COMPONENT IMAGES CDR CDR CDR CDR 0.3 cdr.clinical 0.2 G1 0.35 0.4616 0.4733 0.3174 G2 0.35 0.41668 0.3267 0.3251 0.1 G3 0.25 0.2689 0.2584 0.2216 0 G4 0.5 0.4013 0.7523 0.4956 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 G5 0.35 0.6794 0.4236 0.3331 G6 0.35 0.6041 0.5011 0.3712 Images G7 0.35 0.5136 0.4389 0.3543 G8 0.4 0.5887 0.4889 0.3627 G9 0.3 0.601 0.3006 0.3189 G10 0.55 0.5892 0.5399 0.506 Fig.6. Comparison of clinical CDR and CDR from G11 0.7 0.5679 0.6912 0.6104 ROI method G12 0.45 0.5817 0.5297 0.4877 G13 0.35 0.512 0.4188 0.3272 G14 0.35 0.4689 0.4298 0.3638 IV. POSTULATION G15 0.2 0.245 0.244 0.2073 This work proposes the recognition of Glaucoma in advance. There are several methods of treatment available to impede progression of this disease. It is important to diagnose Manual T hreshold Glaucoma as early as possible to minimize the damage to optic 0.8 nerve head. The detection of Glaucoma by Optical Coherence 0.7 Tomography (OCT) and Heidelberg Retinal Tomography Cup to Disc Ratio 0.6 (HRT) was used to differentiate between the glaucoma and 0.5 cdr.thres non-glaucoma eye using neural network is very expensive. But 0.4 in this approach, the funds photograph which is less expensive 0.3 cdr.clinical is used. The Promising result suggests good potential for use of 0.2 ROI and component analysis in detection of Glaucoma in early 0.1 0 stage. The accuracy of classification also depends on the 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 environment lighting conditions. Hence this method provides Images an additional tool for physician to verify their diagnosis. V.CONCLUSION Fig.4. Comparison of clinical CDR and CDR from The algorithms for the Advance identification of manual method Glaucoma by estimating CDR were developed in this paper. ROI based segmentation is proposed to localize optic disk, Component Analysis which is estimated by using contour method, exactly when compared with other methods even though the image is in low 0.8 contrast. The optic cup was segmented using the component 0.7 analysis and the threshold methods separately. The Cup to Disc Ratio 0.6 performance of various methods was evaluated using the 0.5 cdr.com proximity of the calculated CDR to the clinical CDR. It was 0.4 0.3 cdr.clinical found that ROI, combined with the component analysis method 0.2 provides the better estimation of CDR. The implementation of 0.1 the above said method would be more fruitful with the 0 availability of more suitable data. The Algorithm proposed in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 this paper has high value in clinic practice for automatic Images screening of early diagnosis of Glaucoma. This proposed method can be used as an adjunct tool by the physicians to Fig.5. Comparison of clinical CDR and CDR from cross check their diagnosis. component analysis method ACKNOWLEDGMENT  N.Inoue, K.Yanishima, K. Magatani, and T.A.K.T Kurihara, “Development of a simple diagnostic method for The authors would like to thank Dr.Srinivasan from the glaucoma using ocular Fundus pictures,” in Madurai Eye Care Hospital, Coimbatore for his suggestions proceedings of the 2005 IEEE Engineering in Medicine and for supplying the fundus Photographs for our tests. and Biology 27th annual international conference , pp, 3355 -3358, sep 2005.  H. Li and O.Chutatape, “A model – based approach for REFERENCES automated feature extraction in fundus images,” in proceedings of the Ninth IEEE International conference  Acharya, U. R., Ng, E.Y.K., and Suri, J. S., “Image on computer vision ,vol .1, pp 394-399,oct 2003. modeling of human Eye” Artech House, MA, USA, 2008a,  M. Kass, A. Witkin and D. Terzopoulos, “Snakes: Active April. contour models,” Int. J. Computer vision, vol. 1, pp 321-  U.R., Chau, K. C., Ng, E. Y. K., Wei, W., and Chee, C., 331, 1988. “Application of higher order spectra for the identification  Jagadish Nayak. Rajendra Acharya U. P.Subbanna Bhat. of diabetes retinopathy stages”. J. Med syst. USA. 2008b. Nakkul Shetty. Teik –Cheng Lim, “Automated Diagnosis doi: 10.1007/s10916-008-9154-8. of Glaucoma Using digital Fundus Images,” June 2008.  Song, X., Chen, Y., Song, K., and Chen, Y., “A Computer  Walter, T., Klein, J. C., Massin, P., and Erginay,A.,“A – based diagnosis of glaucoma using an artificial neural Contribution of image processing to diagnosis of diabetic network”. Proceedings of 17th Annual Conference IEEE retinopathy – detection of exudates in color fundus images Engineering in Medicine and Biology, 1, 847-848, 1995. of the human retina”. IEEE trans. Med. Imaging.21:1236-  Viranee Thongnuch, Bunyarit Uyyanonvara, “Automatic 1243,2002. doi: 10.1109/TMI.2002.806290. optic disc detection from low contrast retinal images of  J. Liu, D. W .K Wong, J. H. Lim, X. Jia, F. Yin, H. Li, W. ROP infant using mathematical morphology”, 2000. Xiong, T.Y. Wong, “Optic Cup and Disc extraction from  Nayak, J. Bhat, P.S., Acharya, U. R., Lim, C.M., and Retinal Fundus Images for Determination of Cup- to- Disc Kagathi, M., “Automated identification of different stages Ratio”, in proceedings of 2008 IEEE Engineering pp 978- of diabetic retinopathy using digital fundus images”. J. 1-4244-1718-6/08/ . Med. Sys.USA. 32:2107-115, 2008, doi:10.1007/s10917-  Giri Babu Kande, T.Sathya Savthri, P.Venkata Subbaiah, 007-9113-9. M.R.N Tagore, “Automatic Detection and Boundary  Wong, L.Y., Acharya, U.R., Venkatesh, Y.V., Chee, Estimation of Optic Disc in Fundus Images using C.Lim, C.M., and Ng, E. Y. K., “Identification of different Geometric Active Contours,” Feb 2009. stages of diabetic retinopathy using retinal optical images.” Inf.sci. May 2008.  J. A. Sethian, “Level Set Methods”, Cambridge Monographs on Applied and Computational Mathematics”, Cambridge, 1996.