A Seminar Report on

Image Processing Technique for Glaucoma Detection

Submitted by
Amit Kumar Singh
15CS02F
Himanshu Patel
15CS09F
Vikas Solanki
15CS25F
II Sem M.Tech (CSE)

in partial fulfillment for the award of the degree of

MASTER OF TECHNOLOGY
in

COMPUTER SCIENCE & ENGINEERING

Department of Computer Science & Engineering

National Institute of Technology Karnataka, Surathkal.

April 2016

Abstract

Glaucoma is a neurodegenerative disorder of the optic nerve, which causes partial loss of
vision. Glaucoma is recognized to be the second most common cause of blindness. Large number of
people suffers from eye diseases in rural and semi urban areas all over the world. Early detection
and treatment of glaucoma is hence important for the prevention of disease. Current diagnosis of
retinal disease relies upon examining retinal fundus image using image processing. This paper
describes the application of various image processing techniques for automatic detection of
glaucoma. The key image processing techniques to detect eye diseases include image fusion, image
segmentation, feature extraction, image enhancement, morphology, pattern matching, image
classification, analysis and statistical measurements. . Many autonomous glaucoma detection
systems analyze fundus image by calculating Cup to Disc Ratio (CDR) and categorize the image as
glaucoma or healthy. The proposed methodology provides a novel algorithm to detect glaucoma
using a fusion of CDR and hybrid textural and intensity features. Image categorization (glaucoma,
non-glaucoma, suspect) is done based on the results from both CDR and classifier. This fusion of
CDR with hybrid features has improved the sensitivity of system to 1, specificity 0.88 and accuracy
92%.

1. INTRODUCTION
Glaucoma is an ocular disorder which might leads to permanent vision
loss if not detected at an early stage. Origin of glaucoma is the increase in
intraocular pressure (IOP) which if continues, destroys the optic nerve. Optic
nerve is responsible for the information transfer from eye to brain vice versa. A
destruction in optic nerve leads to blindness. Glaucoma is often called a silent
thief of sight, as it has no clear symptoms at an early stage and if the
progression of glaucoma is not stopped at an early stage the severe
destruction to the optic nerve will result in incurable blindness. The World
Health Organization has declared Glaucoma to be the second largest cause of
blindness all over the world and it encompasses 15% of the blindness cases in
world which makes 5.2 million of the world's population [1] and the number is
expected to increase up to 80 million by 2020 [2]. An early and accurate
detection of glaucoma is one of the major requirements to stop glaucoma
progression.
3. PROPOSED METHODOLOGY
The proposed methodology as shown in Fig. 3, takes an input image, after
preprocessing the image Value plane is extracted for disc detection. After
binary conversion, image is post processed to remove noise and extra
protrusions. Finally the boundary of disc is smoothened using ellipse fitting.
Cup extraction starts from Green plane processing to remove veins from image.
Region growing is done on the resultant image to extract cup. Post-processing
is applied on the image and boundary is smoothened using circular fitting. CDR
is calculated and
compared with
values. Color and
texture based
the input image
calculated and
classified by the
healthy or
Final result
is concluded by
results from CDR

the

clinical

features of
are
classifier as
glaucoma.
merging the
and classifier

VI. COMPUTATION OF CUP TO DISK RATIO AND RESULTS

The vertical cup to disk ratio was calculated from the output
which was obtained from the circle detection stage. According
to the proposed algorithm, if the cup to disk ratio is more than
0.75 then the patients eye is affected by the
Glaucoma
disorder. If the ratio is less
than the 0.75 then it is a healthy
eye. These results have been compared with the medical
experts diagnostic results to check the effectiveness and
accuracy of the algorithm. The algorithm was tested over large
data set. Some of the test results are tabulated in Table 1.

Result
Analysis

Retinal image
sample No.

Sample 1
Sample 2
Sample 3

Detection
result

Detection Result and Comparison of Result

with Expert
Result

Detection Result (Proposed

Method)

Cup to disk
Ratio

Detection
Results from
Proposed
Algorithm

0.538

Healthy

0.742
0.580

Glaucoma

Medical
Expert's
Diagnostic
Resluts

Compared
Result

Healthy

Positive

Glaucoma

Healthy

Healthy

Positive
Positive

Sample 4

0.651

Healthy

Glaucoma

Negative

Sample 5

0.718

Glaucoma

Glaucoma

Positive

Healthy

Glaucoma

Negative

Sample 6

0.496

Sample 7

Positive

Sample 8

Positive

V. CONCLUSION
Glaucoma is a chronic disease and whose progression can only be stopped if
detected at an early stage. CDR measurement is an important structural
change that is being used in autonomous glaucoma detection systems. Varying
intensities and other noise in fundus images might not enable the autonomous
systems to extract cup and disc accurately, thus affects the CDR and
misinterprets the results. This might results in a failure to detect glaucoma.
Intensity and texture based features can also be used as a biomarker to classify
glaucoma and non-glaucoma images. Feature based classification of fundus
images is also being used in autonomous glaucoma detection systems.
However, such systems sometimes also misclassify the images resulting in

inaccurate glaucoma detection. Proposed algorithm improves the sensitivity of

system to 1 by combining the results from CDR and featured based
classification. The algorithm reduce the false negatives to 0 thus lowers the
chance to skip any glaucoma patient. This algorithm will help in telemedicine
as it screens out the images labeled as non-glaucoma by both CDR and
featured based classification and keeps the images considered as glaucoma by
both of the systems or either of the system, thus improving the sensitivity of
system. Proposed algorithm screens out Patients with glaucoma and suspects
and recommends to ophthalmologists for further tests and verification.