Automatic Detection of Optic Disc and Blood Vessel in Retinal Images using Morphological Operations and Ipachi Model

 

 Ganesan P^1*, B. S. Sathish²

¹Department of Electronics & Communication Engineering, Vidya Jyothi Institute of Technology, Hyderabad

²School of Electrical &Electronics Engineering, Sathyabama University, Chennai

 

ABSTRACT:

Automated detection of Optic disc and blood vessel fabrication is becoming of virtual interest for better management of diabetic disease. In this paper, we propose morphology technique and new infinite active contour model that uses hybrid region information of image to approach this problem. For optic disc detection scheme utilizes operations such as edge detection, binary threshold and morphological operation. For vessel detection we use an infinite perimeter regularize, provided by using L2 Lebesgue measure of the  boundaries, allows for better detection of small oscillatory  structures than the  models based on length of a feature’s boundaries. The advantage of using different types of region information, Eigen based enhancement, such as the combination of image intensity information and local phase based enhancement. The local phase based enhancement map is used for preserving vessel edges while the given intensity information will give a correct feature’s segmentation.

 

 

 

 

INTRODUCTION

Blood vessels  and optic disc can be conceptualized anatomically as intricate network, or treelike structure , of hollow tubes of different sizes , compositions are  include arteries, arterioles, capillaries, venules, and veins. The identification of retinal features such as the optic disc and the retinal vessels is a prerequisite before systems can achieve more complex tasks identifying pathological entities¹. The location of the optic disc is crucial in retinal image; any damage to them could lead to profound diabetes, and hypertension, to name only the most obvious.  The drive for better understanding of these conditions naturally the need in improved imaging techniques. The detection and analysis of the OD and vessels in medical images is a fundamental task in many clinical applications for early detection, diagnosis and optical treatment².

 

With the proliferation of imaging modalities, there is increasing demand for automated detection of vessel analysis systems for blood vessel segmentation is the first and most important step. The blood vessels can be seen as linear structures distributed at different orientation and scales in an image, various enhancement filters have been used to enhance them in the segmentation problem³. In particular, a local phase based filter recently introduced by Lather at all seems to be superior to intensity based filters as it is immune to intensity inhomogenity and is capable of faithfully enhancing vessels of different widths. Morphological filters such as path opening in multi scale Gaussian filters¹⁴. The main disadvantage of morphological methods is they do not consider as known vessel cross-sectional shape information, and the use of a structuring element may cause difficulty in detection of highly tortuous vessels¹⁶.

 

PROPOSED SYSTEM:

This work proposed morphology technique for optic disc and new infinite active contour model that uses hybrid region information of the image. More specifically, an infinite perimeter regularize, provided by using L2 Lebesgue measure of the neighbor of boundaries, allows for better detection of small oscillatory  structures than the models based on the length of a feature’s boundaries⁴. The mathematical morphology operators include the following process: Dilation, Erosion, Opening and Closing.                                                  

 

Dilation is used as adds pixels to the boundaries of objects in an image. Erosion removes the pixels of object boundaries⁵. The morphological open operation is erosion followed by dilation, using the same structuring element for both operations

 

A◦B = (AΘB) ⊕B

The closing operator is a dilation followed by erosion.

A●B = (A ⊕ B) ΘB

 

The energy of the IPACHI model is:  

 

 

Where L2 is the 2D Lebesgue measure, Rn is the nth region information, and N is the total number of different region terms¹⁷. The first term L2 is the area of the -neighborhood of the edge set. Here we consider L2, for a large and even number, which is an approximation of the neighborhood area in a given image.

 

The optic disc (OD) is usually the brightest portion of retinal image. From a given RGB retinal image, unlike conventional approaches implemented for detecting the OD region from a given retinal image⁶. First, optic disc is localized using global threshold and feature extraction. Next, form the location obtained in first step, edge detection and morphological operation are performed to obtain the boundary region of the OD¹⁸.

 

A. Pre-processing by Global Thresholding:

At first, for the blue plane image a global threshold is determined by using most commonly used. This threshold is scaled to a higher value and used to transform the image into a binary one ⁷. For a disease free image with perfect illumination condition, most of the bright pixels stay inside OD. Hence a crude estimate of bright OD region is obtained with so many disjoint bright pixels. In order to obtain smooth well connected OD region, some morphological operations are proposed. In this regard, first the erosion morphological operation is performed considering a square object Sμ, where μ×μ(μ>1). Then dilation operation followed by again erosion operation is performed with a disc object Dλ, where λ(λ>1). This will connect bright pixels and some blobs will be formed. The choice of λ and μ may be dependent on the illumination condition, resolution, and size of image⁸.

 

Fig.1 Proposed Model for the detection of optical disc and blood vessel

 

B. Feature Extraction and OD Detection:

The features need to be chosen in such a way that they are sufficiently distinctive; less sensitive to varying environment, With a view to localize the OD, following features are extracted¹⁹.

 

Area (A): This is the total number of pixels of the Blob.

Ratio of major axis and minor axis (R): The length of the major axis and minor axis are obtained from the ellipse that fits the Blob best.

 

Compactness(C): This feature determines the closeness to circular shape of the segmented region of retinal image, which is evaluated as   Compactness =Area/ (Perimeter) ²

 

C. Post-processing:

To better segment OD region, a sub-image is taken around the previously segmented OD region from the blue portion of image, whose area is extended 80X80 outside the detected region⁹. Now applying median filtering in the sub-image small noise are removed. Then each 10X10 region of the sub-image was analyzed. From the summation of pixel values of all the regions, mean area is calculated¹⁰. The regions having area more than mean are transformed completely white and the other are transformed black¹¹. Now, after filling the holes and removing small areas, we will obtain a better segmented OD region. From the estimated bounding box, a circle’s centre and radius is calculated.

TYPICAL VESSEL ENHANCEMENT FILTER:

Filters which can enhance vessel-like structures have played an important role in the vessel segmentation problems¹². Here, we review the three most influential filters.

1)       Eigen value-based Filter:

Proposed by Frangi et al., this filter is based on eigen values of the Hessian matrix H(x). For each pixel of a 2D image with intensity f(x), the Hessian matrix can be formed by its 3 second derivatives, fx1x1, fx2x2, and fx1x2, from which eigen values can be computed and ordered¹³.

2)       Isotropic Undecimated Wavelet Filter:

The isotropic undecimated wavelet transform (IUWT) has recently been used for vessel segmentation and it show good accuracy and computational efficiency.

3)       Chan-Vese (CV) Model:

The CV model was initially proposed by Chan - Vese to solve the constant segmentation problem. It has been widely used and extended to address a wide range of segmentation¹⁴. Without loss of the generality, here we choose the 2-Dimensional (2D) segmentation problem as an example.

4)       Infinite Perimeter Active Contour (IPAC) Model:

The IPAC model is proposed for vessel segmentation of objects with irregular boundaries. The energy function is given as:

 

5)       IPACHI:

Inspired by the IPAC model, we propose a novel extension so as to integrate hybrid region information into the segmentation model. The energy of the IPACHI model is:

 

 

Figure 2 shows the input image of the retinal image. We give the input image for the detection of optic disc and blood vessels.

 

Fig.2 Input image

 

Fig. 3 Optical disc detected image

 

Figure 3 shows the output of the optic disc .We can easily get this image with the help of morphology process. Output of the Eigen value based enhanced result. With the help of enhancement process we can easily detect the blood vessels is illustrated in fig 4.

     

Fig. 4 Eigen value based enhanced result

 

The figure 5 depicts the output of wavelet based result. This image gives the segmentation of blood vessels.

 

Fig. 5 Wavelet based result

 

Figure 6 illustrates the output of the local phase based result. With the help of this we can easily detect the boundary of the image.

 

Fig. 6 Local phase based result

 

Outcome of the infinite perimeter active contour with hybrid region information. It is used in the segmentation of blood vessels is illustrated in fig 7.

 

Fig. 7 Ipachi result

 

Fig. 8 Input images

 

Figure 8 shows the input image of the retinal image. We give the input image for the detection of optic disc and blood vessels.

 

Fig.9 Optical disc detected image

Figure 9 shows the output of the optic disc .We can easily get this image with the help of morphology process. Figure 10 depicts the output of the eigen value based enhanced result. With the help of enhancement process we can easily detect the blood vessels.

 

Fig.10 Eigen value based enhanced result

 

Fig. 11 Wavelet based result

 

The figure 11 illustrates the output of wavelet based result. This image gives the segmentation of blood vessels.  The output of the local phase based result. With the help of this we can easily detect the boundary of the image is demonstrated in fig 12.

 

Fig.12 Local phase based result

 

Fig.13 Ipachi result

 

The yield of the infinite perimeter active contour with hybrid region information is illustrated in fig 13. It is used in the segmentation of blood vessels.

 

CONCLUSION:

In this paper, we use morphology for optic disc from retinal image. This method reduces computational load to a great extent. In vessel segmentation, it is shown that the Gaussian curve is suitable for modeling the intensity of the cross section of retinal vessels in the color fundus images. Based on this elaboration, the amplitude modified second-order Gaussian filter for retinal blood vessel detection is proposed and its performance is demonstrated. The mathematical analysis, blood vessel detection simulation and its assessment on various fundus images show that the Gaussian filter with modified amplitude can be efficiently applied for the detection and measurement of the retinal vessels. This measurement not only presents the information of blood vessel size but it is also valuable for optimizing the matched filter to improve the successful rate of detection.

 

ETHICS AND CONSENT:

This article does not contain any studies with human participants or animals performed by any of the authors. No direct participation of human entertains in this article.

 

CONFLICT OF INTEREST:

We are declaring that, there is no conflict of interest regarding the publication of this paper.

 

REFERENCES:

1.        Luo, C. Opas, and Shankar M. Detection and measurement of retinal vessels in fundus images using amplitude modified second-order Gaussian filter. IEEE Trans. Biomed. Eng. 49(1); 2008; 168–172.

2.        Zhang B, Zhang L, Zhang L, and Karray F. Retinal vessel extraction by matched filter with first-order derivative of Gaussian. Comput. Biol. Med.  40(1); 2010; 438–445.

3.        Li H, Chutatape, O. Automated feature extraction in color retinal images by a model based approach. IEEE Trans. Biomed. Eng. 51; 2004; 246–254.

4.        Rossant F, Badellino M, Chavillon A, Bloch I, and Paques M. A morphological approach for vessel segmentation in eye fundus images, with quantitative evaluation. J. Med. Imaging. Health. Inf.  1(2); 2011; 42–49.

5.        Soares J and Cree M. Retinal vessel segmentation using the 2D Gabor wavelet and supervised classification.  IEEE Trans. Med. Imag.  25; 2006; 1214–1222.

6.        Kalist V, Ganesan P, Sathish BS, and Jenitha JMM. Possiblistic-Fuzzy C-Means Clustering Approach for the Segmentation of Satellite Images in HSL Color Space.  Procedia Computer Science. 57; 2015; 49-56.

7.        Shaik KB, Ganesan P, Kalist V, and Sathish BS. Comparative Study of Skin Color Detection and Segmentation in HSV and YCbCr Color Space.  Procedia Computer Science. 57; 2015; 41-48.

8.        Ganesan P and Rajini V.  A method to segment color images based on modified fuzzy possibilistic-c-means clustering algorithm. Recent Advances in Space Technology Services and Climate Change (RSTSCC), IEEE. 2010; 157-163.

9.        Ganesan P and Rajini V. Assessment of satellite image segmentation in RGB and HSV color space using image quality measures. Advances in Electrical Engineering (ICAEE), 2014 International Conference on. 2014; 1 -5.

10.     Premalatha J, Anitha U, Manonmani V, Ganesan P.  Survey on energy saving methods for green communication network. Indian Journal of Science and Technology.  8(19); 2015; 1-5.

11.     Ganesan P and Shaik KB. HSV color space based segmentation of region of interest in satellite images.  2014 International Conference on Control, Instrumentation, Communication and Computational Technologies (ICCICCT). 2014; 101-105.doi: 10.1109/ICCICCT.2014.6992938

12.     Krishna RV, Sathish BS and Abilash R. Design of voice and gesture controlled Quadcopter. 2015 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS).  2015; 1-6.

13.     Sajiv G and Ganesan P.  Comparative Study of Possiblistic Fuzzy C-Means Clustering based Image Segmentation in RGB and CIELuv Color Space.  International Journal of Pharmacy and Technology. 8(1); 2016; 10899-10909.

14.     Malay Bhushan, Niraj Bishwash, and kalist V. Wireless Power Transfer Platform for Smart Home Appliances. International Journal of Pharmacy and Technology. 8(3); 2016; .15669-15674.

15.     Sajiv G. Unsupervised Clustering of Satellite Images in CIELab Color Space using Spatial Information Incorporated FCM Clustering Method.  International Journal of Applied Engineering Research. 10(20); 2015.

16.     Sathish BS, Ganesan P and Khamar Basha. Shaik. Color Image Segmentation based on Genetic Algorithm and Histogram Threshold. International Journal of Applied Engineering Research. 10(6); 2015; 123-127.

17.     Rajneesh Kumar and Rajeev Kumar. Robust Control of Instruments using 2D Virtual Reality Sensing System. International Journal of Pharmacy and Technology. 8(3); 2016; 15698-15703.

18.     Shekhar RS, Shekhar P and Ganesan P. Automatic detection of squats in railway track. 2015 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS). 2015; 1-5.
Thakur M, Raj I and Ganesan P. The cooperative approach of genetic algorithm and neural network for the identification of vehicle License Plate number. International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS). 2015; 1-6.

 

 

 

 

Accepted on 28.06.2017           © RJPT All right reserved