Review of Related Literature

The US Center for Disease Control and Prevention estimates that 29.1
million people in the US have diabetes and the World Health Organization
estimates that 347 million people have the disease worldwide. Diabetic
Retinopathy (DR) is an eye disease associated with long-standing diabetes.
Around 40% to 45% of Americans with diabetes have some stage of the
disease. Progression to vision impairment can be slowed or averted if DR is
detected in time, however this can be difficult as the disease often shows
few symptoms until it is too late to provide effective treatment.

Currently, detecting DR is a time-consuming and manual process that

requires a trained clinician to examine and evaluate digital color fundus
photographs of the retina. By the time human readers submit their reviews,
often a day or two later, the delayed results lead to lost follow up,
miscommunication, and delayed treatment.

Clinicians can identify DR by the presence of lesions associated with

the vascular abnormalities caused by the disease. While this approach is
effective, its resource demands are high. The expertise and equipment
required are often lacking in areas where the rate of diabetes in local
populations is high and DR detection is most needed. As the number of
individuals with diabetes continues to grow, the infrastructure needed to
prevent blindness due to DR will become even more insufficient.

The need for a comprehensive and automated method of DR screening

has long been recognized, and previous efforts have made good progress
using image classification, pattern recognition, and machine learning. With
color fundus photography as input, the goal of this competition is to push an
automated detection system to the limit of what is possible ideally resulting
in models with realistic clinical potential. The winning models will be open
sourced to maximize the impact such a model can have on improving DR
detection.

Diabetic Retinopathy Detection

https://www.kaggle.com/c/diabetic-retinopathy-detection

Diabetic retinopathy (DR) is damage to the retina caused by

complications of diabetes, which can eventually lead to blindness. It is an
ocular manifestation of diabetes, affects up to 80% of all patients suffering
from diabetes for over 10 years. Research in this direction indicated
retinopathy does not show any early warning signs, but include new blood
vessels formed at the back of the eye, vitreous hemorrhage, specks of blood
or spots, floating in a persons visual field, and blur vision. On funduscopic
exam, a doctor will see cotton wool spots, hard exudates because of which
the perimeter and spread area of the retinal blood vessels appear to be
deviating from normal observations.

Diabetic retinopathy is graded in to the different types based on the

severity. The earliest visible change to the retina is known as background
retinopathy. The capillaries in the retina become blocked, they may bulge
slightly and may leak blood or fluid. This type of retinopathy will not affect
the eyesight, but it needs to be carefully monitored. Mild non-proliferative
Retinopathy is the first stage of Diabetic Retinopathy. It is characterized by
the presence of dot and blotch hemorrhages and microannuerysms in the
Retina Microannuerysms are areas of swelling of the tiny blood vessels in the
Retina caused by the weakening of their structure. Moderate Non-
proliferative Retinopathy is the second stage of Diabetic Retinopathy. During
this stage, some of the small blood vessels in the retina may actually
become blocked, causing a decrease in the supply of nutrients and oxygen to
certain areas of the retina. In Severe Non-proliferative Retinopathy a
significant number of small blood vessels in the retina becoming blocked,
resulting in more areas of the Retina being deprived of nourishment and
oxygen. Lack of sufficient oxygen supply to the retina results in a condition
called Retinal Ischemia. Proliferative Retinopathy is a stage of Diabetic
Retinopathy that carries a significant risk of vision loss. The Retina responds
to a lack oxygen, or ischemia, by attempting to compensate for the reduced
circulation by growing new abnormal blood vessels. This process is called
neovascularization. When retinal neovascularization is present, DR
progresses into the Proliferative Retinopathy. The abnormal blood vessels are
extremely fragile and tend to break easily. This will result in hemorrhaging in
the Vitreous Humour that leads to scarring and ultimately to retinal
detachment.

Image Processing for Identifying Different Stages of Diabetic

Retinopathy

R. Manjula Sri, M.Raghupathy Reddy, K.M.M.Rao

 Image processing can do both reduce the workload of the screeners
and play a central role in quality assurance tasks. Therefore, there has been
an increase in the application of digital image processing techniques for
automatic detection of DR. For example color features on Bayesian statistical
classifier were used to classify each pixel into lesion or non-lesion classes.

Blood vessels, exudes, hemorrhages, microaneurysms and

maculopathy detection techniques, these were the detection techniques
yield most of the features which are used in automated DR detection
systems.

Digital fundus photography from the human eyes gives clear images of
the blood vessels in the retina. This method provides an excellent window to
the health of a patient affected by DR. The blood vessel structure was
obtained by subjecting the green component of the RGB fundus image to a
number of image processing algorithms.

Blood vessels were detected using two-dimensional matched filters.

Gray-level profile of cross section of blood vessel approximated by Gaussian
shaped curve. The concept of matched filter detection of signals was used to
detect piecewise linear segments of blood vessels after the vessel
approximation.

Vessel points in a cross section are fount with a fuzzy C-means

classifier. They have located and outlined blood vessels in images by the use
of a novel method to segment blood vessels that compliments local vessel
attributes with region-based attributes of the network structure.

Hayashi et al. have developed a computer aided diagnosis system to

assists physicians in detecting abnormalities associated with fundus images
of the retina. Their proposed system can detect blood vessel intersections
and it can identify abnormal widths in blood vessels.

Computerized system for both extraction and quantitative description

of the main vascular diagnostic signs from fundus images in hypertensive
retinopathy was presented. The features they have taken into account are
vessel tortuosity, generalized and focal vessel narrowing, presence of Gunn
or Salus signs.

A new system is proposed for the automatic extraction of the vascular

structure in retinal images, based on a sparse tracking technique was
proposed. Blood vessel points in a cross section are found by means of a
fuzzy c-means classifier. After tracking the vessels, identified segments were
connected using greedy connection algorithm. Finally bifurcations and
crossing were identified analyzing vessel end points with respect to the
vessel structure.

Blood vessel tracker algorithm was developed to determine the retinal

vascular network captured using a digital camera. The tracker algorithm
detects optic disk, bright lesions such as cotton wools spots, and dark lesions
such as hemorrhages. This algorithm identifies arteries and veins with an
accuracy of 78. 4% and 66.5% respectively.

Vallabha et al. have proposed a method for automated detection and

classification of vascular abnormalities in diabetic retinopathy. They detected
vascular abnormalities using scale and orientation selective Gabor filter
banks. The proposed method classifies retinal images as either mild or
severe cases based on the Gabor Filter outputs.

The microaneurysms in retinal fluorescein angiograms was identified

by first locating the fovea by sub-sampling image by factor of four in each
dimension. Subsequently, the image was subjected to median filtering with a
5 by 5 mask to reduce high-frequency components. Then the image was
correlated with a two-dimensional circularly symmetric triangular function
with modeled gross shading of the macula.

Blood-vessel detection algorithm bases on the regional recursive

hierarchical decomposition using quadtrees and post-filtration of edges to
extract blood vessels was studied. This method was able to reduce false
dismissals of predominately significant edges and faster in comparison to the
existing approach with reduced storage requirements for the edge map.

Li et al. have used the arteriolar-to-venular diameter ratio of retinal

blood vessels as an indicator of disease related changes in the retinal blood
vessel tree. Their experimental results indicate at 97.1% success rate in the
identification of vessel starting points, and a 99.2% success rate in the
tracking of retinal vessels.

A new method of texture based vessel segmentation to overcome this

problem was proposed. The Fuzzy C-Means (FCM0 clustering algorithm was
used to classify the feature vectors into vessel or non-vessel based on the
texture properties. They compared their method with hand-labeled ground
truth segmentation for five images and achieved 84.37% and 99.6%
specificity.
Algorithms for the Automated Detection of Diabetic Retinopathy
Using Digital Fundus Images: A Review

Oliver Faust et al.