Printer Friendly

A comprehensive of symptoms of retinal disease and its segmentation: survey.


According to estimates from the World Health Organization (WHO) Prevention of Blindness and Deafness Programme:

* Around 285 million people are visually impaired around the world: 39 million are visually impaired and 246 million have low vision (serious or moderate visual disability).

* 4 out of 5 visually impaired or visually impaired people are avoidably so--preventable causes are as high as 80% of the total global visual impairment burden.

* Around 90% of the world's visually impaired individuals live in developing countries.

* 65% of visually impaired, and 82% of blind people are over 50 years of age, although this age group comprises only 20% of the world population.

* Top causes of visual impairment: refractive errors, cataracts and glaucoma.

* Top causes of blindness: cataracts, glaucoma and age-related macular degeneration.

* The number of people visually impaired from infectious diseases has greatly reduced in the last 20 years.


A cataract may be the obfuscation of the crystalline lens in the eye that brings about diminished dream barely in searching through an iced glass window shown in Figure 1.1 b) (Cataract Eye). The primary danger part to creating an cataract will be agnostics which implies that there isn't a great part that might make carried will keep a cataract framing. Sadly there is no cure to ageing yet! Fortunately, cataracts might be treated with surgery the place an eye specialist will uproot your shady lens and repeatability it with an acceptable lens insert. When a cataract need been uprooted it can't develop once again.

Macular degeneration(MD):

is a constant illness that influences your focal vision show in Figure 1.2 (Macular degeneration). It is likewise generally known as MD, age related macular degeneration or ARMD. Dangers: Age. Anybody beyond 50 years old is at danger of creating ARMD and the hazard increments with age. 1 in 7 individuals beyond 50 years old have some type of ARMD.

1. Smoking expands the hazard by 3 overlap

2. Hereditary qualities.

3. Consume less calories in high soaked fats.


otherwise known as 'the sneak cheat of sight' - is a sickness in which the optic nerve head is gradually harmed, for the most part by expanded weight in the eye where shown in Figure 1.3 (Glaucoma Eye).

The eyeball is loaded with liquid to keep up it's exquisite round shape. In the event that a lot of liquid is being created, or all the more generally insufficient is depleting out, then the "weight" in the eye can increment. This ascent in weight can push down on the optic nerve take and progressively murders off the nerve strands that send our visual message to the cerebrum.

The harm to the optic nerve head advances gradually and steadily causes changeless loss of vision, beginning with your fringe vision. The vision misfortune is irreversible and ordinarily goes unnoticed until it is past the point of no return. Treatment can not get back the vision that has been lost, yet it can stop or back off the movement and spare what sight is cleared out. That is the reason it is extremely import to have standard eye examinations to recognize glaucoma changes as ahead of schedule as could be expected under the circumstances.


A genuine eye difficulty of diabetes is diabetic retinopathy which is the place retinal veins release liquid bringing about vision misfortune and at times visual deficiency shown in Figure 1.4(Diabetic Retinopathy).

Both sort 1 and sort 2 diabetics are at danger of creating diabetic retinopathy which is the reason all diabetics ought to have a far reaching eye examination in any event once every year.

Examines have demonstrated that hoisted circulatory strain and cholesterol levels increment the danger of diabetic retinopathy. So, tune in to your General Practitioner and attempt and gain a tight power on your glucose, circulatory strain and cholesterol levels.


The cornea (the reasonable front some portion of the eye) is sporadic fit as a fiddle Figure 1.5 (Keratoconics Eye). Rather than being decent and round like a soccer ball, it is cone-like fit as a fiddle. Since the front surface is unpredictable, glasses don't make the vision 100% clear. Hard contact focal points, RGPs, are the best visual alternative for keratoconics.

2. Related Work:

Glaucoma Screening[8] propose a technique for cup to disc ratio(CDR) appraisal utilizing 2-D retinal fundus pictures. Techniques: In the proposed strategy, the optic circle is first portioned and reproduced utilizing a novel sparse dissimilarity constrained coding (SDC) approach which considers both the divergence imperative and the sparsity requirement from an arrangement of reference plates with known CDRs. Therefore, the recreation coefficients from the SDC are utilized to process the CDR for the testing plate. Comes about: The proposed technique has been tried for CDR appraisal in a database of 650 pictures with CDRs physically measured via prepared experts beforehand. Exploratory outcomes demonstrate a normal CDR blunder of 0.064 and relationship coefficient of 0.67 contrasted and the manual CDRs, superior to the best in class techniques. Our proposed technique has likewise been tried for glaucoma screening. The technique accomplishes zones under bend of 0.83 and 0.88 on datasets of 650 and 1676 pictures, individually, outperforming different techniques. This technique accomplishes great exactness for glaucoma identification. Significance. The technique has an awesome potential to be utilized for vast scale populace based glaucoma screening.

The region types included microaneurysms[18], exudates, neovascularization on the retina, hemorrhages, normal retinal background, and normal vessels patterns. The cumulative distribution functions of the instantaneous amplitude, the instantaneous frequency magnitude, and the relative instantaneous frequency angle from multiple scales are used as texture feature vectors. Using distance metrics between the extracted feature vectors to measure interstructure similarity. The results demonstrate a statistical differentiation of normal retinal structures and pathological lesions based on AM-FM features. Further demonstrate our AM-FM methodology by applying it to classification of retinal images from the MESSIDOR database. Overall, the proposed system indicates significant capacity for use in programmed DR screening.

A technique for the programmed detection of microaneurysms (MAs)[10] in color retinal pictures is proposed. The recognition of MAs is a fundamental step in the determination and reviewing of diabetic retinopathy. The proposed strategy recognizes MA location through the examination of directional cross-segment profiles focused on the neighborhood greatest pixels of the preprocessed picture. Crest recognition is connected on each profile, and a set of attributes regarding the size, height, and shape of the peak are calculated subsequently. The statistical measures of these attribute values as the positioning of the cross-area changes constitute the list of capabilities that is utilized as a part of a naive Bayes classification to reject spurious candidates. Method give a formula for the final score of the remaining candidates, which can be thresholded assist for a binary output. The proposed strategy has been tested in the Retinopathy Online Challenge, where it verified by being focused with the state-of-the-art approaches. Current experimental results for a private image set utilizing the same classifier setup.

Segmentation of Optical Disc(OP) in Retinal Image[1] is fully automatic OD localization and segmentation algorithm. First, OD size estimation adaptive to different image resolution. OD location candidates are identified by using RGB red channel, color retinal image more tends to be saturated. Background normalization, oversmoothed background image by average filtering using window three times the size of the estimated OD radius. Template matching to locate the OD canditates a binary template where the disk, is white then the value is 1 and the black background is assigned a value 0.

OD localization examples. First column: Input retinal images. Second column: Background normalized CIElab lightness images. Third column:OD candidates (green) and detected OD location (red) on matched filtering response images. Fourth column: Results of OD localization shown in Figure 2.1 OD Localization.

In next step OD Segmentation Algorithm carries Image Prepcocessing as saturation detection in the red channel, blood vessel removal and bright region removals are processed.

Optical Disc Localization [11] an automatic method for the optic disc localization in retinal images, which is effective and reliable with multiple datasets. A new vessel enhancement method based on a modified corner detector. The vessel enhancement is combined with morphological operators, to detect the four main vessels orientations {0, 45, 90, 135}. These four image functions has the details to determine an initial optic disc localization, and as a result in two images that are respectively divided along the vertical or horizontal orientations with different division sizes. Final localization of the OD is achieved on the two most distinct features of the OD: high convergence regions and high intensity values.


OD(Optic Disc) and OC(Optic Cup) [12] were segmented using superpixel classification for glaucoma screening. For segmentation 650 images from database with optic disc and optic cup boundaries are marked manually by professional trainers. Experimental results gives an average overlapping for error for optic disc and optic cup segmentation is 9.5% and 24.1% respectively. This method achieves areas under curve of 0.800 and

0.822 in two datasets.

OD and OC regions [15] are obtained from monocular retinal images. A novel OD segmentation strategy is proposed which integrates the local image information around each point of interest in multidimensional component space to give robustness against variations found in and around the OD area.

Segmenting Retinal Blood Vessels [13] future a supervised segmentation strategy that uses a deep neural network prepared on a large (up to 400 00) test cases, pre-processed with global contrast normalizations and gamma corrections.

Blood Vessels and OD [16] in funds retinal images. Strategy makes extraction of the retina vascular tree utilizing the graph cut method. The blood vessel information is then used to assess the location of the OD. The OD segmentation is performed using two alternative strategies. The Markov random field (MRF) image reconstruction strategy segments the OD by removing vessels from the OD area, and the compensation factor technique segments the OD utilizing the prior local intensity knowledge of the vessels.

Retinal Vessel Segmentation [14] technique produces segmentations by classifying each image pixel as vessel or nonvessel, founded on the pixel's feature vector. Feature vectors are made out of the pixel's intensity and two-dimensional Gabor wavelet transform responses taken at multiple scales.

Automatic exudate segmentation [17] in color retinal fundus images, which incorporates three phases: anatomic structure removal, exudate area and exudate segmentation. The technique evaluates both exudate-level and image-level. For exudate-level evaluation tested on e-ophtha EX dataset, provides pixel level annotation. This technique achieves 76% insensity and 75% in PPV. For image-level evaluation, tested on DiaRetDB1 and achieves 83%, 75%, 79% on sensitivity, specificity, accuracy respectively.

IV Measures:

True Positive (TP):

is used to represent the vessel pixels properly detected as vessel pixels itself.

True Negative (TN):

is the number of non-vessel pixels properly detected as non-vessel pixels itself.

False Positive (FP):

is the number of non-vessel pixels itself.

False Negative (FN):

id the number of vessel pixels wrongly detected as non-vessel pixels.

Sensitivity, Specificity, Predictive Value, Accuracy are calculated as,

Sensitivity = TP/TP + FN

Specificity = TN/TN + FP

Predictive value (PV) = TP/TP + FP

Accuracy = TP + TN/TP + FP + TN + FN

F-Score = 2TP/2TP + FN + FP

Fall-Out of False Positive Rate (FPR)

FPR =(FP/FP + TN) = 1 - SPC

False Negative Rate (FNR)

FNR = (FN/TP + FN) = 1 - TPR

False Discovery Rate (FDR)

FDR = (FP/TP + FP) = 1 - PPR

Database Comparison:
Table 1.1: Comparison of Various Disease detecting techniques
with obtained result.

Paper   Disease          Techniques

[1]     Optic Disc       Five Popular Swarm Intelligence Algorithms:
        detection        Artificial Bee Colony, Particle Swarm
                         Optimization, Bat Algorithm, Cuckoo Search
                         and Firefly Algorithm.

                         Pre-processing involves background
                         subtraction, median filtering and mean
                         filtering and is named as Background
                         Subtraction-based Optic Disc Detection(BSODD)

[2]     Diabetic         Stages of DR are Non-Proliferate Diabetes
        retinopathy      Retinopathy and Proliferate Diabetes

                         Probabilistic Neural Network(PNN), Decision
                         Tree (DT) and Support Vector Machine(SVM)

                         For Optimization Techniques Genetic Algorithm
                         and Particle Swarm Optimization

[4]     Diabetic         Machine Learning Classifiers

[3]     Microaneurysms   Hybrid Classifier combines Gaussian Mixture
                         Mode(GMM) and Support Vector Machine(SVM)

[5]     Bright Lesions   Linear Support Vector Machine(SVM)

[6]     Hard Exudates    Dynamic threshold, edge detection

Paper   Database     Obtained Result

[1]     DRIVE        Obtained Accuracy of
        DiaRetDB1    DRIVE=100%
        DMED         DiaRetDB 1=100%
        STARE        DMED=98.82%

[2]     Fundus       Accuracy=96.15%
        Images       Sensitivity=96.27%

[4]     Messidor     Sensitivity=90%

[3]     DIARETDB1    Sensitivity=98.64%

[5]     STARE        Normal Class

                     Drusen Class

                     Exudates Class

[6]     80 Retinal   Sensitivity=90.2%
        Images       Predictive=96.8%


Images of retina captured using fundus camera are used for processing the retinal diseases. Fundus image consist retina, optic disc, macula & posterior pole. Image databases are essential since all image processing algorithms developed have to be tested and confirmed. An overview of all publicly available retinal image databases known to us is given in this section. [19], [20],[21],[22],[23],[24]


This paper has been presented the survey and comparative study of retinal image diseases identification techniques. Like Retinal Diabetic, Microaneurysms, Hard Exudates, blood vessels segmentation has been performed accurately using automatic detection by classifier based vessel tracking methods. The Messidor, STARE, DRIVE, database also been described has been provided better performance in terms of resolution, speed, and accuracy.


[1.] Yu, H., E.S. Barriga, 2012. "Fast Localization and Segmentation of Optic Disk in Retinal Images Using Directional Matched Filtering and Level Sets" IEEE TRANSACTIONS ON INFORMATION TECHNOLOGY IN BIOMEDICINE, 16: 4.

[2.] Mookiah, M.R.K., U. Rajendra Acharya, 2013. "Evolutionary algorithm based classifier parameter tuning for automatic diabetic retinopathy grading: A hybrid feature extraction approach".--Knowledge-Based Systems, 39: 9-22.

[3.] Usman Akrama, M., Shehzad Khalidb, Shoab A. Khana, 2013. "Identification and classification of microaneurysms for early detection of diabetic retinopathy" - Pattern Recognition, 46(1): 107-116.

[4.] Imani E1, Pourreza HR2, Banaee T3. 2015. "Fully automated diabetic retinopathy screening using morphological component analysis". Comput Med Imaging Graph, 43: 78-88.

[5.] Sidibe, D., I. Sadek, F. Meriaudeau, 2015. "Discrimination of retinal images containing bright lesions using sparse coded features and SVM" - Comput Biol Med., 62: 175-84.

[6.] Sanchez, C.I.1., M. Garcia, A. Mayo, M.I. Lopez, R. Hornero, 2009. "Retinal image analysis based on mixture models to detect hard exudates"- Med Image Anal., 13(4): 650-658.

[7.] Balint Antal, Andras Hajdu, 2014. "An ensemble-based system for automatic screening of diabetic retinopathy"- Knowledge-Based Systems, 60: 20-27.

[8.] Jun Cheng, Fengshou Yin, Damon Wing Kee Wong, 2015. "Sparse Dissimilarity-Constrained Coding for Glaucoma Screening"- IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, 62: 5.

[9.] Tamil Pavai, G., S. Tamil Selvi, 2013. "Identification Of Proliferative Diabetic Retinopathy Using Texture Segmentation", Journal of Computer Science., 9(3): 358-367.

[10.] Hanung Adi Nugroho, Dhimas Arief Dharmawan, "Automated microaneurysms (MAs) detection in digital colour fundus images using matched filter"-

[11.] Ivo Soares, Miguel Castelo-Branco, Antonio M. G. Pinheiro, 2016. "Optic Disc Localization in Retinal Images Based on Cumulative Sum Fields"- IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, 20: 2.

[12.] Jun Cheng*, Jiang Liu, Yanwu Xu, Fengshou Yin, 2013. "Superpixel Classification Based Optic Disc and Optic Cup Segmentation for Glaucoma Screening" - IEEE TRANSACTIONS ON MEDICAL IMAGING, 32: 6.

[13.] Pawel Liskowski * and Krzysztof Krawiec, 2016. "Segmenting Retinal Blood Vessels With Deep Neural Networks"- IEEE TRANSACTIONS ON MEDICAL IMAGING, 35(11).

[14.] Joao V. B. Soares *, Jorge J.G. Leandro, Roberto M. Cesar Jr., Herbert F. Jelinek and Michael J. Cree, 2006. "Retinal Vessel Segmentation Using the 2-D Gabor Wavelet and Supervised Classification" - IEEE TRANSACTIONS ON MEDICAL IMAGING, 25: 9.

[15.] Gopal Datt Joshi, Jayanthi Sivaswamy, S.R. Krishnadas, 2010. "Optic Disk and Cup Segmentation from Monocular Colour Retinal Images for Glaucoma Assessment" - IEEE TRANSACTIONS ON MEDICAL IMAGING.

[16.] Ana Salazar-Gonzalez, Djibril Kaba, Yongmin Li, 2014. Xiaohui Liu- "Segmentation of the Blood Vessels and Optic Disk in Retinal Images"- IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, 18: 6.

[17.] Qing Liu (a,b,c), Beiji Zou (a,b), Jie Chen (c), Wei Kec, Kejuan Yue (d), Zailiang Chen (a,b) *, Guoying Zhao (c), 2017. "A location-to-segmentation strategy for automatic exudate segmentation in colour retinal fundus images" Computerized Medical Imaging and Graphics, 55: 78-86.

[18.] Agurto, C., V. Murray, E. Barriga, S. Murillo, M. Pattichis, H. Davis, S. Russell, M. Abramoff, P. Soliz 2010. "Multiscale AM-FM Methods for Diabetic Retinopathy Lesion Detection"- IEEE TRANSACTIONS ON MEDICAL IMAGING, 29: 2.

[19.] Etienne Decenciere, Xiwei Zhang, Guy Cazuguel, Bruno Lay, Beatrice Cochener, et al. FEEDBACK ON A PUBLICLY DISTRIBUTED IMAGE DATABASE: THE MESSIDOR DATABASE.

[20.] 1/



[23.] Prof. S.V. Pattalwar, MR.L. SKalkonde, 2014. "An Overview of Different Analysis Algorithm and Database for Early Detection of Diabetic Retinopathy International Journal of Pure and Applied Research in Engineering And Technology-. IJPRET, 2(9).

(1) B. Sakthi Karthi Durai B.E and (2) M.Tech, Dr.J.Benadictraja M.E, Ph.D,

(1) Assistant Professor, Department of Computer Science and Engineering, Psna College Of Engineering

(2) Assistant Professor, Department of Computer Science and Engineering, Psna College Of Engineering

Received 28 January 2017; Accepted 22 May 2017; Available online 28 May 2017

Address For Correspondence:

B. Sakthi Karthi Durai B.E, Assistant Professor, Department of Computer Science and Engineering, Psna College Of Engineering

Caption: Fig. 1.1 a: Normal eye

Caption: Fig. 1.1 b: Cataract Eye

Caption: Fig. 1.2: Macular degeneration

Caption: Fig. 1.3: Glaucoma Eye

Caption: Fig. 1.4: Diabetic Retinopathy

Caption: Fig. 1.5: Keratoconics Eye

Caption: Fig. 2.1: OD Localization
Table 1.2: Lists of Database and Description

Database Name         Description

DRIVE: Digital        Obtained from a diabetic retinopathy screening
Retinal Images for    program in The Netherlands. The screening
Vessel Extraction     population consisted of 400 diabetic subjects
                      between 25-90 years of age. Forty photographs
                      have been randomly selected, 33 do not show any
                      sign of diabetic retinopathy and 7 show signs of
                      mild early diabetic retinopathy.

                      The images were acquired using a Canon CR5
                      non-mydriatic 3CCD camera with a 45 degree field
                      of view (FOV). Each image was captured using 8
                      bits per color plane at 768 by 584 pixels. The
                      FOV of each image is circular with a diameter of
                      approximately 540 pixels.

                      The set of 40 images has been divided into a
                      training and a test set, both containing 20

STARE                 20 retinal fundus slides and their ground truth
(STructured           images in the STARE (Structured Analysis of
Analysis of the       Retina) database. The images are digitized
Retina)               slides captured by a TopCon TRV-50 fundus camera
                      with 35 degree field of view. Each slide was
                      digitized to produce a 605 x 700 pixel image
                      with 24-bits per pixel. All the 20 images were
                      carefully labeled by hand to produce ground
                      truth vessel segmentation by an expert.

                      The full set of 400 raw images in the STARE
                      database. A total of 44 possible manifestations
                      were queried to the experts during data
                      collection and then reduced to 39 values during

                      Blood vessel segmentation work including 40 hand
                      labeled images.

                      Optic nerve detection work including 80 images
                      with ground truth.

MESSIDOR              The 1200 eye fundus color numerical images of
DATABASE              the posterior pole for the MESSIDOR database
                      were acquired by 3 ophthalmologic departments
                      using a color video 3CCD camera on a Topcon TRC
                      NW6 non- mydriatic retinograph with a 45 degree
                      field of view. The images were captured using 8
                      bits per color plane at 1440*960, 2240*1488 or
                      2304*1536 pixels.

                      800 images were acquired with pupil dilation
                      (one drop of Tropicamide at 0.5%) and 400
                      without dilation. The 1200 images are packaged
                      in 3 sets, one per ophthalmologic department.
                      Each set is divided into 4 zipped sub sets
                      containing each 100 images in TIFF format and an
                      Excel file with medical diagnoses for each

DIARETDB0 -           The current database consists of 130 color
Standard Diabetic     fundus images of which 20 are normal and 110
Retinopathy           contain signs of the diabetic retinopathy (hard
Database              exudates, soft exudates, micronaneuyrysms,
Calibration level 0   hemorrhages and neovascularization). Images were
                      captured with a 50 degree field-of-view digital
                      fundus camera with unknown camera settings. The
                      data correspond to practical situations, and can
                      be used to evaluate the general performance of
                      diagnosis methods. This data set is referred to
                      as "calibration level 0 fundus images".

DIARETDB1 -           The database consists of 89 colour fundus images
Standard Diabetic     of which 84 contain at least mild
Retinopathy           non-proliferative signs (Microaneurysms) of the
Database              diabetic retinopathy, and 5 are considered as
Calibration level 1   normal which do not contain any signs of the
                      diabetic retinopathy according to all experts
                      who participated in the evaluation. Images were
                      captured using the same 50 degree field-of-view
                      digital fundus camera with varying imaging
                      settings. The data correspond to a good (not
                      necessarily typical) practical situation, where
                      the images are comparable, and can be used to
                      evaluate the general performance of diagnostic
                      methods. This data set is referred to as
                      "calibration level 1 fundus images".
COPYRIGHT 2017 American-Eurasian Network for Scientific Information
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2017 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Durai, B. Sakthi Karthi; Benadictraja, J.
Publication:Advances in Natural and Applied Sciences
Article Type:Report
Date:May 1, 2017
Previous Article:Information security and resource sharing using AWS public cloud.
Next Article:Big Data analysis: a step to define.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters