IEEE - 40222

A COMPARISION OF OCT AND RETINAL FUNDUS

IMAGES FOR AGE-RELATED MACULAR
DEGENERATION
Dr Punal.M.Arabi1, Nanditha Krishna2, Vamsha Deepa.N3, Ashwini.V4, Prathibha.H.M5
#1
HOD, Dept of BME, ACSCE, Bangalore, India
arabi.punal@gmail.com
#2
Asst.Professor, Dept of BME, ACSCE, Bangalore, India
nanditha13@gmail.com
#3
Asst.Professor, Dept of BME, ACSCE, Bangalore, India
vamshi.deepa@gmail.com
*4
Student, fourth year B.E, Dept of BME ACSCE, Bangalore, India
ashu.ashwini1921@gmail.com
*5
Student, fourth year B.E, Dept of BME ACSCE, Bangalore, India
prathibha1895@gmail.com

Abstract— Retinal diseases are the most important related Macular Degeneration. In the Dry Age- related
issue in the medical field. Macular retinal commonly Macular Degeneration (DAMD), there will be existence of
known as drusen present in the eye is infected with Age- small yellow deposits in macula known as Drusen, which
related Macular Degeneration (AMD) and can be will be present under the RPE layer and converts its smooth
visualized with both optical coherence tomography curve shape into a bumpy one. In the Wet Age- related
(OCT) and Retinal fundus imaging. Earlier, Macular Degeneration (WAMD) there will an irregular
fundography (fundus imaging) was the most widely used growth of blood vessels under the layer of macula, this
test to detect AMD. But now a days, OCT is widely used condition is known as choroidal neo-vascularisation, where
for detecting AMD due to its ability to detect any small these blood vessels pours out the blood and fluid into retina,
changes in the retinal layers. This paper compares the this results in vision deterioration, which may further lead to
two imaging techniques of detecting AMD by the permanent blindness.In the normal eye the fundus area
extraction of blood vessels for fundus images and consists of retina, optic disc, fovea, macula and even the
extraction of RPE layer for OCT images using a set of blood vesssels, whereas in the affected retina various other
images acquired for twenty different subjects. These pathologies will also be present. Hence, the
images were studied for OCT and fundus images out of ophthalmologists can detect these various pathologies with
which ten subjects were affected by AMD and remaining the study of fundus region of the eye. There are several
ten subjects were of a healthy condition. methods used to detect and study the retinal disorders. They
are fundus photography (fundus imaging), Fundus
Fluorescein Angiography (FFA) and Optical coherence
Keywords--Optical Coherence Tomography(OCT) , Age-
related Macular Degeneration(ARMD) , Retinal tomography (OCT). Fundus imaging was the widely used
Pigmented Epithelium(RPE) , Fundus imaging. technique to detect the retinal disorders, earlier. One needs
proper perfection in order to segment the fundus images
manually. Now a day, OCT is being widely used technique
I. INTRODUCTION for detecting the retinal disorders due to its ability to detect
small changes in retinal layers. Optical Coherence
Age-related macular degeneration (AMD) is the major cause Tomography is very sensitive in the detection of AMD,
for a critical blindness and complete visions loss. It is a CNV, and CNV activity; however, it cannot fully replace
complex disease with both environmental and genetic Fluoresce in Angiography. It detects the macular diseases
factors. Age- related Macular Degeneration can be classified such as Macular degeneration, Macular Edema, Macular
into Dry Age- related Macular Degeneration and Wet Age- Hole, Central Serous Retinopathy, Choroidal Neo-
vascularisation and Pigment Epithelium Detachment. For

8th ICCCNT 2017

July 3 - 5, 2017, IIT Delhi,
Delhi, India
 IEEE - 40222

these retinal disorders, there will be no major symptoms the intensity value in a particular range. Philippa warr [7]
witnessed by the patients at the early state. But, OCT has an developed a neuro-prosthetic device used to augment
advantage over the other techniques for an early detection of function, literally enabling a normal animal to acquire a sixth
these retinal disorders based on some clinically observed sense. Yanling Ouyang et.al [8] proposed a method to
symptoms. As it shows a cross sectional region of macula determine the sensitivity of three-dimensional optical
and optic disc, it also gives an objective evaluation of coherence tomography (3D-OCT) versus single field non-
macular and ocular pathology. Removal of interferences and my driatic fundus photography (FP) for detection of a variety
noises is the major step before the detection of these retinal of retinal abnormalities. The approximate location of
diseases for a better and accurate diagnosis. A novel findings in relation to the 3D-OCT field was noted as in field
algorithm was implemented to automatically extract the and out of field. Findings from both modalities were
RPE layer. combined to form the gold standard for comparison for each
modality. In this study, sensitivities of 3D-OCT were higher
than non my driatic fundus images for overall detection of
II. RELATED WORKS retinal abnormalities or irregularities in each category. 3D-
OCT demonstrated good ability to detect most features;
C.M.G.Cheung et.al[1] represents a chronic disease with however, with limitation to intra retinal haemorrhage and
different stages, and variable rates of progression over time. pigment migration. Pratul P Srinivasan et.al [9] proposed an
People more than 65 years of age usually become the victims automated algorithm for the detection of retinal diseases
of Age- related Macular Degeneration hence it is commonly through optical coherence tomography (OCT) imaging
referred as Age related Macular Degeneration.For this which utilized multi-scale histograms of oriented gradient
condition, the central part of retina, which is known as descriptors as feature vectors of a support vector machine
macula becomes progressively poor. Samra Naz et.al [2] classifier. This was impactful tool for the remote diagnosis
proposed an algorithm, which automatically segment the of ophthalmic diseases. Yasser M Helmy et.al [10] proposed
retinal pigment epithelium (RPE) layer of the eye using a method for the classification of Diabetic Cystoid Macular
optical coherence tomography (OCT) images for the Edema (CME) based on optical coherence tomography
detection of drusen. The bottom layer of retina known as (OCT) findings. Some patients were diagnosed to have
RPE layer is extracted and analyzed by fitting a second order CME, as assessed by OCT, with measurement of the vertical
polynomial curve. The difference between both the curves size of the largest macular cyst and maximum macular
decides the class of the image. Giovanni Gregori et.al [3] thickness, and the relation between them. This concluded
proposed a novel algorithm which was used to quantitatively that OCT is a useful technique for quantitative measurement
assess drusen area and its volume. For each of the eye, the and also helped in better anatomical characterization and
standard deviations in the drusen area and the volume classification of CME. Qiang Chen et.al [11] showed a
measurements were calculated. Main Outcome of the normal retinal morphology and anatomical features
method is to measure Drusen area and volume. F. Mokwa described in an automatic drusen segmentation method for
et.al [4] in their proposed work had a comparison with the SD-OCT retinal images which The highly reflective and
color fundus photography (FP), fluoresce in angiography locally connected pixels located below the retinal nerve fibre
(FA), and spectral domain optical coherence tomography layer were used to generate a segmentation of the retinal
(SDOCT) for the detection of age-related macular pigment epithelium layer. This experimental evaluation of
degeneration (AMD), choroidal- neovascularisation (CNV), results demonstrated that the method was effective for
and CNV activity. They concluded that, CNV lesions and
drusen segmentation.
activity may be missed by FP alone, but FP may help
identifying drusen and pigmentary changes. Taimur Hassan
.
et.al [5] presented a paper which gives a detailed comparison III. METHODOLOGY
between Fundography and OCT imaging technology for the
detection of Macular Edema using their own dataset and In the proposed method, four sets (twenty images in total) of
concluded that, OCT images provide an objective evaluation healthy fundus, healthy OCT, retinal fundus and OCT
of early macular edema as compared to Fundography. images of Age related Macular regeneration are taken for
Nimishal, Rana Gi et.al[6] proposed a method of the pre- experimentation. These images are first acquired and then
processing and segmentation. Pre-processing is done on the converted to gray scale. Further the converted gray scale
original image using Gamma Normalization which brings image undergoes the Pre-processing stage for the removal of

8th ICCCNT 2017

July 3 - 5, 2017, IIT Delhi,
Delhi, India
 IEEE - 40222

Gaussian noise which are present in the images by

preserving the edges of the images and enhancing the retinal
layers.
Pre-Processing stage undergo 2 steps- Fig2b. Set of healthy OCT images
1. Image filtering by bilateral filter.
2. Image enhancement by Contrast stretching.
After Pre-processing , Binary thresholding is performed to
partition the image into a foreground and background. This
Fig2c. Set of AMD fundus images
technique helps in isolating objects by converting gray scale
images into binary images. The next step after binary
thresholding is to apply intensity based thresholding
technique to extract blood vessels in fundus images and RPE
layer in OCT images which is of high pixel value from the
Fig2d. Set of AMD OCT images
overall image.
Fig 2a, 2b, 2c ,2d - shows the set of healthy fundus, healthy
OCT, AMD fundus and AMD OCT images.
Image Acquisition
Original image pre-processed image Binary thresholding

RGB to Gray
Conversion

Pre-processing
Blood vessel extraction

Binary Thresholding Fig3a. Experimental result of healthy fundus image1

Original image pre-processed image Binary thresholding

Blood Vessel and RPE

layer Extraction

Healthy Unhealthy
RPE layer extracted
images images (AMD)
Fig3b. Experimental result of healthy OCT image1

Fig1. Flow Diagram Original image pre-processed image Binary thresholding

Fig1- shows the flow diagram of the proposed method

IV. RESULTS

Blood vessel extraction

Fig3c. Experimental result of AMD Fundus image1
Fig2a. Set of healthy Fundus images

8th ICCCNT 2017

July 3 - 5, 2017, IIT Delhi,
Delhi, India
 IEEE - 40222

Original image pre-processed image Binary thresholding Calculation of Average Mean for OCT images

Threshold (T) = m3+m4

2
T = 4596+742 = 2669
2
Table-1 shows the number of pixels for healthy and AMD
fundus images
RPE layer extracted Table-2 shows the number of pixels for healthy and AMD
Fig3d. Experimental result of AMD OCT image1 OCT images

Fig 3a, 3b, 3c 3d- shows the set of experimental results of V. DISCUSSION
healthy fundus1, healthy OCT1, AMD fundus1 and AMD
OCT1 images. A set of twenty images as five in each category of healthy
Table-1 Number of pixels in fundus images fundus, OCT images and AMD affected fundus and OCT
images are taken for analysis. Bilateral filtering is performed
on retinal fundus imaging and OCT imaging images for the
Healthy fundus AMD Fundus
Sl.No removal of Gaussian noise. Upon inspecting the number of
images (m1) images (m2)
pixels present after binarization (Table-1) are in few hundred
1 241 2296 in healthy fundus images whereas AMD fundus images have
2 240 5586 them in thousands which means that extracted blood vessels
3 727 1833 are more in number in AMD fundus images compared to
normal fundus images. The increase in number of blood
4 116 3041
vessels is an important biomarker for AMD. From Table-2,
5 490 4889 the number of pixels present in the extracted RPE layer for
Mean 319 3529 healthy images is in thousands as the RPE layer extracted is
Average thick and continuous whereas AMD OCT images have
Mean 1924 pixels in hundreds as RPE layer extracted is thin and
(Threshold) discontinuous which is a disease marker.

Calculation of Average Mean for fundus images From the results obtained a decision rule can be framed as-

Threshold (T) = m1+m2 (i)For fundus images,

2 If image pixels< threshold 1924, healthy image
T = 319+3529 = 1924
If image pixels> threshold 1924, AMD affected image
2
(ii) For OCT images,
Table-2 Number of pixels in OCT images If image pixels< threshold 2669, AMD affected image
If image pixels> threshold 2669, healthy image
Healthy OCT AMD OCT By the proposed method, fundus images have an overall
Sl.No accuracy of 90% and OCT images have an overall accuracy
images (m3) images (m4)
of 60%.This method works well for fundus images, but
1 1472 877
sophisticated image processing techniques can be used to
2 1591 549 improve the accuracy of OCT images.
3 2521 958
4 15807 449 VI. CONCLUSION
5 1591 881
The proposed method compares fundus images and OCT
Mean 4596 742 images for AMD affected eye. The increase in number of
Average blood vessels is an important biomarker for detection of
Mean 2669 AMD in fundus images; whereas RPE layer is an important
(Threshold) biomarker for detection of AMD in OCT images. The

8th ICCCNT 2017

July 3 - 5, 2017, IIT Delhi,
Delhi, India
 IEEE - 40222

proposed method shows that there is a possibility of [5] Taimur Hassan1, M. Usman Akram2, Bilal Hassan3,
developing an automated classification technique to classify Ammara Nasim4 and Shafaat Ahmed Bazaz5, “Review of
AMD. The results obtained show that the proposed method OCT and Fundus Images for Detection of Macular Edema”
is more suitable for fundus images than for OCT images.
Since, OCT imaging is superior to fundus imaging as it has [6] Nimishal, Rana and Inderjeet Kaur “ Segmentation of
an ability to detect smallest changes in the retina by Retinal Area by Adaptive SLIC Superpixel”
acquiring the cross sectional images with semi histologic
resolution, it defines the location, nature of the changes in
[7] PHILIPPA WARR "Implant gives rats’ sixth sense for
retina and adjacent structures ,also evaluates the thickness of
infrared light"
retina and surrounding structures.The proposed method
should be improved with additional feature identification [8] Yanling Ouyang, Florian M Heussen, Pearse A Keane,
and by sophisticated image processing techniques to Srinivas R Sadda, and Alexander C Walsh. “The retinal
improve the accuracy for OCT images. This may be disease screening study: Retrospective comparison of
achieved by sampling the image and then analysing the nonmydriatic fundus photography and three-dimensional
number of pixels present in the samples and also probing optical coherence tomography for detection of retinal
their inter sample relations to grade the images into healthy irregularities comparison of non mydriatic fundus
and affected AMD and also further grade the images into photography”
early AMD and severely affected AMD.
[9] Pratul P Srinivasan, Leo A Kim, Priyatham S Mettu,
ACKNOWLEDGMENT Scott W Cousins, Grant M Comer, Joseph A Izatt, and Sina
Farsiu. “Fully automated detection of diabetic macular
The authors thank the Management and Principal of ACS
edema and dry age-related macular degeneration from
College of Engineering, Mysore road, Bangalore for
optical coherence tomography images”
permitting and supporting to carry out the research work.

REFERENCES [10] Yasser M Helmy, Heba R Atta Allah, “Optical

Coherence Tomography classification of diabetic cystoid
[1] C.M.G.Cheung & T.Y.Wong “Is age related macular macular edema”
degeneration a manifestation of systemic disease?New
prospects for early intervention & treatment” Journal of [11] Qiang Chen, Theodore Leng, Luoluo Zheng, Lauren
Internal medicine Kutzscher, Jeffrey Ma, Luis de Sisternes, and Daniel L
Rubin “Automated drusen segmentation and quantification
[2] Samra Naz, Aneeqa Ahmed, M. Usman Akram, Shoab in sd-oct images”
A. Khan “Automated Segmentation of RPE Layer for the
Detection of Age Macular Degeneration Using OCT
Images”

[3] Giovanni Gregori, Fenghua Wang, Philip J Rosenfeld,

Zohar Yehoshua, Ninel Z Gregori, Brandon J Lujan, Carmen
A Puliafito, and William J Feuer, “Spectral domain optical
coherence tomography imaging of drusen in nonexudative
age-related”

[4], F. Mokwa, Tina Ristau, Pearse A. Keane, Bernd

Kirchhof,1 Srinivas R. Sadda, and Sandra Liakopoulos ,
“Grading of Age- Related Macular Degeneration:
Comparison between Color Fundus Photography,
Fluorescein Angiography, and Spectral Domain Optical
Coherence Tomography”

8th ICCCNT 2017

July 3 - 5, 2017, IIT Delhi,
Delhi, India