A Literature Review of Extraction of Optic Disc in Ophthalmology Diagnosis

A Literature Review of Extraction of Optic Disc in Ophthalmology Diagnosis

R.SubbuLakshmi Prof.P.Mariselvam

PG Scholar Asst.Professor

Sardar Raja College of Engineering, Alankulam Sardar Raja College of Engineering, Alankulam

Abstract

Optic Disc diagnosis helps to identifying the possible disorder in eye. Diseases of the optic nerve usually result in reduced visual acuity, color vision, a relative afferent pupillary defect, visual field defects and cupping. The optic disc is the entry point for the major blood vessels that supply the retina. Optic disc Analysis is one of the critical issues because these images contain very small nerves and some artifacts in it. After Extraction, the abnormal object is diagnosed by analyzing the image shapes and selecting specific image objects and Detecting Abnormal Roots. Our main goal has been to compile an introduction to the subject of Extraction of retinal Images. There exist a number of studies of other algorithms are used for best describe the various parts. In this paper, we explain about different approaches used to detect optic disc effectively.

Keywords Extraction, Diagnosis, Morphology, Optic Disc

1. Introduction

   Optic disc is the location where ganglion cell axons exit the eye to form the optic nerve. The goal of Extraction of optic disc is to simplify the representation of an image into something that is more meaningful and easier to analyze. Retinal Image Extraction is typically used to locate objects and boundaries in images. Every pixel in a region is same with respect to some characteristic or computed property, such as colour, intensity. Image Extraction is usually the first task of any image analysis process. All subsequent tasks, such as feature extraction and object recognition rely heavily on the quality of the Extraction. Without a good Extraction algorithm an object may never be recognizable. The purpose of image Extraction is to partition an image into meaningful regions with respect to a particular application. Extractions of easy gray-level pictures can provide helpful data regarding the surfaces in the scene. Usually image Extraction is an initial and vital step in a series of processes aimed at overall image understanding. Extractions of simple gray-level images can

   provide useful information about the surfaces in the scene. Extraction problems are the bottleneck to achieve object extraction, object specific measurements, and fast object rendering from multi-dimensional image data. Simple Extraction techniques are based on local pixel-neighbourhood classification.

   .

2. Literature Review

   1. Cup Segmentation

      This paper describes about Automatic retinal image analysis [1] is rising as an important screening tool for early detection of eye diseases. Eye disease is one in all the most common causes of visual impairment. The manual examination of point is a standing operating procedure used for detect eye disease. This paper present Associate in treatment automatic OD parameterization technique supported segmented OD and cup regions obtained from monocular retinal image. a novel OD segmentation technique is projected which integrates the native image data around every purpose of interest in four-dimensional feature area to provide hardiness against variations found in and round the OD region. Conjointly propose a novel cup segmentation technique which relies on anatomical proof like vessel bends at the cup boundary, considered relevant by eye disease consultants. Bendsin a vessel area unit robustly detected employing a region of support construct, which mechanically selects the right scale for analysis. A multi-stage strategy is used to derive a reliable set of vessel bends known as r-bends followed by a neighborhood spline fitting to derive the required cup boundary. The strategy has been evaluated on 138 images comprising traditional and one zero five eye diseases images against three glaucoma consultants. The obtained segmentation results show consistency in handling numerous geometric variations found across the dataset. The estimation error of the strategy for vertical cup-to- disk diameter ratio is zero.09/0.08 (mean/standard deviation). Overall, the obtained qualitative and quantitative results show effectiveness in both segmentation and consequent OD parameterization for eye disease assessment.

   2. Detection of Exudates

      This paper explains the detail of exudates among the macular region [2] is a main hallmark of diabetic macular hydrops and permits its detection with a high sensitivity. Hence, detection of exudates is a very important diagnostic task, within which computer help may play a serious role. Exudates are found victimisation their high grey level variation and their contours are determined by means that of morphological reconstruction techniques. The detection of the blind spot is indispensable for this approach. Detect the blind spot by means that of morphological filtering techniques and also the watershed transformation. The rule has been tested on a tiny low image knowledge base and compared with the performance of somebody's critic. Robustness with relevance changes of the parameters of the rule has been evaluated. This rule consisting in shade correction, contrast sweetening, sharpening, and a manually chosen threshold are applied to this downside. In colour normalisation and local contrast sweetening are followed by fuzzy C-means clustering and neural networks [2] are used for the final classification step. This has been shown to work well, but it relies on the native contrast sweetening that has been introduced and that amplifies noise particularly in areas, within many features.

      Figure 1:-(a) Exudates Image

      (b) Segmented Image

   3. Localization of Optic Disc

      Most of the algorithms developed for OD detection are especially applicable to normal and healthy retinal images. Its a challenging task to discover OD in all types of retinal images that is normal healthy images likewise as abnormal that is images affected because of disease. This paper presents an automated system to locate Associate in Nursing OD and its centre in all types of retinal images. The ensemble of steps supported different criteria produces a lot of accurate results. The planned algorithmic rule [3] gives glorious results and avoids false OD detection. The technique is developed and tested on commonplace databases provided for researchers on net, Diaretdb0 (130 images), Diaretdb1 (89 images), Drive (40 images) and local information (194 images). The local information images are collected from ophthalmic

      clinics. The results achieved by different algorithms will be compared once algorithms are applied on same commonplace databases.

      An automated method has been conferred that is in a position to locate Associate in Nursing OD in retinal images. The results show that the system is in a position to locate the OD accurately in 98.45% of all tested cases. The share of successful detection of OD is enlarged mistreatment method conferred in this paper. The tactic of OD localization is tested on retinal images Associate in nursing qualitatively valuated by comparison the automatically divided OD with manual ones detected by an older eye doctor. Original detection of OD leads towards the development of a totally automated retinal image analysis system to help clinicians in sleuthing and designation retinal diseases. Compared to the approaches by other researchers, in this algorithmic rule for OD detection has the advantage that it's applicable to any or all types of retinal images, healthy likwise as abnormal affected because of disease.

      Table 1:- Database Used For OD Localization

      S No.	Test Database	No. of Images
      1	Diaretdb0	130
      2	Diaretdb1	89
      3	Drive	40

   4. Model Based Approach

      A modified active shape model is projected in the shape finding of optic disc. A body structure coordinate system is established to provide a much better description of the options in the retinal images; associate approach to detect exudate by the combined region growing and edge detection is projected. The success rates of disk localization, disk boundary detection, and fovea localization area unit ninety nine, 94%, and 100%, respectively. The sensitivity and specificity of exudates detection area unit are 100 pixels. The success of the projected algorithms may be attributed to the utilization of the model-based methods [4]. The detection and analysis could be applied to automatic mass screening and designation of the retinal diseases. Modification in the shape, color or depth of optic disk is associate indicator of various ophthalmic pathologies particularly for glaucoma. The correct detection of the optic disk boundary may be accustomed assess the progress of eye disease and the treatment results. Some elements of the disk boundary don't seem to be well defined and some elements area unit partially obscured by the blood vessels in

      retinal images, that create the detection of roundness difficult. A modified active shape model (ASM) is projected to find the disk boundary in retinal images.

      PCA based mostly model is projected to localize optic disk in the candidate regions. The boundary of optic disk is extracted by a modified ASM method. A body structure coordinate system is established to provide a much better description of the options in the retinal images. Associate approach to find exudates [2] by the combined region growing and edge detection supported the color difference is projected. Substantial experiments are performed, that show the relative simplicity and blessings of the projected algorithms. The success of the projected algorithms may be attributed to the utilization of the model-based methods. Tests are administrated on the projected algorithms once additional appropriate data area unit on the market clinically. Such tests could contribute to any enhancements on the algorithms, leading to additional robust and additional correct detection that eventually may be accepted for the clinical functions.

      Figure 2:- Comparison of modified ASM with ASM for Boundary detection

   5. Optic Nerve Head Segmentation

      This paper explains concerning efficient point localization and segmentation area unit vital tasks in machine-driven retinal screening. All purpose edge detection algorithms [5] usually fail to section the point because of fuzzy boundaries, inconsistent image distinction or missing edge options. This paper presents associate degree algorithmic rule for the localization and segmentation of the nervous optics head boundary in low-resolution images. Point localization is achieved exploitation specialised template matching and segmentation by a deformable contour model. The latter uses a world elliptical model and an area deformable model with variable edge-strength dependent stiffness. The algorithmic rule is evaluated against an arbitrarily elect info of a hundred images from a diabetic screening programme. 10 pictures were classified as unusable; the others were of variable quality. The localization algorithmic rule succeeded on all bar one usable image; the contour estimation algorithmic rule was qualitatively assessed by an eye doctor.

      Nervous optics head segmentation by active contours has not been extensively examined within the past. There are unit important issues in addressing distracters on blood vessels edges and also the skin colour, and with the terribly variable look of the nervous optics head. Antecedent printed techniques need careful formatting of the model position, pre-processing of the image exploitation morphological operations, and perform badly wherever the rim is faint or undetectable. In distinction, the algorithmic rule given during this paper exploits specific options of the nervous optics head anatomy to attain smart localization whereas avoiding distracters. The temporal lock algorithmic rule exploits the form of the neuro retinal rim to bypass blood vessels and avoid the skin colour, and also the world and native deformable model deals effectively with weak areas of rim and vessel crossings. Shaping energy functions and employing a quasi-Newton optimisation strategy makes the algorithmic rule fairly quick.

   6. Fuzzy Convergences

      In this method uses a novel algorithm, fuzzy convergence [6] to determine the origination of the blood vessel network. This technique used thirty one images of healthy retinas and fifty images of diseased retinas, containing such numerous symptoms as tortuous vessels and haemorrhage that fully obscure the particular nerve. This technique experimented and results, detection of 89 correct image. Additionally compare our technique against 3 less complicated ways, demonstrating the performance improvement. This algorithmic rule identifies the nervous optics because the centre of attention of the vessel network. Within the absence of a robust convergence, this technique identifies the nervous optics because the brightest region within the image when illumination effort. Fuzzy convergence

      [9] may be a voting-based technique. The balloting takes place on the number grid of the initial image. Every line-like form is sculptured by a fuzzy section, whose space contributes votes to its constituent pixels. The summation of votes at every pel produces an image map wherever every pel contains a worth proportionate to its strength of convergence. The map is then blurred and threshold to supply one points of strongest convergence. The projected technique runs in on time, wherever n is that the range of line-like shapes. It doesn't need any quantity of inliers; instead, associate degree threshold for strength is also applied to see if any space ought to be deemed focused. During this ways use the convergence of the vessel network because the primary features for detection, in conjunction with the brightness of the nerve as a secondary feature.

   7. Ridge-Based Vessel Segmentation

      This technique is given for machine- driven segmentation of vessels in two-dimensional colour images of the tissue layer. This technique is utilized in pc analyses of retinal images, e.g., in machine-driven screening for diabetic retinopathy. The system is predicated on extraction of image ridges [7] that coincide around with vessel centre lines. The ridges area unit used to compose primitives within the variety of line parts. With the road parts an image is partitioned off into patches by assignment every image to the highest line element. Each line part constitutes an area coordinate frame for its corresponding patch. For each pel, feature vectors area unit computed that build use of properties of the patches and also the line parts. The feature vectors area unit classified employing a NN-classifier and serial forward feature choice. The algorithmic rule was tested on a info consisting of forty manually labeled images. The tactic achieves a neighbourhood beneath the receiver operative graphical record of 0.952.In this paper, analysis has been done exploitation accuracy of exhausting classifications and values of soppy classifications. Alternative analysis measures may be a lot of acceptable, looking on the appliance at hand. As an example, if one is fascinated by examining the tortuousness of the vessels, the breadth of the vessels may not be vital, solely the centrelines. The measures used don't take into consideration the quantity of branches, the connectedness of the vessels or the quantity of branching points, that all may be relevant in specific applications. Another smart analysis measure may be the performance of a computer- aided diagnosing system for retinal pictures that uses the results of a vessel segmentation algorithmic rule in its analysis.

   8. Geometrical Model of Vessel Structure

      This paper presents the plan of the preliminary detection of the most retinal vessels. All retinal vessels [8] originate from the OD and their path follows an analogous directional pattern all told images. To explain the overall direction of retinal vessels at any given position within the image, a geometrical constant model was projected, wherever 2 of the model parameters area unit the coordinates of the OD centre. Exploitation as experimental information samples of vessel centre line points and corresponding vessel directions, provided by any vessel identification procedure, model parameters were known by means that of a simulated hardening optimisation technique. These calculable values offer the coordinates of the middle of OD. The performances of the projected technique supported a model of the structure, area

      unit passionate about the supply of a decent portion of this structure within the image, whereas area unit freelance of the particular visibility of the OD. Being the structure unfolds everywhere the image, it's a lot of less laid low with the presence of contradictory or obscuring pathological areas. The supply of a vessel extraction procedure may be a necessary necessity for this system, and also the performances of this step directly have an effect on the right positioning of the OD.

   9. Active Visual Segmentation

      Attention is associate degree integral a part of the human sensory system and has been widely studied within the visual attention [9] literature. The human eyes fixate at vital locations within the scene, and each fixation purpose lies within a selected region of absolute form and size, which might either be a complete object or a locality of it. exploitation that fixation purpose as associate degree identification marker on the article, this technique was projected to section the article of interest by finding the optimal closed contour round the fixation purpose within the polar space, avoiding the perennial downside of scale within the mathematician house. The projected segmentation method is applied in 2 separate steps: initial, all visual cues area unit combined to get the probabilistic boundary edge map of the scene; second, during this edge map, the optimal closed contour around a given fixation purpose is found. Having 2 separate steps additionally makes it attainable to determine an easy feedback between the mid-level cue and also the low-level visual cues. Finally, this experiments show the promise of the projected technique as associate degree automatic segmentation framework for a general purpose sensory system. The framework combines static cues with motion and stereo to clear up between the interior and also the boundary edges. The approach is intended by biological vision, and it's going to have connections to neural models developed for the matter of border possession in segmentation. Though the framework was developed for an active observer, it applies to image databases further, wherever the notion of fixation amounts to choosing an image purpose that becomes the middle of the polar transformation.

   10. Gray-Level and Moment Invariants- Based options

       This paper presents a brand new supervised technique for vessel detection in digital retinal images. This technique uses a neural network theme for pel classification and computes a 7-D vector composed of gray-level and moment invariants-based options [10] for pel illustration. the tactic was evaluated on the publically out there

       DRIVE and STARE databases, wide used for this purpose, since they contain retinal images wherever the structure has been exactly marked by specialists. This technique performance on each sets of take a look at images is best than alternative existing solutions in literature. The tactic proves particularly correct for vessel detection in STARE images. Its application to the current outperforms all analyzed segmentation approaches. Its effectiveness and strength with totally different image conditions, beside its simplicity and quick implementation, build appropriate for retinal image pc analyses like machine-driven screening for early diabetic retinopathy detection.

3. Conclusion

   This paper is useful to analysis various strategies methods used to segment the optic disc efficiently. This paper briefly explains the diagnosis of infectious region simpler, more accurate by using special and attentive methods for interpretation of the medical images. In this paper the available classifications of methods was reviewed as well as a categorized for applying this techniques to decreasing human intervention in optic disc extraction. There are many Extraction algorithms that can be universally used to solve many problems. So the goal is to search for best algorithms that can be used to segment the retinal images and compared their performance with other Extraction approaches for better analysis.

4. References

1. Sivaswamy, and S. R. Krishnadas Optic Disk and Cup Segmentation From Monocular Color Retinal Images for Glaucoma Assessment in IEEE Transactions on Medical Imaging, vol.30, no. 6, june 2011

2. T. Walter, J. C. Klein, P. Massin, and A. Erginay, A contribution of image processing to the diagnosis of diabetic retinopathy-detection of exudates in color fundus images of the human retina, IEEE Trans.Med. Imag., vol. 21, no. 10, pp. 12361243, Oct. 2002.

3. M. Park, J. S. Jin, and S. Luo, Locating the optic disc in retinal images, in Proc. IEEE Int. Conf. Comput. Graph. Imag. Visualisat. , 2006, pp. 141 145.

4. H. Li and O. Chutatape, Automated feature extraction in color retinal images by a model based approach, IEEE Trans. Biomed. Eng., vol. 51, no. 2, pp. 246254, Feb. 2004.

5. J. Lowell, A. Hunter, D. Steel, A. Basu, R. Ryder,

   E. Fletcher, and L.Kennedy, Optic nerve head segmentation, IEEE Trans. Med. Imag., vol. 23, no. 2, pp. 256264, Feb. 2004.

6. A. Hoover and M. Goldbaum, Locating the optic nerve in a retinal image using the fuzzy convergence of the blood vessels, IEEE Trans. Med. Imag., vol. 22, no. 8, pp. 951958, Aug. 2003.

7. 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

8. 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,

   Vol. 23, No. 10, October 2004

9. Ajay K. Mishra, Yiannis Aloimonos, Loong-Fah Cheong, And Ashraf A. Kassim, Active Visual Segmentation,Ieee Transactions On Pattern Analysis And Machine Intelligence, Vol. 34, No. 4,

   April 2012

10. Diego Marín, Arturo Aquino, M.E.G.Arias, and J.M Bravo , A New Supervised Method for Blood Vessel Segmentation in Retinal Images by Using Gray-Level and Moment Invariants-Based Features, IEEE Transactions On Medical Imaging , Vol. 23,N0.11, April 2011.

Acknowledgement

I would like to express my gratitude towards my HOD for her kind co-operation and encouragement which helps me in completion of this paper effectively. I am highly indebted to my guide for their valuable guidance and constant supervision as well as providing necessary information regarding the review of this paper. I would like to extend my sincere thanks to all my Staff Members of Department of CSE of Sardar Raja College of Engineering,Alankulam.

R.SubbuLakshmi received her B.E degre in Computer Science and Engineering , National college of Engineering, Maruthakulam in 2010 from Anna University, Chennai and she is currently doing

M.E degree in Computer Science and Engineering, Sardar Raja

College of Engineering, Alankulam. She published paper in National Conference. Her research interest includes Digital Image Processing.

P.MariSelvam received his B.E degree in Computer Science and Engineering, Sardar Raja College of Engineering, Alankulam in 2006 from Anna University, Chennai and

M.E Computer Science and Engineering in Dr. Sivanthi Aditanar College of Engineering, Tiruchendur, Anna University of

Technology, Tirunelveli. He is currently working as an Assistant Professor in Sardar Raja College of Engineering; Alankulam.He published papers in International Conference. His research interest includes image analysis and segmentation.