Various Retinal Problems and Techniques to analyze it Using image processing
Various Retinal Problems and Techniques to analyze it
Using image processing: A review
S.D.Mendhule1, Dr. R.J.Bhiwani2
1M.E. (Scholar), Dept. of Electronics and Telecommunication, Babasaheb Naik College of Engineering, Pusad
2Professor, Dept. of Electronics and Telecommunication, Babasaheb Naik College of Engineering, Pusad,
E-mail:- 1 [email protected],
2 [email protected].
ABSTRACT
Diabetic-related eye disease is a major cause of preventable blindness in the world. It is a complication of diabetes
which can also affect various parts of the body. When the small blood vessels have a high level of glucose in the
retina, the vision will be blurred and can cause blindness eventually. This is known as diabetic retinopathy (DR).
Regular screening is essential in order detect the early stages of diabetic retinopathy for timely treatment to prevent
or delay further deterioration. The various retinal problems may include Microaneurysm (MA), Exudates, and
Haemorrhages. In this paper various techniques are reviewed to reduce the cost and increase productivity and
efficiency for ophthalmologists.
Keywords: – Diabetic Retinopathy (DR), Image Processing, Retinal problems.
1. INTRODUCTION
According to recent estimates, approximately 425 million people worldwide in the 20–79 year age group will have
diabetes in 2017 and by 2045, 629 million people of the adult population, is expected to have diabetes 1. The
largest increases will take place in the regions dominated by developing economies. The prevalence of diabetes is
rising all over the world due to population growth, aging, urbanization and an increase of obesity and physical
inactivity. Unlike in the West, where older persons are most affected, diabetes in Asian countries is
disproportionately high in young to middle-aged adults. This could have long-lasting adverse effects on a nation’s
health and economy, especially for developing countries. The International Diabetes Federation (IDF) estimates
the total number of people in India with diabetes to be around 72.9 million in 2017, rising to 134.3 million by
20451,making the India diabetic capital of the World. The ratio of ophthalmologists to the number of diabetic
patient is very low. Ophthalmologists in India are insufficient to support the growing diabetic population. India
has one ophthalmologist per 1 Lac. Patients and this ratio is even smaller for rural areas. Today diabetic retinopathy
is a third cause of blindness in India.
The “Top 10” countries in the world, in terms of the number of people with diabetes, for 2017 and 2045, are shown
in Table 1.
Table 1: Top ten countries/territories for number of people with diabetes (20-79 years), 2017 and 2045 1
Rank Country /Territory Number of people with diabetes 2017 (millions) Country / Territory Number of people with diabetes 2045 (millions)
1 China 114.4 India 134.3
2 India 72.9 China 119.8
3 U.S. 30.2 U.S. 35.6
4 Brazil 12.5 Mexico 21.8
5 Mexico 12.0 Brazil 20.3
6 Indonesia 10.3 Egypt 16.7
7 Russian Federation 8.5 Indonesia 16.7
8 Egypt 8.2 Pakistan 16.1
9 Germany 7.5 Bangladesh 13.7
10 Pakistan 7.5 Turkey 11.2
Diabetic Retinopathy is a complication of diabetes, especially for those with type 2 diabetes. High blood sugar
levels over a period of time can damage the blood vessels in the retina, making them, swell and leak. In some cases
it may also happen that they block blood from passing through. These lead to growth of abnormal new blood
vessels in the retina. All of these conditions affect the vision adversely leading ultimately to loss of vision.
Although diabetic retinopathy is caused by prolonged high blood sugar level, prolonged high blood pressure has
an additive effect. The risk of the condition increases with age and duration of uncontrolled blood sugar. The high
glucose levels in the blood vessels damage the sensitive small blood vessels in the retina causing haemorrhages,
exudates and even swelling of the retina. When this occurs, the retina is deprived of oxygen which in turn leads to
the growth of abnormal blood vessels. A condition known as proliferative diabetic retinopathy. The retina detects
light and converts it to signals sent through the optic nerve to the brain. Diabetic retinopathy may stop this process
and cause loss of vision as shown in figure 1.
.
A B
Figure 1: Influence of diabetes on vision: (A) normal vision; (B) diabetic retinopathy
Types of Diabetic retinopathy:
Nonproliferative diabetic retinopathy (NPDR): It is an early stage of diabetic retinopathy. In this stage, tiny
blood vessels within the retina leak blood or fluid. The leaking fluid causes the retina to swell or to form deposits.
Proliferative diabetic retinopathy (PDR): or advanced diabetic retinopathy, is the stage where new blood vessels
begin to grow within the retina. These new blood vessels are usually abnormal and grow in the center of the eye.
This growth is accompanied by haemorrhages, exudates, occlusions and vision defects.
Abnormalities associated with Diabetic Retinopathy:
Following are the main abnormalities that are associated with the Diabetic Retinopathy
Microaneurysms: These 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 haemorrhages within the light sensitive retina. Their sizes
ranges from 10-100microns i.e. less than 1/12th the diameter of an average optics disc and are circular in shape, at
this stage, the disease is not eye threatening.
Haemorrhages: Occurs in the deeper layers of the retina and are often called ‘blot’ haemorrhages because of their
round shape.
Hard exudates: These are one of the main characteristics of diabetic retinopathy and can vary in size from tiny
specks to large patches with clear edges. As well as blood, fluid that is rich in fat and protein is contained in the
eye and this is what leaks out to form the exudates. These can impair vision by preventing light from reaching the
retina.
Soft exudates: These are often called ‘cotton wool spots’ and are more often seen in advanced retinopathy.
Neovascularisation: This can be describe as abnormal growth of blood vessels in areas of the eye including the
retina and is associated with vision loss. This occurs in response to ischemia, or diminished blood flow to ocular
tissues. If these abnormal blood vessels grow around the pupil, glaucoma can result from the increasing pressure
within the eye. These new blood vessels have weaker walls and may break and bleed, or cause scar tissue to grow
that can pull the retina away from the back of the eye. When the retina is pulled away it is called a retinal
detachment and if left untreated, a retinal detachment can cause severe vision loss, including blindness. Leaking
blood can cloud the vitreous (the clear, jelly-like substance that fills the eye) and block the light passing through
the pupil to the retina, causing blurred and distorted images. In more advanced proliferate retinopathy; diabetic
fibrous or scar tissue can form on the retina.
Figure 2: Showing Abnormalities associated with Diabetic Retinopathy
This paper reviews the Various Retinal Problems and Techniques to analyze the elementary lesions for seriousness
of the retinal disease. The microaneurysm and haemorrhage detection can be used to grade the progression of DR
into four stages: no DR, mild DR, moderate DR and severe DR, as shown in Table 2.
Table 2.Grading criteria of Diabetic Retinopathy
DR Grade
Grade 0 (No DR) MA = 0 ; H = 0
Grade 1 (Mild) 1 ? MA 5 ; H = 0
Grade 2 (Moderate) 5 ; MA ; 15 or 0 5
Captions: MA = Microaneurysm, H = Haemorrhage
Literature Survey on Image Pre-Processing:
Detecting abnormalities associated with fundus image, the images have to be Pre-Processed in order to correct the
problems of uneven illumination problem, non-sufficient contrast between lesions and image background pixels
and presence of noise in the input fundus image. Aside from aforementioned problems, this section is also
responsible for colour space conversion and image size standardization for the system. In this section, various
techniques for the preprocessing has been discuss.
In the method proposed by C.I.O. Martins et al. 2, 6 image preprocessing is performed on the green-plane Igreen
of the original RGB color image Iorg. This operation can be viewed as a shade correction procedure and consists in
removing low gradient regions from the green channel image Igreen. The image Ibg results from median filtering the
Igreen image with a kernel of 25×25 pixels. The size of the median filter kernel was chosen such that it is wider than
the widest blood vessel in the used set of images. The shade corrected image (Isc) is the difference between Igreen
and Ibg, i.e., Isc = Igreen ? Ibg.
A.J. Frame et al. 3 uses a technique of shade correction to remove illumination variations in the image by
smoothing the image with a large scale median filter and subtracting the filtered image from the original. The
shade-corrected image is then normalized to have the same maximum and minimum grey levels as the original
image.
Akara Sopharak et al. 4 apply median filter on green band image to reduce the noise before Contrast Limited
Adaptive Histogram Equalization 8, then shade correction algorithm is applied to the green band in order to
reduce background variation due to non-uniform illumination. 35X35 median filter applied to the image to correct
background variation.
Literature Survey on Lesion detection:
1. Microaneurysms:
Akara Sopharak and et al. 7 used segmentation technique to detect fine microaneurysms, from non-dilated pupils.
The process consists of two segmentation steps: a) Coarse segmentation using mathematic morphology and b) Fine
segmentation using naive Bayes classifier. A total of 18 microaneurysms features were extracted by naive Bayes
classifier. The detected microaneurysms are validated by comparing at pixel level with ophthalmologists’ data.
Balint Antal et al. 12 proposed an ensemble-based framework to improve microaneurysm detection. They
proposed a combination of internal components of microaneurysm detectors, namely preprocessing methods and
candidate extractors, and approach for microaneurysm detection.
Atsushi Mizutani et al. 15 investigated a computerized method for the detection of microaneurysms on retinal
fundus images. After image preprocessing, candidate regions for microaneurysms were detected using a double-
ring filter. Any potential false positives located in the regions corresponding to blood vessels were removed by
automatic extraction of blood vessels from the images. Twelve image features were determined, and the candidate
lesions were classified into microaneurysms or false positives using the rule-based method and an artificial neural
network. The true positive fraction of the proposed method was 0.45 at 27 false positives per image. Forty-two
percent of microaneurysms in the 50 training cases were considered invisible by the consensus of two co-
investigators. When the method was evaluated for visible microaneurysms, the sensitivity for detecting
microaneurysms was 65% at 27 false positives per image.
Pallawala et al 11 proposed a microaneurysm segmentation and detection algorithm that is based on generalized
eigenvectors of affinity matrix. This technique is robust and has a minimal interference from other structures and
lesions achieve 93% accuracy in the detection of microaneurysms.
Quellec et al. 22 proposed lesion template in sub bands of wavelet transformed images. The optimization process
is based on a genetic algorithm followed by Powell’s direction set descent. Results are evaluated on 120 retinal
images analyzed by an expert and the optimal wavelet is compared to different conventional other wavelets. These
images are of three different modalities: there are colour photographs, green filtered photographs and angiographs.
Depending on the imaging modality, microaneurysms were detected with a sensitivity of respectively 89.62%,
90.24% and 93.74% and a positive predictive value of respectively 89.50%, 89.75% and 91.67%.
2. Exudates:
Amel et al. 5 combines k means clustering algorithm and mathematical morphology to detect hard exudates in
retinal images and obtained sensitivity of 95.92%, predictive value of 92.28% and accuracy of 99.70% using a
lesion-based criterion.
Hussain et al 23 adopt a combination of coarse and fine segmentation to detect hard exudates. The result is
achieved with 89.7% sensitivity, 99.3% specificity and 99.4% accuracy. A limitation of this method is that it
occasionally fails to exclude some non-exudate objects particularly those that have similar features to real exudates.
A limitation of their work is that it occasionally fails to exclude some non-exudate objects particularly those that
have similar features to real exudates.
Jaafar et al. 16 presents an automated method for the detection of bright lesions (exudates) in retinal images. In
their work, an adaptive thresholding based on a novel algorithm for pure splitting of the image is proposed. A
coarse segmentation based on the calculation of a local variation for all image pixels is used to outline the
boundaries of all candidates which have clear borders. A morphological operation is used to refine the adaptive
thresholding results based on the coarse segmentation results. Using a clinician reference standard (ground truth),
images with exudates were detected with 91.2% sensitivity, 99.3% specificity, and 99.5% accuracy. Limitations
occurs due to some incorrect non-exudates detection which are caused by those artifacts that have similar features
of real exudates.
Zohra et al. 9 proposed a computer based approach for the detection of diabetic retinopathy stage using color
fundus images .The features are extracted from the raw images using the image processing techniques and fed to
the support vector machine (SVM) and demonstrated a sensitivity of 97.5% for the classifier with the specificity
of 100%.
3. Haemorrhages
Joshi et al. 14 propose algorithm in three main steps 1. Color image enhancement 2.Image subtraction to extract
blood vessels and haemorrhages and 3.Use of set of optimally adjusted morphological operators to suppress blood
vessels and to highlight only haemorrhages. These automatically detected haemorrhages are validated by
comparing with expert ophthalmologists’ hand-drawn ground-truths. Quantitative performance of algorithm is
evaluated by calculating sensitivity and specificity and predictive value (PV). The overall sensitivity, specificity
and PV obtained are 89.49%, 99.89%, and 98.34% respectively. Draw of this method is that it cannot detect faint
and small haemorrhages.
Acharya et al. 24 based on the morphological operation and „Ball' shaped structuring element. The Haemorrhages
were detected by subtracting blood vessels from Haemorrhage candidates.
Godlin et al. 25 analyzed hemorrhage detection in retinal fundus images using classifier and segmentation
methods. All the database images into the pre-processing steps and some meaning full features are extracted from
the images. Then ANFIS classifier utilized to normal and abnormal images, this abnormal category into the
hemorrhage detection process with help of segmentation technique. Here Region growing (RG) with threshold
optimization techniques are considered its known as Modified RG (MRG) to get the maximum accuracy in the
hemorrhage segmenting process.
Publically available data:
A list of public data repositories for retinal image analysis.
This appendix lists the public retinal data sets known to us. Unless otherwise stated, all data sets listed are easily
reachable by a Google search. Most descriptions are excerpts from the referred websites.
DRIVE (http://www.isi.uu.nl/Research/Databases/DRIVE/) is another much cited test set; it was created to enable
comparative studies on segmentation of blood vessels in retinal images. The photographs were obtained from a
DR screening program in The Netherlands. The screening population consisted of 400 diabetic subjects between
25-90 years of age. 40 photographs were randomly selected, 33 without and 7 with DR signs. 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 colour plane at 768 by 584 pixels. The FOV of each image is circular with a diameter of
approximately 540 pixels. 13
STARE (http://www.ces.clemson.edu/~ahoover/stare/) is one of the earliest and most cited test sets in the ARIA
literature, created for validating OD location. It consists of 31 images of healthy retinas and 50 images of retinas
with disease, acquired using a TopCon TRV-50 fundus camera at 35 field-of view, and subsequently digitized at
605×700 pixels in resolution, 24 bits per pixel (standard RGB). The nerve is visible in all 81 images, although
partially visible in 14 as appearing on the image border. In 5 images the nerve is completely obscured by
haemorrhaging. 17
MESSIDOR (http://messidor.crihan.fr/download-en.php) contains 1200 eye fundus colour digital images of the
posterior pole, acquired by 3 ophthalmologic departments using a colour video 3CCD camera on a Topcon TRC
NW6 non-mydriatic retinograph with a 45-degree FOV, 8 bits per colour plane and resolutions of 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 image. Currently there are no annotations (markings) on the images. Annotations by a single clinician for
OD diameter and the fovea centre, for the whole MESSIDOR set, have been made available by the Department of
Electronic, Computer Systems and Automatic Engineering, University of Huelva, Spain at
www.uhu.es/retinopathy/muestras/Provided_Information.zip.18
DIARETDB1 (http://www2.it.lut.fi/project/imageret/diaretdb1/) consists of 89 colour fundus images. 84 contain
at least mild non-proliferative DR signs (microaneurysms) and 5 are considered normal, not containing DR signs
according to all experts who participated in the evaluation. Images were captured using the same 50-degree FOV
digital fundus camera with varying imaging settings. The data correspond to a good (not necessarily typical)
practical situation, where images are comparable and can be used to evaluate the general performance of diagnostic
methods. 4 medical experts were asked to mark the areas related to the microaneurysms, haemorrhages, and hard
and soft exudates. Ground truth confidence levels, { 50%, ~100%}, represented the certainty of the
decision that a marked finding is correct, are included. 19
ROC (http://roc.healthcare.uiowa.edu/) is a set of 100 digital colour fundus photographs selected from a large
dataset (150 000 images) acquired at multiple sites within the Eye Check DR screening program (see ROC website
references), marked as gradable by the screening program ophthalmologists and including microaneurysms. Three
different types of images with different resolutions are included, acquired by a Topcon NW 100, a Topcon NW200,
or a Canon CR5-45NM and resulting in two differently shaped FOVs. All images are JPEG and compression was
set in the camera. The substantial black background around the FOV present in the original type II and III images
was cut off using specialized software. This complete set was randomly split into a training and a test set each
containing 50 images. Four retinal experts, all from the Department of Ophthalmology at the University of Iowa,
were asked to annotate all microaneurysms and all irrelevant lesions in all 100 images in the test and training set.
20
REVIEW (http://reviewdb.lincoln.ac.uk/) contains several subsets, with a mix of patients with disease and no
disease, and a focus on validating accurate measurements. It includes 16 images with 193 vessel segments,
demonstrating a variety of pathologies and vessel types. These image sets contain 5066 manually marked profiles.
Images were assessed by three independent experts, who marked the vessel edges. AREDS (Age Related Eye
Disease Study; https://web.emmes.com/study/areds/, http://www.areds2.org/) is a major clinical trial sponsored by
the National Eye Institute (NEI) at the National Institutes of Health (http://www.nei.nih.gov/amd/), involving
several US centers working on ARMD. The dataset includes several thousands of analog and digitized fundus
images showing various stages of AMD. Longitudinal studies were performed over ten years showing disease
progression. The images were graded by national centers for AMD as well as for lens opacity. The ground truth
does not include image-level delineation of drusen. The fundus photographs consist principally of 30° images
including stereo images centered on temporal margin of the disc and including an oblique view of the center of the
macula near the temporal margin of the field, stereo images centered on the center of the macula, and monoscopic
images centered temporal to the macula and including an oblique view of the center of the macula near the nasal
margin of the field. 21
ARIA (http://www.eyecharity.com/aria_online/) contains colour fundus images collected at St Paul’s Eye Unit and
the University of Liverpool, UK, as part of the ARIA project. All subjects were adult. All images were taken using
a Zeiss FF450+ fundus camera, originally stored as uncompressed TIFF files and converted to compressed JPG
files for WWW publication. All photographs were taken at a 50-degree FOV. Blood vessel masks created by
trained image analysis experts are available. The optic disk and fovea, where relevant, are outlined in separate file
sets. The data is organised into three categories, namely, age-related macular degeneration subjects, healthy
control-group subjects, and diabetic subjects.
BIOIMLAB (http://bioimlab.dei.unipd.it/Data%20Sets.htm) at the Univ of Padova, Italy, maintains a number of
publicly available datasets for several measurements, including vessel tortuosity ( 60 images from normal and
hypertense patients; 30 images of retinal arteries of similar length and calibre, 30 images of retinal veins of similar
length and calibre, Matlab data structures).
HEI-MED (http://vibot.ubourgogne.fr/luca/heimed.php) is a collection of 169 fundus images to train and test
image processing algorithms for the detection of exudates and diabetic macular edema. The images have been
collected as part of a telemedicine network for DR diagnosis. The images contain manual segmentation of
exudation, and include a machine segmentation of the vascular tree and optic nerve locations. The dataset contains
a mixture of ethnic groups, with roughly 60% African- American, 25% Caucausian, and 11% Hispanic.
Author Detected Lesion Method Sensitivity Specificity
Sopharak et al 4 Microaneurysms Mathematical Morphology 81.61% 99.99%
Amel et al 5 Hard Exudates Mathematical Morphology,
k-means clustering 95.92% 99.78%
SujithKumar et al 6 Microaneurysms Thresholding 94.44% 87.5%
Sopharak et al 7 Microaneurysms Naïve Bayes classifier 85.68% 99.99%
Datta et al 8 Microaneurysms Contrast Limited Adaptive Histogram
Equalization (CLAHE) 82.64% 99.98%
Zora et al 9 Hard Exudates Support Vector Machine (SVM) 97.5% 100%
Ravishankar et al 10 Microaneurysms/
Hemorrhage Mathematical Morphology 95.1% 90.5%
Pallawala et al 11 Microaneurysms Generalized eigenvectors 93% NA
Joshi et al. 14 Hemorrhages Mathematical Morphology 89.49 99.89
Jaafar et al. 16 Exudates Pure Splitting Technique 91.2 99.3
Sopharak
et al 4
Amel et
al 5
SujithKu
mar et al
6
Sopharak
et al 7
Datta et
al 8
Zora et al
9
Ravishan
kar et al
10
Pallawal
a et al
11
Joshi et
al. 14
Jaafar et
al. 16
Sensitivity81.61%95.92%94.44%85.68%82.64%97.50%95.10%93%89.49%91.20%
Specificity99.99%99.78%87.50%99.99%99.98%100%90.50%099.89%99.30%
81.61%
95.92%94.44%85.68%82.64%
97.50%95.10%93.00%89.49%91.20%99.99%99.78%
87.50%
99.99%99.98%100%90.50%
0
99.89%99.30%
0.00%10.00%20.00%30.00%40.00%50.00%60.00%70.00%80.00%90.00%100.00%
% Sensitivity, Specificity
Author
Chart Title
References:
1 IDF Diabetes Atlas, 8th edition. International Diabetes Federation, 2017.
2 Martins, Charles Iury Oliveira, F. N. S. Medeiros, Rodrigo MS Veras, Francisco Nivando Bezerra, and
R.M. Cesar Jr. “Evaluation of retinal vessel segmentation methods for microaneurysms detection.” In Image
Processing (ICIP), 2009 16th IEEE International Conference on, pp. 3365-3368. IEEE, 2009.
3 Frame, Allan J., Peter E. Undrill, Michael J. Cree, John A. Olson, Kenneth C. McHardy, Peter F. Sharp, and
John V. Forrester. “A comparison of computer based classification methods applied to the detection of
microaneurysms in ophthalmic fluorescein angiograms.” Computers in biology and medicine 28, no. 3
(1998): 225-238.
4 Sopharak, Akara, Bunyarit Uyyanonvara, and Sarah Barman. “Automatic microaneurysm detection from
non-dilated diabetic retinopathy retinal images using mathematical morphology methods.” IAENG
International Journal of Computer Science 38, no. 3 (2011): 295-301.
5 Amel, Feroui, Messadi Mohammed, and Bessaid Abdelhafid. “Improvement of the hard exudates detection
method used for computer-aided diagnosis of diabetic retinopathy.” International Journal of Image,
Graphics and Signal Processing (IJIGSP) 4, no. 4 (2012): 19.
6 SujithKumar, S. B., and Vipula Singh. “Automatic detection of diabetic retinopathy in Non-dilated RGB
Retinal fundus images.” International Journal of Computer Applications 47, no. 19 (2012): 26-32.
7 Sopharak, Akara, Bunyarit Uyyanonvara, and Sarah Barman. “Fine microaneurysm detection from non-
dilated diabetic retinopathy retinal images using a hybrid approach.” In Proc. of the World Congress on
Engineering, vol. 2. 2012.
8 Datta, Niladri Sekhar, Himadri Sekhar Dutta, Mallika De, and Saurajeet Mondal. “An effective approach:
image quality enhancement for microaneurysms detection of non-dilated retinal fundus image.” Procedia
Technology 10 (2013): 731-737
9 Zohra, Berrichi Fatima, and Benyettou Mohamed. “Automated diagnosis of retinal images using the
Support Vector Machine (SVM).” Faculte des Science, Department of Informatique, USTO, Algerie (2009).
10 Ravishankar, Saiprasad, Abhishek Jain, and Anish Mittal. “Automated feature extraction for early detection
of diabetic retinopathy in fundus images.” In Computer Vision and Pattern Recognition, 2009. CVPR 2009.
IEEE Conference on, pp. 210-217. IEEE, 2009.
11 Pallawala, P. M. D. S., Wynne Hsu, Mong Li Lee, and Say Song Goh. “Automated microaneurysm
segmentation and detection using generalized eigenvectors.” In Application of Computer Vision, 2005.
WACV/MOTIONS’05 Volume 1. Seventh IEEE Workshops on, vol. 1, pp. 322-327. IEEE, 2005.
12 Balint Antal and Andras Hajdu, “An Ensemble-Based System for Microaneurysm Detection and Diabetic
Retinopathy Grading,” IEEE Transactions on Biomedical Engineering, VOL.59, NO.6, June 2012
13 Niemeijer, Meindert, Joes Staal, Bram van Ginneken, Marco Loog, and Michael D. Abramoff, DRIVE
(2004): digital retinal images for vessel extraction,http://www.isi.uu.nl/Research/Databases/DRIVE.
14 Joshi Manisha Shivaram, Dr.Rekha Patil and Dr. Aravind H. S, “Automated Detection and Quantification
of Haemorrhages in Diabetic Retinopathy Images Using Image Arithmetic and Mathematical Morphology
Methods,” International Journal of Recent Trends in Engineering, Vol 2, No. 6, November 2009
15 Atsushi Mizutani, Chisako Muramatsu, Yuji Hatanaka, Shinsuke Suemori, Takeshi Hara, Hiroshi Fujita,
“Automated microaneurysm detection method based on double-ring filter in retinal fundus images,”
Medical Imaging 2009: Computer- Aided Diagnosis
16 Jaafar, Hussain F., Asoke K. Nandi, and Waleed Al-Nuaimy. “Detection of exudates in retinal images using
a pure splitting technique.” In Engineering in Medicine and Biology Society (EMBC), 2010 Annual
International Conference of the IEEE, pp. 6745-6748. IEEE, 2010.
17 http://www.ces.clemson.edu/~ahoover/stare/
18 Messidor (2004) Messidor: digital retinal images. http://messidor.crihan.fr/index-en.php.
19 Kauppi T, Kalesnykiene V, Kamarainen J-K, Lensu L, Sorri I,Raninen A, Voutilainen R, Uusitalo H,
Kalviainen H, Pietila J (2007) DIARETDB1 diabetic retinopathy database and evaluation protocol,
Technical report.
20 Niemeijer M, van Ginnerken B, Cree MJ, Mizutani A, Quellec G, Sanchez CI, Zhang B, Hornero R,
Lamard M, Muramatsu C, Wu X, Cazuquel G, You J, Mayo A, Li Q, Hatanaka Y, Cochener B, Roux C,
Karray F, Garcia M, Fujita H, Abramoff MD (2010) Retinopathy online challenge: automatic detection of
microaneurysms in digital color fundus photographs. IEEE Trans Med Imaging 29(1):185–195.
21 Al-Diri B, Hunter A, Steel D, Habib M, Hudaib T, Berry S (2008) REVIEW—a reference data set for
retinal vessel profiles. In: 30th Annual international IEEE EMBS conference, Vancouver, pp 2262–2265.
22 Quellec, Gwénolé, Mathieu Lamard, Pierre Marie Josselin, Guy Cazuguel, Béatrice Cochener, and
Christian Roux. “Optimal wavelet transform for the detection of microaneurysms in retina photographs.”
IEEE TMI 2007, no. 0694 (2007): 1.
23 Hussain F, Jaafar, Asoke K. Nandi and Waleed Al-Nuaimy, Automated Detection of Exudates In Retinal
Images Using A Split and-Merge Algorithm, European Signal Processing Conference, EUSIPCO, 2010
24 Acharya UR1, Lim CM, Ng EY, Chee C, Tamura T., “Computer-based detection of diabetes retinopathy
stages using digital fundus images,” Proceedings of the Institution of Mechanical Engineers, Part H: Journal
of Engineering in Medicine July 1, 2009, 223: 545-553
25 Godlin Atlas L, Kumar Parasuraman, “Detection of retinal hemorrhage from fundus images using ANFIS
classifier and MRG segmentation,” Biomedical Research 2018; 29 (7): 1489-1497
