Research Scholar, Bharathiar University, Assistant Professor, Nehru Memorial College, Puthanampatti, Trichy, Tamilnadu, India

Transcription

1 Research Article SCIFED Publishers Saraswathi K,, 2017, 1:1 SciFed Journal of Diabetes and Endocrinology Open Access Pixel Count Method to Pigeonhole the Placid and Brutal Juncture of Non proliferative Diabetic Retinopathy *1 Saraswathi K, 2 Ganesh Babu V *1 Research Scholar, Bharathiar University, Assistant Professor, Nehru Memorial College, Puthanampatti, Trichy, Tamilnadu, India 2 Assistant Professor of Computer Science, Government College for Women, Maddur, Mandya, Karnataka, India Abstract Diabetic Retinopathy is a common complication of diabetes and the primary cause for visual impairment and blindness in adults that is caused by changes in the blood vessels of the retina. Regular screening is essential in order to detect the early stages of diabetic retinopathy for timely treatment to prevent further damage of vision. An important aspect of DR is the micro-vascular changes that cause detectable changes in the appearance of retina blood vessels. Prolongation of Non proliferative Diabetic Retinopathy may result in permanent blindness. A group of eye conditions that affect people with diabetes which causes progressive damage to the tiny blood vessels in the retina. The placid and brutal juncture are identified using Pixel Count Method. This Method scrutinizes the retina to know the number of red pixels in placid and brutal juncture of non proliferative Diabetic Retinopathy. The Microaneurysms appear as small blood clots on the surface of the retina layers. The Hemorrhages appear as large round red dots. The placid stage is indicated by the presence of at least 1 Microaneurysms on the surface of the retina layer and the brutal stage is indicated by the presence of Hemorrhages on the surface of the retina layer. The level of Microaneurysms and Hemorrhages are identified through the count of red pixels. The number of red pixels in brutal juncture is higher than placid juncture. This method pigeonholes the placid and brutal juncture of non proliferative diabetic retinopathy through the count of red pixels. Keywords Retina; Microaneurysms; Hemorrhages; Exudates; Placid; Brutal 1. Introduction The level of Microaneurysms and Hemorrhages are detected through the count of red pixels in the retina. Initially the number of red pixels is counted in normal retina without the effect of Diabetic Retinopathy. Then the same process is applied in retina with Microaneurysms and Hemorrhages. The Placid and Brutal juncture of Non proliferative Diabetic Retinopathy are detected through the count of red pixels. The retina is easily detected whether it is affected by Diabetic Retinopathy or not using the red pixels count. The existing methods detect and display the microaneurysms and hemorrhages. The pixel count method counts the number of red pixels in placid and brutal juncture of non-proliferative diabetic retinopathy and display them. So the affected area or level of microaneurysms and hemorrhages are detected through the count of red pixels in retina fundus images. So the hemorrhages are identified whenever the number of red pixels is increased. Methods The retina image is converted into green channel image. Adaptive histogram equalization technique is adopted to perform the Contrast enhancement. Then Morphological filling is performed on the green channel *Corresponding author: Saraswathi K, Research Scholar, Bharathiar University, Assistant Professor, Nehru Memorial College, Puthanampatti, Trichy, Tamilnadu, India. saraswathimuruganmsc@gmail.com Received June 19, 2017; Accepted June 28, 2017; Published August 29, 2017 Citation: Saraswathi K (2017) Pixel Count Method to Pigeonhole the Placid and Brutal Juncture of Non proliferative Diabetic Retinopathy. SF J Diabetes Endocrin1:1. Copyright: 2017 Saraswathi K. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. page 1 of 7

2 image. The unfilled green channel image is then subtracted from the filled one. The RGB color image is converted into binary image. Then the background and foreground pixels are changed to yield an image with Micro aneurysms and Hemorrhage patches. The level of Micro aneurysms and Hemorrhages are detected through the count of red pixels. 2. Diabetic Retinopathy Diabetic retinopathy (DR) is a symptom of early blindness. It causes due to high pressure on blood vessels. Pressure on eye may be origin of hemorrhages and microaneurysms, hard exudates, and cotton-wool spots. Hence to detect early lesions of diabetic retinopathy computer system has been developed using image processing [1]. Based on the damage done to the retina the disease is classified into two stages. Non-proliferative diabetic retinopathy (NPDR) is the early state of the disease in which there are no or minimal symptoms of disease. In NPDR, the blood vessels in the retina are weakened causing tiny bulges called micro-aneurysms to protrude from their walls. Proliferative diabetic retinopathy (PDR) is the more advanced form of the disease. At this stage, the retina becomes severely oxygen deprived and so new blood vessels starts forming to supply oxygen to the retina. These new vessels are abnormal so they are very fragile and tends to burst, leaking blood into the vitreous (fluid that fills the space in front of retina) clouding vision through a process called vitreous hemorrhages. Micro aneurysms, Hemorrhages and Exudates are the Characteristic features seen at different stages of Non proliferative Diabetic Retinopathy. Microaneurysms are the first sign of Diabetic Retinopathy which can be observed on the retinal fundus image [2]. As the disease progresses some of the capillaries rupture and appear as small dots or larger blots or flame-shaped hemorrhages which appear red on fundus image. Based on the extent of the presence of these features the diabetic retinopathy is classified into Mild Non-proliferative Diabetic Retinopathy, Moderate Non-proliferative Diabetic Retinopathy, and Severe Nonproliferative Diabetic Retinopathy [3]. 3. Methods of Detection 3.1 Literature Survey The Table 1 represents Literature Survey for the detection of non-proliferative diabetic retinopathy. Table 1: Literature Survey for the Detection of Non-Proliferative Diabetic Retinopathy S.No Paper Image Type Initial candidates method Classifier used Reported Performance 1. Spencer, [8] Florescence Gaussian Filter Rule-based Private dataset (4 images) 2. Cree, 1997[9] Florescence Gaussian Filter Rule-based Private dataset (20 images) 3. Hipwell,2000 [10] Colour Basic Thresholding Rule-based Private dataset (3783 images) 4. Sinthanayothin, 2002 [17] Colour Moat operator N/A Private dataset (14 images) 5. AbdelAzeem, 2002 [18] Florescence Hough transform Rule-based Private dataset (3 images) 6. Streeter, 2003 [11] Colour Gaussian filter Linear Discriminant Analysis Private dataset 7. Niemeijer, 2005 [22] Colour Gaussian Filter pixel classification K-Nearest-Neighbours Private dataset (100 images) 8. Fleming, 2006 [12] Colour Gaussian Filter K-Nearest-Neighbours Private dataset (1441 images) 9. Quellec, 2008 [15] Colour N/A N/A 10. Mizutani, 2009 [16] Colour Double-ring filter Neural network 11. Sánchez, 2009 [7] Colour Mixture model-based clustering N/A 12. Zhang, 2010 [13] Colour Multiscale Gaussian Rule-based 13. Giancardo, 2010 [5] Colour Basic Thresholding N/A 14. Lazar, 2011 [2] Colour Local Maxima scanlines N/A 15. Sopharak, 2011 [23] Colour Extended-minima Naïve Bayes Private dataset (45 images) 16. Giancardo, 2011 [6] Colour Basic Thresholding N/A page 2 of 7

3 S.No Paper Image Type Initial candidates method Classifier used Reported Performance 17. Lazar, 2013 [1] Colour Local Maxima scanlines N/A 18. Rocha, 2012 [24] Colour N/A Support Vector Machine DIARETDB1 v1 MESSIDOR 19. Sopharak, 2013 [25] Colour Extended-minima Bayesian Private dataset (80 images) 20. Akram 2013 [26] Colour Gabor filter Hybrid classifier DIARETDB0, DIARETDB1 v1 21. Li, 2013 [4] Colour Multi-orientation Gaussian (MSMF) N/A 22. Junior, 2013 [3] Colour Extended Minima N/A DIARETDB1v1 23. Inoue, 2013 [19] Colour Hessian Matrix Eigenvalues Neural network 24. Adal, 2014 [21] Colour 25. Ram, 2015 [27] Colour 26. Wu, 2015 [14] Colour Hessian Matrix Eigenvalues Morphological reconstruction Multiscale Multiorientation Gaussian (MMMF) Support Vector Machines, K-Nearest-Neighbours, Naïve Bayes, Random Forest K-Nearest-Neighbours Support Vector Machines, K-Nearest-Neighbours, Linear Discriminant Analysis DIARETDB1 v1 Private dataset 27. Romero, 2015 [28] Colour Frangi-based filters Support Vector Machines MESSIDOR+ DIARETDB1 v1 28. Srivastava, 2015 [20] Colour Hit-or-miss transform Neural networks DIARETDB1v Haloi, 2015 [29] Colour N/A Nearest-mean classifier DIARETDB1v Fundus Image Images of retina are taken by a device called fundus camera. Retinal fundus images (RFI) is the name given to these images. This camera takes images of the internal surface of retina, posterior pole, macula, optic disc, and blood vessels. Image acquisition is a leading step for medical diagnosis. Some benchmark databases are openly available for the assessment of algorithms introduced for the computerized screening and analysis of DR. The purpose of databases is to check the strength of automatic screening of DR and then compare the results with current techniques. Seven datasets are available openly including DRIVE, STARE, DIARETDB, E-ophtha, HEI-MED, Retinopathy Online Challenge (ROC), and Messidor [4]. In Figure 1 (a) represented original retina fundus image obtained from the datasets available in Internet. 3.3 Morphological Filling The original image is converted into green channel image. In Figure 1 (a) represented original image and (b) represented green channel image. The contrast of the grayscale image is enhanced by transforming the values using contrast-limited adaptive histogram equalization (CLAHE) [5]. Histogram equalization redistributes the histogram of each color channel in the input image such that the output image contains a uniform pixel value distribution. The assumption is that for each color plane the pixel rank order is maintained even with variations in illumination. A monotonic, non-linear transformation function is applied to equalize the histogram of each separate color channel [6]. An output image is produced by applying the function to map each gray level value in the input image to a new corresponding value in the output image [7]. In Figure 1(c) represented histogram equalization image. The structuring element function created STRELs of any arbitrary size and shape. Here this function created square shaped two STRELs objects. Then the morphological close operation is performed on the output image which joined the individual objects in the output image together by filling in the gaps between them and by smoothing their outer edges [8]. The holes in the output image is filled by imfill function. In Figure 1(d) represented the output image after morphological operations. page 3 of 7

4 3.4 Micro Aneurysms Image and Red Pixels Count in Placid Juncture The output image generated by imclose operation is subtracted from morphological filling image which is represented by a Figure 2(a).The subtracted image is converted into binary image through the intensity value in the subtracted image. The noise in the original image is removed through median filter and preserved edges [9]. The constant value is subtracted from original image to receive the marker image. The image after noise removal is reconstructed using marker image. The background image is retrieved through the subtraction of marker image from original image. Then the binary image is received through the intensity value of the background image. The components are connected in the image. The new binary image is received after the removal of objects that have fewer than some constant pixels (p) from old binary image. The new binary image is represented by Figure 2(b).The new binary image is subtracted from old binary image. Then the vessels are extracted from the retina by using some constant threshold value in histogram equalization image. The eroded image is received through the binary erosion. Then the each element in the eroded image is multiplied with each element in the original image. The morphological operation is performed in the output image. The image after morphological operation is converted into binary image. Micro aneurysms are the first clinical abnormality to be noticed in the eye. They may appear in isolation or in clusters as tiny, dark red spots or looking like tiny hemorrhages within the light sensitive retina. Their sizes range from microns i.e. less than 1/12th the diameter of an average optics disc and are circular in shape [10].From analysis and experiment, the pixel count for candidate micro aneurysms ranges from 30 to 5000 pixels for a (1320x1024) image. An area less than the range of 30 to 5000 pixels are regarded as a background noise. The micro aneurysms image is received by using the following statements. IM1(i,j)=double(BW1(i,j))-double(IB3(i,j))-double(I_ exudates(i,j)) level4 = graythresh (IM1) I_micro=~im2bw (IM1, level4) Then the binary image [I_micro] is converted into RGB image. In this image the microaneurysms are displayed in red color which is represented by Figure 2(c). The red pixels are counted and displayed as red pixels count[66800] in placid juncture which is represented by work space WSP.1. page 4 of 7

5 WSP. 1 Work Space for Red Pixels Count in Placid Juncture 3.5 Hemorrhages Image and Red Pixels Count in Brutal Juncture The original image is converted into green channel image. In Figure 3 (a) represented original image with hemorrhages. In Figure 3 (b) represented green channel image. The holes in the output image are filled by infill function. In Figure 3 (c) represented the output image after morphological operations. The green channel image is subtracted from morphological filling image which is represented by a Figure 2(d).The subtracted image is converted into binary image through the intensity value in the subtracted image. The binary image is represented by a Figure 3 (e). Then the binary image is converted into RGB image. In this image the hemorrhages are displayed in red color which is represented by Figure 3 (f).the red pixels are counted and displayed as red Pixel count [138290] in brutal juncture which is represented by work space WSP Results Figure 4 compares the results obtained by the Pixel Count Method. 5. Conclusion An eye disease like Diabetic retinopathy (DR) is responsible for blindness in human eye. Therefore it is necessary to detect such diseases at early stage with the help of image processing technologies and methods. Pixel Count Method is able to pigeonhole the placid and brutal juncture of non proliferative diabetic retinopathy. So that ophthalmologist can detect the level of damaged blood vessels due to pressure in eye. Here the retina images are taken only from Internet datasets such as DRIVE and STARE. In this paper we discuss about the variation of red pixels count in placid and brutal juncture of Non proliferative diabetic retinopathy. page 5 of 7

6 WSP. 2 Work Space for Red Pixels Count in Brutal Juncture Figure 4: Red Pixels Count in Placid and Brutal Juncture page 6 of 7

7 6. Acknowledgement I am thankful to Dr. V. Ganesh Babu (Assistant Professor in Computer Science, Government College for Women, Maddur, Mandya, Karnataka) for his excellent guidance and K. Vasantha (Mother), N. Karuppannan (Father) and T. Murugan (Spouse) for supporting me in all the ways. References 1. Patil JD, Chaudhari LC (2012) Tool for the Detection of Diabetic Retinopathy using Image Enhancement Method in DIP. International Journal of Applied Information Systems 3: Tarannum ZA, Srilatha B (2015) Detection of Diabetic Retinopathy with Feature Extraction using Image Processing. International Journal of Electrical Electronics and Computer Systems 3: Zhang X (2014) Image processing methods for computeraided screening of diabetic retinopathy. Other Ecole Nationale Superieure des Mines de Paris. Citation: Saraswathi K (2017) Pixel Count Method to Pigeonhole the Placid and Brutal Juncture of Non proliferative Diabetic Retinopathy. SF J Diabetes 2. Thakkar F, Parikh R (2016) A Survey on Automatic Detection of Diabetic Retinopathy Exudates from Retinal Fundus Images. International Journal of Advanced Research in Computer and Communication Engineering 5: Aniruddha L, Prabhu S, Sampathila N (2015) Detection of Abnormal Features in Digital Fundus Image Using Morphological Approach for Classification of Diabetic Retinopathy. International Journal of Innovative Research in Computer and Communication Engineering 3: Amin J, Sharif M, Yasmin M (2016) A Review on Recent Developments for Detection of Diabetic Retinopathy. Scientifica (Cairo). 5. Valverde C, Garcia M, Hornero R, et al. (2016) Automated detection of diabetic retinopathy in retinal images. Indian Journal of Ophthalmol 64: Datta NS, Sarker R, Dutta HS, et al. (2012) Software based Automated Early Detection of Diabetic Retinopathy on Non Dilated Retinal Image through Mathematical Morphological Process. International Journal of Computer Applications 60: Prentasic P, Loncaric S (2016) Detection of exudates in fundus photograph using deep neural networks and anatomical landmarks detection fusion. Computer Methods and Programs in Biomedicine 137: Jadhav AS, Patil BP (2015) Classification of Diabetes Retina Images Using Blood Vessel Area. International Journal of Cybernetics and Informatics 4: Jagannath M, Adalarasu K (2015) Diagnosis of Diabetic Retinopathy From Fundus Image Using Fuzzy C-Means Clustering Algorithm. IIOAB 6: 3-9. page 7 of 7