Automatic Fovea Detection and Grading of Diabetic Maculopathy Severity Levels

Automatic Fovea Detection and Grading of Diabetic Maculopathy Severity Levels

Shobha Christila.S1,

1Assistant Professor, Biomedical Engineering,

Sri Ramakrishna Engineering College,Coimbatore, Tamil

Nadu, India.

Rajeshwari R²

2Assistant Professor, Biomedical Engineering,

Sri Ramakrishna Engineering College Coimbatore, Tamil

Nadu, India.

AbstractIn this work, localization of the fovea automatically is done by getting information of the Optic Disk and the blood vessel structure and hence to localize the fovea region further accurately by mathematical morphology method. The algorithm which proposed has two stages such as the first stage deals with the detection of the blood vessels of the retinal fundus image and then the proceeding stage utilizes the geometrical distance between Optic Disk and fovea region and the structure of the blood vessels to perfectly localize the fovea region. The extracted macula and fovea region may help in further automatic detection and grading the severity level of diabetic maculopathy.

Index Terms blood vessel, optic disk, fovea, diabetic maculopathy, mathematical morphology.

I. INTRODUCTION

Image processing, analysis and computer vision techniques are widely used today in all fields of medical sciences and especially to modern ophthalmology, as it is heavily dependent as visually oriented signs[1]. Hence retina is the only location where blood vessels can be directly visualized non-invasively in vivo.

.

Fig.1.The human retina

In colour fundus images optic disk and fovea are two important anatomical landmarks that is used in retinal disease screening .This paper presents a new, fast, fully automatic optic disk and fovea localization algorithm has been developed for diabetic maculopathy.

The optic disc or optic nerve head is the location where ganglion cell axon exits the eye to form the optic nerve. The fovea is the part of the eye located in the centre of the macula region of the retina. The fovea is responsible for sharp central vision [2] .The foveal region is made up of cones if the delicate cones in the fovea are destroyed the person become blind .Manual detection of fovea region by ophthalmologist is time consuming hence automation is highly needed. Hence this paper mainly deals with fovea localization in which the fundus retinal image the macula is the most darker part approximated by a circle and also geometrically fovea is said to be located at a distance 2.5 times the diameter of the optic disk from its center. Thus fovea localization thus have done using morphological operation and geometric features. After locating the fovea and macular region depending upon the diameter variations the disease such as diabetic maculopathy severity level is estimated. The diabetic maculopathy is a pathological disorder of the macula which frequently develops in diabetic patients. It is charactized by odemea, hard exudates, micro aneurysms and ischemia in the macular area. If the odemea is severe visual acuity will be reduced but a blue-yellow colour vision defect is usually noted before the loss of acuity.

1. DATA COLLECTION

   The retinal images are collected from DRIVE database. The drive database images are acquired using a canon CR5 non-mydriatic 3CCD camera with a 45 degree field of view [3]. Each image was captured using 8 bits per color plane at 768 by 584 pixels. The FOV of each image is circular with a diameter of approximately 540 pixels. For this database, the images have been cropped around the FOV.

2. PROPOSED METHOD

   The method has four parts:

   Blood vessel extraction Optic disk detection Fovea detection

   Grading the diseased part

   1. ALGORITHM FOR BLOOD VESSEL EXTRACTION

      The blood vessel is a network like structure. The optic disk originates from the blood vessel and hence the blood vessel extraction is mainly used for optic disk detection. Fovea is an region where there is no blood vessel and hence this criteria is also helpful in finding the foveal region[4].

      There are several methods adopted for blood vessel extraction such as blood vessel extraction by two dimensional matched filters[5], the ridge based blood vessel extraction[6] .The method adopted in this blood vessel extraction is morphology based scheme which is an simple enough and most accurate[7][8].

      Step 1 The image is taken and read in MATLAB.

      Step 2 The colour image is converted to gray scale image.

      Step 3 The image is resized for further processing.

      Step 4 The morphological opening operator is applied on image with disk shaped structuring element.

      Step 5 The morphological closing operator is used to remove vessel structure.

      Step 6 The top hat transformation is used on the image to extract the blood vessel structure.

      Step 7 The image is then converted to binary image by thresholding.

      Step 8 By using connected component analysis extract the blood vessel from the image.

   2. ALGORITHM FOR OPTIC DISK DETECTION

      There are various methods that has been proposed for optic disk detection such as geometric model of vessel structure [9], optic disk detection by adaptive morphological approach [10]. In this work the optic disk detection is based on median filter method since optic disk is having highest intensity pixel the suitable median filter is used to extract optic disk [11].

      Step 1 A mask is generated for the particular region of interest (i.e) optic disc over the image.

      Step 2 The mask is the applied over the image.

      Step 3 The morphological operators closing and opening were applied to give the final ROI.

      Step 4 After the detection of optic disk the diameter of the optic disk is calculated.

   3. ALGORITHM FOR FOVEA DETECTION

      The fovea detection is done depending on the information of the optic disk diameter estimation [4]. It has been found that fovea is located at 2.5 times the diameter of the optic disk .Hence using this calculation fovea detection is done [12].

      Step 1 Locate a point P horizontally at a distance 2.5×d from G towards the centroid.

      Step 2 Consider a vertical strip of width k pixels around P perpendicular to GP.

      Step 3 Apply a k × k sliding window along the strip and form the chain of numbers denoting the black pixels in the window.

      Step 4 Find the maximum run length of zeros, L in the number chain.

      Step 5 Let S and E are the start and end position corresponding to L and D be the mid position of S and E.

      Step6 Consider a binary image of size same as the input image with only a black pixel at position D. Dilate by a disc of radius DS to obtain .

      Step 7 Obtain R as the portion of the gray-scale image, 1 corresponding to the black region in .

      Step 8 Binarize R to approximate macula region.

      Step 9 Refine binarized R by removing noise and fitting the circle to obtain final macula region.

      Step 10 Detect fovea region as the small area around the center of macula.

   4. ALGORITHM FOR DIABETIC MACULOPATHY SEVERITY LEVELS ESTIMATION

      The grading of diabetic maculopathy is done in such a way that once the optic disk diameter and fovea region is estimated [13] .The macula which is surrounding the fovea is drawn three circles with radii 1/3 of optic disc diameter such as 1 DD (Disk Diameter ) ,2 DD,3 DD. Then the grading is done in such a way that [14]

      Fig. 2.Grading of diabetic maculopathy severity level

      LOCATION	CONDITION
      Exudates absnt	Normal
      Exudates present but outside 2 DD	Clinically Non Significant Diabetic Macular Edema
      Exudates present inside 2 DD	Clinically Significant Diabetic Macular Edema
      Outside 1DD	Mild
      Inside 1DD	Moderate
      1/3 DD	Severe

      Table 1: Classifications of various levels

3. RESULTS AND DISCUSSIONS

We have tested our proposed algorithm in a publically available DRIVE database. The method was implemented in MATLAB and testing of fovea was done in following images.

Fig.3. Input retinal image

Fig.4. Gray scale image

Fig.5. Opening the image

Fig.6. Closing the image

Fig.7.Top Hat transformed image

Fig.8. Binarised image

Fig.9. Blood vessel extracted image

Fig.8.Fovea detected image

IV CONCLUSION AND FUTURE ENHANCEMENT

In this paper fovea detection has been done by getting information of the optic disk and extracted blood vessel. The main algorithm proposed are based on image morphology. This algorithm has been tested on retinal images and this method has been used for automatic fovea detection which replaces any manual intervention. Thus in the future work fovea detections are then mainly used for grading of diabetic maculopathy severity levels.

V ACKNOWLEDGMENT

The authors would like to thank the Karunya University for giving immense support and also like to acknowledge the reviewers for this accomplishment.

VI REFERENCES

1. Digital Image Processing, Third Edition, By Gonzalez, Refael C.; Woods, Richard. E, Published By Pearson Education ,Inc

   ,Publishing As Prenctice Hall ,Copyright 2008.

2. Michael D. Abràmoff ,Mona K. Garvin, and Milan Sonka, Retinal Imaging And Image Analysis., IEEE Reviews In Biomedical Engineering, Vol. 3, 2010.

3. DRIVEdatabase,http://www.Isi.Uu.Nl/Research/Databases/DRIV E

4. Soumitra Samanta, Sanjoy Kumar Saha and Bhabatosh Chanda,A Simple And Fast Algorithm To Detect The Fovea Region In Fundus Retinal Image., IEEE Computer Society ,2011 Second International Conference On Emerging Applications Of Information Technology

5. Subhasis Chaudhuri, Shankar Chatterjee,Norman Katz, Mark Nelson, and Michael Goldbaum, Detection of Blood Vessels in Retinal Images Using Two-Dimensional Matched Filters, IEEE Transactions On Medical Imaging, Vol. 8, No. 3, September 1989.

6. Joes Staal, Michael D. Abràmoff, Meindert Niemeijer, Max A. Viergever, and Bram van Ginneken, Ridge-Based Vessel Segmentation in Color Images of the Retina, IEEE Transactions On Medical Imaging, Vol. 23, No. 4, April 2004.

7. Pierre Soille, 2003: Morphological Image Analysis, Principles And Applications (Practical Approach).

8. Jean Serra and Luc Vincet, An Overview Of Morphological Filtering (Mathematical Approach).

9. M. Foracchia , E. Grisan, and A. Ruggeri Detection of Optic Disc in Retinal Images by Means of a Geometrical Model of Vessel Structure, IEEE Transactions On Medical Imaging-2004.

10. Daniel Welfer A, Jacobscharcanski A, Cleysonm. Kitamura B, Melissam.Dalpizzol B, Laura W.B.Ludwig B, Dianeruschelmarinho C, Segmentation Of The Optic Disk In Color Eye Fundus Images Using An Adaptive Morphological Approach, Computers in Biology and Medicine 40 (2010) 124137.

11. Aliaa Abdel-Haleim Abdel- Razik Youssif, Atef Zaki Ghalwash, and Amr Ahmed Sabry Abdel- Rahman Ghoneim, Optic Disc Detection From Normalized Digital Fundus Images By Means Of A Vessels Direction Matched Filter, IEEE Transactions On Medical Imaging, Vol. 27, No. 1, January 2008.

12. S.Sekhar, W.Al-Nuaimy and A.K.Nandi, Automated Localization of Optic Disc and Fovea In Retinal Fundus Images, 16th European Signal Processing Conference (EUSIPCO 2008), Lausanne, Switzerland, August 25-29, 2008.

13. P.C. Siddalingaswamy, K. Gopalakrishna Prabhu, Automatic Grading Of Diabetic Maculopathy Severity Levels, IEEE Computer Society , Proceedings Of 2010 International Conference On Systems In Medicine And Biology 16-18 December 2010, IIT Kharagpur, India .

14. Luca Giancardo , Fabrice Meriaudeau, Thomas P. Karnowski, Edward Chaum, Yaqin Li, and Kenneth W. Tobin Jr,Exudation And Diabetic Macular Edema Detection In Retinal Fundus Images With A Publicly Available Dataset.