The human eye is
a complex organ and is of spheroid structure that lies in a bony cavity at the
front of the head. The sclera, the choroids and the retina are the three
covering layers in the eyeball. The outermost layer of eye tissue is called sclera.
The second layer of the tissue underneath the sclera is the choroid which
consists of dense pigments and blood vessels that nourish the tissues. The size
of the pupil is regulated by iris. The inner surface of the eye known as the
retina contains rods and cones that serve the task of detecting the intensity
and the frequency of the incoming light. The light energy is converted to
electric signals by the retina and sent to the brain through optic nerves. Fig.
1 illustrates the cross-section of a human eye and points out its major
Fig. 1 Structure of a Human Eye
Fig. 1 Structure of a Human Eye Retina consists of 120
million rods and 6 million cones which send nerve impulses to the brain and
travel through a network of nerve cells. There are one-million neural pathways
from the rods and cones to the brain.
Structure of Retina
Fig. 2 Structure of Retina Macula is present at the center of
the retina of the human eye and is a small and light sensitive layer of tissue
at the back of the retina. It is responsible for sharp, clear and detailed
central vision. The fovea is found at the center of the macula and consists of
light sensitive cells. Optic Disc (OD) is the brightest region on the retina
and is usually round at the back of the inside of the eye.
Retinopathy (DR) is the most common cause of new cases of blindness among
adults aged between 20 and 74 years. The blood vessels of the retina get
damaged and lead to leakage of blood and fluid into the retinal surface. There
is vision impairment when the nerve cells are damaged. Initially there are no symptoms
for DR and a vision problem would occur only during later stages. If the blood
sugar level is not controlled, it would weaken and damage the blood vessels in
the retina causing abnormal auto regulation. Hence the vessels do not constrict
adequately due to sudden increase of blood flow. There is an increase in shear
stress on the vessel walls which damage the vessels and leads to vascular
complications. It is irreversible and would eventually direct to blindness if
ignored. Therefore, early diagnosis is important to stop the disease from
worsening. WHO lists DR as one of its priority eye diseases and includes it
into its action plan for the prevention of inevitable blindness and visual
retinal disease starts from mild Non-Proliferative Diabetic Retinopathy (NPDR)
which is characterized by abnormal lesions and changes in blood vessels to
moderate and severe NPDR which are characterized by vessel closure. It
progresses from NPDR to Proliferative Diabetic Retinopathy (PDR) and is characterized
by the growth of new blood vessels. Due to the presence of microvascular
complications, loss of vision and blindness can occur in diabetic patients.
In the early
stages of DR, patients may not have symptoms whereas in the more advanced
stages of the disease, the patients suffer from floaters, blurred vision,
distortion and progressive visual acuity loss. The clinical features of DR
include the following.
(MAs): One of the first clinical signs of the presence of DR is the appearance
of MAs. They are small round red dots on the retinal surface and are less than
125 ?m in size. MAs have sharp margins and are less than the diameter of the
major optic veins.
addition to leaking blood, the retinal blood vessels also expand and leak
lipids and proteins causing small bright dots called exudates. They appear as
random yellow color patches of different shapes, sizes and positions. The major
causes of vision loss in NPDR are due to the presence of these exudates and are
visible in retinal surface.
C. STAGES OF DIABETIC RETINOPATHY
Based on the
presence of clinical features, DR is classified into four stages namely Mild
NPDR, Moderate NPDR, Severe NPDR and PDR.
1. Mild NPDR:
During this earlier stage, MAs occur. These MAs are small red dots due to
weakening of blood vessels and are characterized by the presence of “dot” and
“blotch” MAs and HAs in the retina during eye examination. Generally 40% of
diabetic patients have mild signs of DR.
NPDR: As the disease develops, some blood vessels that feed the retina are
blocked. The characteristic features of this stage are HAs and hard exudates.
PDR would develop soon for the patients with moderate NPDR.
3. Severe NPDR:
The blood vessels in the retina are blocked and hence they send signals to the
body to grow new blood vessels for nourishment. During this stage, severe
intra-retinal MAs, HAs, venous beading and intra-retinal micro vascular
abnormalities are present in at least two quadrants of the retina. About 50% of
severe NPDR would progress to PDR in one year.
Screening is a
successful method for the detection of DR at an initial stage (Raman et al
2014). It is used for diagnosis and timely treatment for the patients who are
affected by DR and also helpful in reducing the incidence of vision and
blindness. It is difficult to identify the disease during the early stages
since there are no symptoms, but the treatment would save the sight if the DR
is detected early. It is essential to screen diabetic patients regularly to
prevent them from retinal disease and to reduce the rate of the disease progression
(ETDRS report 1991 & Bresnick et al 2000).
NPDR Moderate NPDR Severe NPDR PDR
3 Normal and Abnormal Fundus Images
retinal abnormalities and various stages of DR are shown in Fig. 3. Different
stages of DR can be identified by screening the diabetic patients regularly.
The process of Microaneurysms Hemorrhages Neovascularisation Blood vessels
Optic disc Exudates Hemorrhages grading DR from retinal image is tedious which
would further lead to errors.
mass screening of retinal images, the digital imaging techniques are more
useful than the ophthalmoscopic techniques. With the rapid development of
technology, computerized techniques for screening are available for the
detection and classification of retinopathy stages. It mainly focuses on the
development of retinal image diagnosis system for the early detection and
classification of DR
Keerthi Ram10 et al, (2010) has proposed that the presence of
tiny microaneurysms is usually an early sign of diabetic retinopathy. A
successive rejection-based strategy is proposed to progressively lower the
number of clutter responses. The processing stages are designed to reject
specific classes of clutter while passing majority of true MA’s, using a set of
specialized features. Results of extensive evaluation of the proposed approach
on three different retinal image datasets is reported, and are used to
highlight the promise in the presented strategy.
Shree Divya R and Balaji, (2014)15 has proposed a
work to detect several defects of the human eye like hemorrhages, exudates.
Among them, microaneurysms should be considered as one of the severe condition
for the early blindness. A new technique called neural network taken for
presentation, helps to detect and determine the severity of microaneurysms
which would be able to give a better performance than the existing techniques.
Meindert Niemeijer11 et al, (2005) have presented
that the robust detection of red lesions in digital color fundus photographs
for diabetic retinopathy. In this proposal, a novel red lesion detection method
is presented based on a hybrid approach, combining prior works by Spencer and
Frame with two important new contributions. The first contribution is a new red
lesion candidate detection system based on pixel classification. After removal
of the connected vasculature the remaining objects are considered possible red
lesions. Second, an extensive number of new features are added to those
proposed by Spencer–Frame. The detected candidate objects are classified using
all features and a k-nearest neighbor classifier. When determining whether an
image contains red lesions the system achieves a sensitivity of 100% at a
specificity of 87%.
Diego Marin et al, (2011)3 presented a new supervised method
for blood vessel detection in digital retinal images. This method uses a neural
network (NN) scheme for pixel classification and computes a 7-D vector composed
of gray-level and moment invariants-based features for pixel representation.
The method was evaluated on the DRIVE and STARE databases. Its application to
this database outperforms all analyzed segmentation approaches. Its
effectiveness and robustness with different image conditions, together with its
simplicity and fast implementation, make this blood vessel segmentation
proposal suitable for retinal image computer analyses such as automated
screening for early diabetic retinopathy detection.
Oliver Faust et al, (2012)13 has analyzed the diabetic
retinopathy based on features, such as blood vessel area, exudates,
hemorrhages, microaneurysms and texture. In this paper they review algorithms
used for the extraction of these features from digital fundus images. The
classifications efficiency of different DR systems is discussed. Most of the
reported systems are highly optimized with respect to the analyzed fundus
images. Therefore a generalization of individual results is difficult. However,
this review shows that the classification results improved has improved
recently, and it is getting closer to the classification capabilities of human
Bob Zhang et al, (2009)2 has presented a new novel approach
to the computer aided diagnosis(CAD) of the diabeticretinopathy(DR) a common
and severe complication of long-term diabetes. Since microaneurysms are regarded
as the first signs of DR. In contrast to existing algorithms, a new approach
based on multi-scale correlation filtering(MSCF) and dynamic thresholding is
developed. This consists of two levels, microaneurysm candidate detection
(coarselevel) and true microaneurysms classification(finelevel). The approach
was evaluated based on two public datasets: ROC and DIARETDB1.
Balint Antal and Andras Hajdu, (2012)1 proposed that
the reliable microaneurysm detection in digital fundus images. They propose a combination
of internal components of microaneurysm detectors, namely preprocessing methods
and candidate extractors. They have evaluated our approach for microaneurysm
detection in an online competition, and also on two other databases. They also
tested the proposed method for this task on the publicly available Messidor
database, where a promising AUC 0.90±0.01 is achieved in a “DR/non-DR”-type
classification based on the presence or absence of the microaneurysms.
Istvan Lazar and Andras Hajdu, (2013)8 introduced a
method realizes MA detection through the analysis of directional cross-section
profiles centered on the local maximum pixels of the preprocessed image. Peak
detection is applied 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 orientation of the
cross-section changes constitute the feature set that is used in a Bayes
classification to exclude spurious candidates.They give a formula for the final
score of the remaining candidates, which can be thresholded further for a
binary output. The proposed method has been tested in the Retinopathy Online
Challenge. They also present the experimental results for a private image set
using the same classifier setup.
Additional attributes may be abstracted as raw features that
are incorporated after the convolutional phase of the network. Finally, we
compare its performance against existing approaches on the challenging problem
of detecting lesions in retinal images.
The main objective is early detection of microaneurysm’s from
Introducing methods for improving the processing time.
To provide automated mechanisms in detecting MA’s and
classifying the disease level.
The method proposed
earlier aims to increase the number of true positives in the first phase of
detection process. Hence a framework for selecting an optimal combination of
preprocessing and candidate extractors are used. The optimal combination of
preprocessing have been used. Thus due to using many sets of preprocessing
combinations processing time may be bit more.
Another method in MA
detection is generally by successive clutter rejection method that is proposed
in order to lower the number of clutter response. That involves two stages of
rejection to filter out the unwanted signals which interfere the MA detection.
The drawback is that also it includes two sets of classification that will cause
the system to be complex to work.
Retinal vascular is
segmented utilising the contrast between the blood vessels and surrounding
background. Hemorrhage candidates were detected using density analysis and
bounding box techniques. Finally, classification of the different stages of eye
disease was done using Random Forests technique based on the area and perimeter
of the blood vessels and hemorrhages. Accuracy assessment of the classified
output revealed that normal cases were classified with 90% accuracy while
moderate and severe NPDR cases were 87.5% accurate.
method is proposed for the microaneurysms detection. Initially, the method
performs a segmentation of the retinal vasculature and defines a global set of
microaneurysms candidates, using both coarser and finer scales. Using the finer
scales, a set of microaneurysms candidates are analysed in terms of shape and
size. Then, a set of Gaussian-shaped matched filters are used to reduce the
number of false microaneurysms candidates. Each candidate is labeled as a true
microaneurysm using a new neighborhood analysis method. The proposed algorithm
was tested with the training Retinopathy Online Challenge (ROC) dataset,
revealing a 47% Sensitivity with an average number of 37.9 false positives per image. The results are not satisfied.
The proposed work is
that the input images are obtained from DRIVE database. Initially 20 images are
used from the database. The main aim is to detect the diabetic retinopathy at
its early stage. The first step is the preprocessing of the image by two
methods namely, Gaussian filtering and Contrast-limited adaptive histogram
equalization. Then the wavelet transformation is applied to that preprocessed
images. The special type of transformation used here is the Mathieu
transformation, that is hoped to give better performance.