extraction of macular disease area using regional ... technique for human retinal optical ......

Applied Mathematical Sciences, Vol. 9, 2015, no. 129, 6437 - 6448

HIKARI Ltd, www.m-hikari.com

http://dx.doi.org/10.12988/ams.2015.53291

Extraction of Macular Disease Area Using Regional

Statistics Technique for Human Retinal Optical

Coherence Tomography (OCT) Image

Mohd Fadzil Abdul Kadir1*, Abd Rasid Mamat1, Azrul Amri Jamal1,

Shinji Tsuruoka2, Haruhiko Takase3, Hirobaru Kawanaka3,

Fumio Okuyama4 and Hisashi Matsubara5

1Faculty of Informatics and Computing, University Sultan Zainal Abidin, 22000

Tembila, Besut, Terengganu, Malaysia

*Corresponding author

2Graduate School of Regional Innovation Studies

Mie University, Tsu, Mie, Japan

3Graduate School of Engineering, Mie University, Tsu, Mie, Japan

4Faculty of Medical Engineering, Suzuka University of Medical Science

Suzuka Mie, Japan

5Graduate School of Medicine, Mie University, Tsu, Mie, Japan

Copyright © 2015 Mohd Fadzil Abdul Kadir et al. This article is distributed under the Creative

Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any

medium, provided the original work is properly cited.

Abstract

Optical Coherence Tomography (OCT) has emerged as a new technology that enables

high-resolution cross-sectional images of the retina for identifying, and quantitatively

assessing of the retina disease. Quantitative information of retina is needed for tracking

progression of ocular disease and evaluatesthe efficacy of treatment. In this paper, we

propose a new border tracking procedure using regional statistics (BTPRS) to extract an

abnormal area that specified by medical doctor. This procedure uses a combination of

regional statistics and border tracking method. The objectives of this research are to extract

6438 Mohd Fadzil Abdul Kadir et al.

the abnormal area in human retina from optical coherence tomography images and to

improve the extraction percentage.This research uses 128 pieces of 2 dimensional OCT

retinal image of one drusenpatient, and 128 pieces of 2dimensional OCT retinal image

of a diabetic macular edema (DME) patient. The part of the diseases are specified by a

medical doctor. Results show that the regional statistic border tracking method provided

the highest extraction of rate percentage and can extract the abnormal area in

bothconditions, white and black. In this paper, we will focus on the abnormal area at

macular part.This research will provide more useful information to medical doctor and

patient for informed consent. We hope that thisprocedure willbe added in the

commercial OCT unit to evaluate the degree of disease and response to the treatment.

Keywords: Optical Coherence Tomography·Image Scanning ·Border Tracking

·Regional Statistics

1. Introduction

Optical coherence tomography (OCT) is an imaging technology performed non-

invasivehigh resolution cross-sectional images of transparent and translucent structures

[1]. OCT imaging is similar to ultrasound imaging except it uses light instead of sound.

In ophthalmology OCT canprovide the structure imaging of retinal morphology at an

intraretinal level. The quantitative information of retinal thickness can be obtained in

order to assessing retinal diseases. The precise visualization of retinal structure

morphology is the most critical in diagnosis retinal diseases.Therefore the needs of

retinal imaging using OCT devices have been growing [2].

Fig. 1: Example of OCT retina image

Measurement the retinal structure layers and thickness performed by directing the light

beam from optical source onto the retina. The back reflected light produced by the

accident with the retinal layers contained interference beams. The OCT images

generated by scanning reflected beams, producing two-dimensional data set of gray

scale images (Fig. 1) [2, 3]. Fig.2 shows the example of normal and abnormal retina

OCT image. The generated abnormal area in the OCT images can be displayed in two

different colours; white (Fig. 3(a)) or black (Fig. 3(b)) in color.

Extraction of macular disease area using regional statistics technique 6439

(a) (b)

Fig. 2: (a) Example of normal retina. (b) Example of abnormal retina.

Fig. 3: White and black abnormal areas.

Retinal nerve fiber layer thickness is important in early diagnosis of retinal disease.

Quantitative information between inner limiting membrane (ILM) and retinal pigment

epithelium (RPG), retinal layer structure and abnormal area in the retina can be used to

access the retinal disease and monitoring the treatments process [4, 5]. Yagi et al. [6]

proposed a method of extraction ILM and RPG using conventional morphological

technique. However, the proposed method still cannot extracts the layer lines when the

original or input OCT images having some disappearance points. To improve Yagi`s

method, Yamakawa et al [7] used one directional activenet (ODAN) to extract the

layers. The correction rate of extraction increased (Fig. 4) to 98.7% for normal retinal

image and 91.0% for abnormal. Kodama et al [8] proposed extraction method using

statistical technique to measure the number of layer boundaries in retina in order to eva-

(a) abnormal area displayed

in white colour.

(b) abnormal area displayed

in black colour.


luate the size of retinal disease in horizontal direction. All of these methods have some

problems are i) cannot extract the abnormal area in the image that contained retinal

layer damages, ii) only can detect the black pixels in gray scale OCT images and iii)

cannot specify the abnormal area by a medical doctor.

Fig. 4: Extracted borders by Yamakawa`s method

Image scanning and border tracking method [9] proposed to fulfill the medical doctors’

need that they can select the abnormal area by themselves, but cannot extract the

abnormal area properly when the OCT image contained retinal layer damage. Regional

statistics scanning method (from here on referred to as: our previous procedure) [10]

was proposed to improve image scanning and border tacking method has a problem that

cannot extract precisely at the small part in abnormal area.

Traditional extraction methods evaluate the retinal disease and treatment effectiveness

by measure the thickness of retinal layer. In this paper we proposed a new border

tracking procedure using regional statistics (BTPRS) that only extract the abnormal area

in the retinal images. These methods will give more precise results and can reduce the

processing time.

2. Border Tracking Procedure Using Regional Statistics (BTPRS)

Our previous method (see Fig. 5(a)) starts with median filter image smoothing,

followed initial pixel selection, and finally the area extraction. In border tracking

procedure (see Fig. 5(b)), starts with initial pixel selection, followed by two threshold

binarization and finally the border extraction.

In this paper, the sample of 128 pieces of 2-dimensional OCT image about one retina of

a drusen patient and 128 pieces of 2-dimensional OCT image about one retina of a

DME patient are digitalized to a pixel size of 6μm x 6μm, 16-bit gray scale with

resolution 512 x 480 pixels.


(a)Previous method (b) Border tracking method.

Fig. 5: Previous and Border tracking method comparision

2.1. Selection of Initial Pixel and Initial Region

The initial position of pixel (i) (see Fig.6(a) and Fig.6(b)) will select by a medical

doctor in the abnormal area in OCT retina image using the computer mouse on the

monitor display.

(a) (b) (c)

Fig. 6: (a) The pointed initial pixel, (b) Pointed initial pixel surrounded by initial

region, (c) Binary image


An initial region is defined as 25 pixels (5 × 5 pixels) surrounding the selected initial

pixel, i. The mean μand standard deviation σof gray level of pixels in the initial region

will be calculated, and stored in program memory.

2.2. Binarization Using Two Thresholds

The calculated values of mean and standard deviation are used to generate two threshold

formulas. The system will calculate mean 𝜇𝑥 and standard deviation 𝜎𝑥 of all regionsand

determines that every pixelisbelonging to the abnormal area or not using equation (1).

(μ – aσ) <𝜇𝑥< (μ + aσ) (1)

where a is the constant

If the calculated values of mean and standard deviation are in the range, the gray level

values of the center pixel of that region will changes to 255 or white in color. If the

calculated values of mean and standard deviation are out of the stated range, the gray

level value of the center pixel of that regionwill changes to 0 or black in color. The

system will scan all regions from top to bottom of the image and changes the raw gray

scale image to binary image. (see Fig. 6(c)).

2.3. Boundary Extraction

From the initial pixel i, the scanning process to detect the abnormal area is started, and

the position of the next pixel is determined by moving the initial pixel to (i + 1) to the

top of the image. The pixel will move one by one until finds the different color of pixel;

from white to black. The border following process will used four points of contact

algorithm to extract the abnormal area (see Fig. 7).

(a) Border tracking image (b) Enlarge of border tracking

image

Fig. 7: Border tracking image from the proposed method


3. Results and Discussions

We determined four values for a that stated in the benchmark formula. The values are

0.5, 1.0, 1.5, and 2.0. In this experiment, only 35 pieces of image from drusen sample

and 36 pieces of image from DME sample that contained the abnormal area at the

macular part were used to extract border line. Below are the sample of binary and

border tracking image from experiment results.

Fig. 8 (a-1 to a-3) and (b-1 to b-3) show the results using proposed BTPRS to the OCT

images that contained abnormal area in white colour.

(a-1) Original image (a-2) Binary image (a-3) Border tracking

image

(b-1) Original image (b-2) Binary image (b-3) Border tracking

image

Fig. 8: Experiment results for abnormal area in white colour.

Fig. 9 (a-1 to a-3) and (b-1 to b-3) below show the results using proposed BTPRS to

the OCT images that contained abnormal area in black colour.


(a-1) Original image (a-2) Binary image (a-3) Border tracking

Image

(b-1) Original image (b-2) Binary image (b-3) Border tracking

image

Fig. 9: Experiment results for abnormal area in black colour.

Table 1 shows that the extracted abnormal area from drusen OCT image samples got

the highest rate of successful when we set the a value at 1.0.

Table 1. Extraction results from drusen OCT images

Table 2 shows that the extracted abnormal area from DME OCT image samples got the

highest rate of successful when we set the a value at 1.5.

a

(value)

Extracted Failed Successful Rate

(%)

0.5 22 13 62.9

1.0 26 9 74.3

1.5 25 10 71.4

2.0 23 12 65.7


Table 2. Extraction results from DME OCT images

Fig. 10 show the experiment results using proposed BTPRS onto the same sample of

OCT image that contained retinal layer with a damage area compare to others method

that proposed before. The result shows the effectiveness of the proposed BTPRS

compare to the image scanning method (Fig. 10(b)), border tracking method (Fig.

10(c)), and previous procedure (Fig. 10(d)). Image scanning and border tracking

method cannot extract the abnormal area when the original OCT image has structure

damage at the retinal layer. Even though the regional statistics image scanning can

extract the abnormal area, but the extracted result is not precise compare to the proposed

method (Fig. 10(f)).

(a) Original image (b) Image scanning method (c) Border tracking

Method

(d) Our Orevious method (e) Binary image by two (f) Extracted border

using threshold BTPRS

Fig. 10: Comparisons results between previous procedures and proposed BTPRS

Even though the proposed BTPRS can solve the problem in extracting abnormal area

that contained damages in the retinal layer image, but in some cases it still need impro-

a

(value)

Extracted Failed Successful Rate

(%)

0.5 25 11 69.4

1.0 28 8 77.8

1.5 30 6 83.3

2.0 27 9 75.0


vements to provide more higher extraction rate. Fig. 11 and Fig. 12 show the failure

example of extraction images from two different conditions of abnormal area.

(a) Original image (b) Binary image (c) Border tracking image

Fig. 11: Example of failure extraction for abnormal area in white colour.

Fig. 11(a) to Fig. 11(c) show the proposed BTPRS cannot extract half of the abnormal

area (in the red circle) when the abnormal area separated into two. The proposed

method also cannot extract the abnormal area when the original image contained bigger

damage in the retinal layer (Fig. 12(a) to Fig 12(b)).

(a) Original image (b) Binary image (c) Border tracking image

Fig. 12.Example of failure extraction for abnormal area in black colour.

4. Conclusion and Future Works

This paper presented a new border tracking procedure using regional statistics (BTPRS)

to extract the abnormal area from the human retina optical coherence tomography

image. We have done four different methods to extract the abnormal area. The first is

image scanning method. Then we tried using image border tracking method. Both of

these method have a same problem that is, the system cannot extract the abnormal area

that contained the damage retinal layer in the OCT image. To resolve that problem, we

proposed the extraction method using regional statistics scanning image; previous

procedure. To improve the result, we proposed the BTPRS by combined the regional

statistics and border tracking methods.


From all experimental results, the proposed BTPRS provides the betteralternative to

extract the abnormal area from human retinal optical coherence tomography images.

This method also can extract the abnormal area from the OCT image in both colours,

white or black. It also can detect the abnormal part that only specified by a medical

doctor.

For the future enhancement experiment to get more precise result, we are thinking to

combine the mean and standard deviation values of every region to compare with the

benchmark. We hope this research will help the task of clinical doctors, and the system

would specify the degree of retina disease accurately.

Acknowlwdgements. We are grateful to Universiti Sultan Zainal Abidin (UniSZA),

Malaysia for supporting this research and to Suzuka University of Medical Science,

Japan for providing the expertise.

References

[1] C. K. Hitzenberger, P. Trost, P. W. Lo, Q. Zhou, Three-dimensional Imaging of the

Human Retina by High-speed Optical Coherence Tomography, Optics Express, 11

(2003), no. 21, 2753-2761. http://dx.doi.org/10.1364/oe.11.002753

[2] D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang, M.

R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, J. G. Fujimoto, Optical Coherence

Tomoghraphy, Science, 254 (1991), 1178-1181.

[3] J. G. Fujimoto, C. Pitris, S. A. Boppart, M. E. Brezinski, Optical Coherence

Tomography: An Emerging Technology for Biomedical Imaging and Optical Biopsy,

Neoplasia, 2 (2000), no. 1-2, 9-25. http://dx.doi.org/10.1038/sj.neo.7900071

[4] H. K. Tony, J. G. Fujimoto, J. S. Schuman, L. A. Paunescu, A. M. Kowalevicz, I.

Hartl, W. Drexler, G. Wallstein, H. Ishikawa, J. S. Duker, Comparision of Ultrahigh

and Standard Resolution Optical Coherence Tomography for Imaging Macular

Pathology, Ophthalmology, 112 (2005), no. 11, 1922-1935.

http://dx.doi.org/10.1016/j.ophtha.2005.05.027

[5] W. Drexler, U. Morgner, R. K. Ghanta, F. X. Kartner, J. S. Schuman, J. G.

Fujimoto, Ultrahigh-resolution Ophthalmic Optical Coherence Tomography, Nature

Medicine, 7 (2005), no. 4, 502-507.

[6] T. Yagi, F. Okuyama, H. Kawanaka, S. Tsuruoka, A Study on Extraction Method of

Internal Limiting Membrane and Retinal Pigment Epithelium from OCT Images,

Proceeding of Joint 4th Interna tional Conference on Soft Computing and Intelligent

Systems and 9th International Symposium on Advanced Intelligent Systems (SCIS &

ISIS 2008), 2008-2013.

http://dx.doi.org/10.1364/oe.11.002753

http://dx.doi.org/10.1038/sj.neo.7900071

http://dx.doi.org/10.1016/j.ophtha.2005.05.027


[7] A. Yamakawa, D. Kodama, S. Tsuruoka, H. Kawanaka, H. Takase, M. F. A. Kadir,

H. Matsubara, F. Okuyama, Extraction Method of Retinal Border Lines in Optical

Coherence Tomography Image by Using Dynamic Contour Model, World Congress

Biomechanics 2010, International Federation for Medical and Biological Engineering

IFMBE Proceedings, 31 (2010), 1558-1561.

http://dx.doi.org/10.1007/978-3-642-14515-5_397

[8] D. Kodama, A. Yamakawa, S. Tsuruoka, H. Kawanaka, H. Takase, M. F. A. Kadir,

H. Matsubara, F. Okuyama, A Retinal Layer Structure Analysis to Measure the Size of

Diesease Using Layer Boundaries Detection for Optical Coherence Tomography

Images, World Congress Biomechanics 2010, International Federation for Medical and

Biological Engineering IFMBE Proceedings, 31 (2010), 1554-1557.

http://dx.doi.org/10.1007/978-3-642-14515-5_396

[9] M. F. A. Kadir, A. Yamakawa, D. Kodama, S. Tsuruoka, H. Kawanaka, H. Takase,

Y. Uji, H. Matsubara, F. Okuyama, Extraction Method of Abnormal Parts in Retinal

Layers from Optical Coherence Tomography Image, Proceedings of the 2nd

International Workshop on Regional Innovation Studies (IWRIS2010), 83-86.

[10] M. F. A. Kadir, S. Tsuruoka, H. Takase, H. Kawanaka, F. Okuyama, Y. Uji, H.

Matsubara, H. Yagami, Extraction of Macular Disease Area Using Ragional Statistics

for Optical Coherence Tomography (OCT) Image, Proceedings of the 3rd International

Workshop on Regional Innovation Studies (IWRIS2011), (2011), 57-60.

Received: April 15, 2015; Published: November 2, 2015

http://dx.doi.org/10.1007/978-3-642-14515-5_397

http://dx.doi.org/10.1007/978-3-642-14515-5_396

extraction of macular disease area using regional ... technique for human retinal optical ......

Documents