A Review on Detection of Diabetic Retinopathy Using Fundus Image

A Review on Detection of Diabetic Retinopathy Using Fundus Image

Pradeepa N

Assistant Professor, Department of ECE SNS College of Technology Coimbatore, Tamilnadu, India

Chenthur Pandian S

Principal

SNS College of Technology Coimbatore, Tamilnadu, India

Abstract- Over the foremost recent decade, there has been development in imaging modalities accessible to clinicians so as to diagnose and observe the behaviour and development of diabetic retinopathy. In recent times, progress in image technologies associated with OCT and OCT roentgenography have enabled higher plan and understanding of the malady. During this paper, the utilization of detection of diabetic retinopathy like feature extraction, detection in blurred image, detection victimization retinal grading algorithmic program are delineated and techniques in adaptive optics and hyperspectral imaging within the designation and management of diabetic retinopathy were conjointly mentioned.

Key words- Diabetic macular oedema, diabetic retinopathy, retinal grading algorithm, microaneurysms.

1. INTRODUCTION

   Diabetic retinopathy (DR) is the most significant cause of vision defect within the working-age population of the urban world and is calculable to have an effect on over ninety three million folks. The grading procedure consists of recognising very minute DETAILS, like microaneurysms, to some large options like exudates and from time to time their spot relative to every alternative on images of the attention.

   The world Health Organization estimates that 347 million folks have the disease international. DR is an eye unwellness connected with long-standing polygenic disorder. De velopment of vision accidental injury will be reduced or stopped if DR is detected in time. On the other hand this may be troublesome because the unwellness typically shows few symptoms till it's too late to provide effective treatment.

   Currently, detection DR could be a long and manual method that needs a trained practician to look at and appraise digital color fundus pictures of the retina. By the time people submit their reviews, typically a day or two later, the delayed results result in lost follow up, miscommunication, and delayed treatment.

   Clinicians will determine DR by the presence of lesions related to the tube-shaped structure abnormalities caused by the unwellness. Whereas this approach is effective, its resource demands are high. The experience and instrumentality needed are typically lacking in areas wherever the rate of polygenic disorder in local populations is high and DR detection is most required. Because the number of people with polygenic disorder continues to grow, the infrastructure required to stop vision defect attributable to DR can become even a lot of light.

   Figure 1.1 Structure of Eye Affected by Diabetic Retinopathy

   The need for a comprehensive and automatic methodology of DR screening has long been recognized, and former efforts have created sensible progress victimization image classification, pattern recognition, and machine learning. With color fundus photography as input, the goal of this competition is to push an automatic detection system to the limit of what's potential ideally leading to models with realistic clinical potential.

2. DETECTION OF DIABETIC RETINOPATHY WITH FEATURE EXTRACTION

   The automatic detection of diabetic retinopathy is completed with the assistance of feature extraction from digital fundus pictures. This feature extraction is achieved using MATLAB. The features studied during this approach are micro-aneurysms, optic disc, exudates and blood vessels.

   The structure of this method contains of varied modules and therefore the complete output depends upon the success rate of each and every individual step. The figure 2.1 provides the general approach of feature extraction and thereby detection severity of the illness.

   Fig.2.1 Overall Approach of Feature Extraction

3. DETECTION OF DIABETIC RETINOPATHY IN

   BLURRED IMAGE

   In this approach an automatic detection structure for fundus pictures underneath a severe scenario of blurred pictures were investigated. During this approach the application of the regular filter deblurring algorithm is usually recommended as a preceding step to the automated detection so as to boost the performance of the automated detection system. Simulation results have shown that regularized filter plays a significant role in enhancing the fundus pictures and achieves high detection rates with sound reproduction options. The system will acknowledge numerous stages with a mean accuracy of over ninety two, a sensitivity of over 880 yards, and a specificity of 100 percent.

   Fig 3.1 Normal Retinal Eye Vs Diabetic Retinopathy Eye

   Retinal image segmentation includes steps like morphological process, thresholding and edge detection followed against this improvement. Mathematical morphology will be employed in the process of retinal pictures. The most processes used here are dilation, erosion, opening, and closing .These processes involve a special mechanism of mixing two sets of pixels. Usually, one set consists of the image being processed and therefore the alternative constitutes the structuring component or kernel. Two vital transformations are opening and closing. Intuitively, dilation expands a picture object and erosion shrinks it.

4. RETINAL GRADING ALGORITHM

   Retinal pictures have pathological noise and backgrounds with varied texture that result in issues in vessel extraction .In order to get rid of the noise, gray opening operation and an opening by reconstruction with a linear structuring component are used to the original image at varied orientations.

   Secondly, the highest Hat transform combined with reconstruction opening and closing operations are projected to strengthen the smoothened image by iteratively filtering the image.

   Fig 4.1 Overall Approach

   The distribution of exudates round the region of the fovea helps to work out the severity of the unwellness. The lesions near the macular region are additional dangerous than the once further away and need immediate intervention. Within the same approach the count, size and therefore the distribution of the hemorrhages and microaneurysms facilitate to determine the severity of the Retinopathy. The International Council of ophthalmology identifies 5 classes of Retinopathy of Diabetic origin like none, delicate moderate, severe and of proliferative state. The Retinal Grading Algorithm helps to mechanically classify the DR (Diabetic Retinopathy) intensity supported these criteria.

5. DETECTION OF MICROANEURYSMS IN FUNDUS

   IMAGE

   The projected approach to resolve the matter of detecting candidates on retinal fundus pictures, wherever candidates are regions probably comparable to microaneurysms (MAs), is separated into three totally different stages, one preprocessing stage (to cut back non-uniform illumination and to normalize the gray scale intensities) and two selection stages (to analyze gray scale content and shape), that are portrayed within the Fig.5.1.

   Fig 5.1 Stages of the process for extraction of Microaneurysm candidates

   The two stages, used for detection of MA candidates, are based on morphologic process. one in all these stages receives a gray-level image as input, classifies a picture region as candidate or non-candidate by analyzing intensity values of the region through the employment of the bottom- hat transform, and generates a binary image as output with most of the red regions detected (red lesions and blood vessels).

   The need for candidates continues on the binary image with a second stage that analyzes shape to pick out candidates through the utilization of the random transform. Its during this stage that the majority of the blood vessels are removed.

   Fig 5.2 Flowchart

   The final classifier distinguishes between real MAs and false positives (FP), comparable to isolated blood vessel regions or vessel crossings, through the extraction of two features that are regional descriptors. The proposed technique differs from others in this the quantity of features to extract, for the ultimate classifier, is extremely tiny (two features).

   The proposed MA detection technique achieves sensitivity, specificity and exactness of 92.32%, 93.87% and 95.93% for the diaretDB1 information and 88.06%, 97.47% and 92.19% for the ROC database. TP, FP, TN, FN are obtained by investigating the entire number of candidates and real lesions detected by the proposed algorithmic rule once applying it on all testing pictures.

6. CONCLUSION

In diabetic retinopathy (DR), retina blood vessels are broken owing to fluid discharge from these vessels. Completely different lesions, i.e., Exudes, hemorrhages, microaneurysms, and textures are used to observe the stage of DR. in this paper, use of many image process techniques for DR lesion detection are mentioned and evaluated. Its found that early diagnosing of DR will cut back the possibility of vision loss up to 500th. Image processing techniques mentioned in this paper will find the DR accurately. Hybrid methodology ought to be utilized in order to get higher end in terms of accuracy and potency for DR detection.

REFERENCES

1. P. Aimmanee, B. Uyyanonvara, P. Jitpakdee, "A survey on hemorrhage detection in diabetic retinopathy retinal images," in Proceedings of the 9th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON), 2012, pp. 1-4.

2. A. Agrawal, C. Bhatnagar, A. S. Jalal, "A survey on automated microaneurysm detection in diabetic retinopathy retinal images," in Proceedings of the International Conference on Information Systems and Computer Networks (ISCON), 2013, pp. 24-29.

3. C. E. Baudoin, B. J. Lay, J. C. Klein, "Automatic detection of microaneurysms in diabetic fluorescein angiography," Revue D' Épidémiologie et de Santé Publique, vol. 32, pp. 254-261, 1984.

4. T. Spencer, R. P. Phillips, P. F. Sharp, J. V. Forrester, "Automated detection and quantification of microaneurysms in fluorescein angiograms," Graefe completo Arch. Clin. Exp. Ophthalmol, vol. 230, pp. 36-41, 1992.

5. T. Spencer, J. A. Olson, K. C. McHardy, P. F. Sharp, J. V. Forrester, "An image-processing strategy for the segmentation and quantification of microaneurysms in fluorescein angiograms," Computers and Biomedical Research, vol. 29, pp. 284-302, 1996.

6. M. J. Cree, J. A. Olson, K. C. McHardy, P. F. Sharp, J. V. Forrester, "A fully automated comparative microaneurysm digital detection system," in Eye, vol. 11, pp. 622-628, 1997.

7. J. H. Hipwell, F. Strachan, K. C. McHardy, J. A. Olson, P. F. Sharp,

   J. V. Forrester, "Automated detection of microaneurysms in digital red-free photographs: a diabetic retinopathy screening tool," Diabetic Medicine, vol. 17, pp. 588-594, 2000.

8. C. Sinthanayothin, J. F. Boyce, T. H. Williamson, H. L. Cook, E. Mensah, S. Lal, D. Usher, "Automated detection of diabetic retinopathy on digital fundus images," Diabetic Medicine, vol. 19, pp. 105-112, 2002.

9. M. Mathivanan and S. Chenthur Pandian, Analysis of Noise Characteristics in VMR-WB speech using subband filters, Australian Journal of Basic and Applied Sciences, ISSN : 1991- 8178, Vol.7, Issue.2, pp.79-88, 2013.

10. M. Niemeijer, B. Van Ginneken, J. Staal, M. S. A. Suttorp-Schulten,

    1. D. Abramoff, "Automatic detection of red lesions in digital color fundus photographs," IEEE Transactions on Medical Imaging, vol. 24, pp. 584-592, 2005.

    2. Pradeepa received her Bachelor of Engineering degree specialized in Electronics and Communication Engineering at SNS College of Technology, Coimbatore, in the year 2009, under Anna University, Chennai, Tamilnadu. She received her Master of Engineering degree specialized in Applied Electronics at Dr.Mahalingam College of Engineering and Technology,

Pollachi, in the year 2012 under Anna University, Chennai. She is now working as Assistant Professor in the Department of Electronics and Communication Engineering at SNS College of Technology, Coimbatore, Tamilnadu, India. She is currently pursuing PhD under Anna University, Chennai. Her area of interest includes Image Restoration (Denoising and Deblurring), Feature Extraction and Digital Signal Processing. She is a member of ISRD. She have attended 4 international conferences and has published 8 papers in International Journals.

S. Chenthur Pandian obtained his AMIE-A(1984) and AMIE- B(1991): Electrical Engineering from Institution of Engineers (India), Calcutta; Master of Engineering in Power Systems from Punjab University, Chandigarh (1994); Doctorate of Philosophy in Fuzzy Application for Power Systems from Periyar University, Salem (2005). He authored a Textbook

Basic Electrical, Electronics & Communication Engg. and Co-Authored a Textbook Total Quality Management and Lab Manuals for Electrical Engineering Laboratories. He joined as an Assistant professor (1994 – 1996) in Sapthagiri College of Engg., Dharmapuri, TN. He worked as Head/EEE (1996 – 2005) in K.S.R College of Technology, Tiruchengode and promoted to the post of Director (2005 – 2006). He served as a Founder Principal (2006 – 2010) in Selvam College of Technology, Namakkal and as a Principal (2010 – 2013) in Dr. Mahalingam College of Engineering and Technology, Pollachi. Presently he is attached with SNS College of Technology, Coimbatore, as Principal. He is a Recognized supervisor of Anna University, Chennai. He is a Reviewer of Elsevier and IEEE International Journals and Indian Examiner for Ph.D Evaluation in few Indian Universities. Under his guidance 16 Research Scholars have completed their Ph.D. and currently 8 research scholars have been pursuing Ph. D. He had published 95 Papers in International Journals and 25 papers in International Conferences. He had presented 43 papers in National Conferencaes and organized 26 Conferences. He is a Fellow of Institution of Engineers (FIE), Member of Indian Society for Technical Education (MISTE), Member of Institute of Electrical and Electronics Engineers, MIEEE (USA).