Imperial Journal of Interdisciplinary Research (IJIR) Vol-3, Issue-2, 2017 ISSN: 2454-1362, http://www.onlinejournal.in
A Survey on Glaucoma Detection Methods Ms. Swathy Ravi.V M. Tech in Computer Science & Engineering, Thejus Engineering College, Vellarakkad Abstract: Glaucoma is a disease that causes damage to our eye's optic nerve and subsequently leads to loss of vision. As the indications just happen, when the infection is very advanced, so the glaucoma is known as the silent thief of sight. Glaucoma can't be cured, yet its advancement can be backed off by treatment. In early stages it was hard to distinguish Glaucoma. The medicinal methods utilized by ophthalmologists like HRT and OCT is expensive and tedious. Henceforth there is a need to create automated system framework which can distinguish glaucoma and in less time. Optic disk and optic cup are prime elements which help in diagnosing glaucoma. Hence appropriate segmentation of optic cup and optic disk assume as an important factor in distinguishing the images as glaucomatic or healthy. The target of this survey is to acquaint a technique with automatically analyze the retinal images of the eye. Automatic detection of glaucoma is very efficient and accurate. keywords—Glaucoma, Optic cup segmentation, Optic disk segmentation, Cup to disk ratio, Nueroretinal rim.
1. Introduction The word glaucoma emanates from ancient Greek word, which means ‘clouded or blue-green hue’, ostensibly which describes someone with a dilated cornea or who's swiftly growing a cataract, each of which may be effected through continual (long-time period) growth inside the intraocular stress of the eye. Glaucoma is a class of sicknesses where in the optic nerve is vandalized main to irreversible loss of vision. In maximum cases, this harm is because of immense boom of strain inside the attention. The eye generates a vitreous fluid called aqueous humor that is secreted by means of ciliary body into the posterior chamber-a space among the iris and the lens and it drains via trabecular mesh community. In wholesome eye, the price of secretion is balanced to the fee of drainage. Glaucoma arises while drainage canal is in part or absolutely blocked which leads to the growth in strain, known as intraocular pressure which damages the optic nerve - used to transmit impulses to the mind where in visible records may be interpreted. If this damage left untreated, may additionally lead to overall blindness. Consequently, the early detection of glaucoma is vital. Presently, a vital indicator of glaucoma is CDR, described as the ratio of the vertical top of the optic cup to the vertical peak of
Imperial Journal of Interdisciplinary Research (IJIR)
the optic disc. For glaucoma detection, first, retinal photographs are acquired the usage of virtual capture gadgets for picture content. Then pre-processing is carried out for equalizing the irregularities at the pictures. In pre-processing, blood vessels are segmented and in painted to benefit a vesselunfastened picture. Then, feature extraction is done to lessen the scale efficaciously to symbolize the fascinated components of an picture as a concise function vector for describing the massive data set exactly. Pixel intensity values, textures, FFT coefficients and Histogram model are the techniques used in feature extraction. Photograph category is executed which analysis the numerical properties of image and organizes the information. Depending on the outcomes obtained, the set of facts is split into discrete classes’ i.e. regular eye or glaucomatous eye. Optic nerve cupping progresses as the cup becomes larger in comparison to the optic disc. A cup to disc ratio value that is greater than 0.65 is generally considered to be suspicious for glaucoma. Many studies have been suggested previously at the computerized segmentation of the optic disc and cup from retinal fundus photographs. Many research proposed a disc detection scheme the usage of variational level set segmentation and then, using threshold degree-set segmentation for cup detection [5][6][7]. This technique uses an elliptical fitting post-processing to deal with deformation because of blood vessels. The method offered in [8] makes use of manual threshold evaluation, coloration issue evaluation and ROI (region of interest) based segmentation for the detection of the cup. For the cup, the factor evaluation technique is used.
2. Automatic Glaucoma Detection Methods Glaucoma is a disease portrayed by degeneration of optic nerves. So the fall in circulatory system to the optic nerve accommodate the visual field surrenders related with glaucoma. Sedate treatment to control the lifted intraocular weight and serial appraisal of the optical nerves is the focal system for curing the ailment. Standard procedures for appraisal of the optic nerve using ophthalmology or stereo photography or evaluation of visual fields. There are manual and modified disclosure methods available. The review is driven on different glaucoma acknowledgment procedures in image processing.
Page 1063
Imperial Journal of Interdisciplinary Research (IJIR) Vol-3, Issue-2, 2017 ISSN: 2454-1362, http://www.onlinejournal.in This review rapidly delineates a few techniques that are used for the discovery of glaucoma. 2.1. Automated Glaucoma using CDR The diagnosis of glaucoma ought to be conceivable through the estimation of CDR (Cup to Disk ratio). At this moment, CDR evaluation is physically performed by means of arranged ophthalmologists or costly hardware, for instance, Heidelberg Retinal Tomography (HRT).CDR appraisal by an ophthalmologist is subjective and the availability of HRT is to a great degree confined. In this procedure CDR is computed from nonstereographic retinal fundus photographs. To naturally remove the disk, two strategies making usage of an edge detection and variational level-set strategy are proposed. For the cup, color component analysis and level-set strategy are utilized. To reshape the got circle and glass restrain from beforehand said methodologies, ellipse fitting is associated with the procured picture. To register the vertical cup to Disk Ratio (CDR), the optic cup and disk first should be distributed from the retinal images.. 2.1.1. Optic Disk Segmentation A coarse localization of optic disc region is presented using the red channel. The red component is utilized as it is found to have higher contrast between the optic disc and non-optic disc area than for other channels. To remove the blood vessels, a morphological closing operation is performed. After performing the closing operation, a median filter is applied to further smoothen the obtained image. The Canny approach is certain for edge detection due to the fact the Canny algorithm can come across edges with noise suppressed on the equal time. This method uses two thresholds, to discover robust and vulnerable edges, and it includes the susceptible edges within the output best if they're related to strong edges. The most advantageous threshold of the every input retinal image is observed to be exclusive due to the variation intensities in every image.
morphological opening operation in order to remove noise around the cup region. Then the edge detection and ellipse fitting is proposed to obtain the cup boundary smoothing. In threshold level set approach, the green channel of the input image is selected as the basis for further segmentation due to the optimum observed contrast between the cup and disc boundaries in this channel. The method to select the top 1/3 of the gray scale intensity is to find the threshold value from the normalized cumulative histogram and then compare it with all the intensity values of the input image. Then an intensity value is that is greater than the threshold value is selected. Text step is to use a morphological opening operation in order to remove noise around the cup region. Next, the intensityweighted centroid method is proposed to find an approximate initial point. This is found to give a good initial approximation for the initial cup region. Then, a threshold level-set algorithm is applied to segment the optic cup. 2.1.3. Ellipse Fitting for Optic Disc and Cup Ellipse fitting calculation can be utilized to smooth the optic cup and disk limit. Ellipse fitting is generally in view of the minimum square fitting algorithm which expects that the best-fit curve of a given sort is the curve that has the insignificant entirety of the deviations squared from, given information focuses. Direct least Square Fitting Algorithm is fit the optic disk from mainstream ellipse fitting algorithms. It is ellipse specific; in this manner the impact of noise around the disk range can be limited while shaping the ellipse. It can likewise be effectively comprehended normally by a generalized Eigen framework. 2.1.4. Calculate CDR The ratio of the size of the optic cup to the optic disc, also known as the cup-to-disc ratio (CDR), the value of CDR which is more than 0.65 is used to assess a patient as a possible glaucoma case. 2.2. Glaucoma Detection using Cup-Disk and
2.1.2. Optic Cup Segmentation Optic Cup Segmentation, two approaches are used. The colour component analysis approach and threshold level set approach. In the colour component analysis method, RGB components of the input images are analyzed, and it is found that the optic cup is more easily discriminated in the green image because the visibility and contrast of the optic cup is superior and its pixels are of higher intensities, while the neuroretinal rim and the retinal vessels are often of lower intensities. The next step is to use
Imperial Journal of Interdisciplinary Research (IJIR)
Rim-Disk Ratio From Fundus Images In this method, RGB fundus image is used as an input. In order to detect glaucoma, the most important region of interest is optic disk. Thus, instead of processing on the whole retinal image, region around optic disk is extracted. This ROI is a small image which helps in faster processing and large automated screening of glaucoma.
Page 1064
Imperial Journal of Interdisciplinary Research (IJIR) Vol-3, Issue-2, 2017 ISSN: 2454-1362, http://www.onlinejournal.in 2.2.1. Region of Interest
2.2.4. Neuroretinal Rim
In order to extract the region around optic disk, centre of optic disk is first computed [4]. A window of dimension r*column is created. Where r is the radius of optic disk (approximate).Then passing the window through the image spatially, for all rows and obtaining the maxima, we find the row of centre of optic disk. Another window of dimensions r*r is created. This window is passed through detected row for all the columns and the maxima are detected. Hence the coordinates of centre of optic cup is obtained. The ROI is defined as a square around optic disk centre with dimensions of thrice the optic disk diameter.
Neuroretinal rim is the region located between the edge of optic disk and optic cup. After segmenting of optic disk Fig1(a) and optic cup Fig 1(b), NRR is obtained by subtracting optic cup from optic disk. NRR is shown in Fig1(c).
2.2.2. Optic Disk Segmentation The extracted ROI image consists of three channels, red, green and blue. To detect optic disk, red channel is used as in this channel, optic disk appears to be the brightest and blood vessels are also suppressed in this channel. Hence it is easier and accurate to segment optic disk in red channel of input fundus image. Otsu thresholding technique is used to segment the optic disk which makes the segmentation independent of image quality. 2.2.3. Optic Cup Segmentation Optic cup is segmented using green component of ROI image. Statistical features such as mean and standard deviation for green channel of ROI image are calculated. These features help in making the method adaptive of image quality. Standard deviation is defined as the square root of variance. Histogram of green channel is examined. It is a graphical representation of number of pixels with respect to gray levels in digital image. The horizontal axis of histogram represents the tonal variation while the vertical axis of histogram represents total number of pixels in each tone. The histogram of each image is examined to determine threshold level for segmenting optic disk and optic cup. From the statistical patterns of histogram, it is known that the mean is the central tendency and the standard deviation is the dispersion from that central value. The addition of mean and standard deviation gives an intensity level that points to the highest number of pixels in the grey region of the image. After analysing the images, it is determined experimentally that optic cup lies at an intensity level given by Tcup = Mean + 3* Standard deviation Where Tcup is the intensity level which is decided as a threshold value for segmenting optic cup.
Imperial Journal of Interdisciplinary Research (IJIR)
(a)
(b)
(c)
Fig 1: (a) Segmented optic disk, (b) Segmented optic cup, (c) Neuroretinal rim 2.2.5. Classification
Classification of images is done on the basis of two parameters; Cup to disk ratio and rim to disk ratio [4]. CDR is defined as ratio of total segmented cup area to total segmented disk area. CDR = Optic Cup Area/Optic Cup Area. Cup area and disk area are obtained by summing all the white pixels in segmented cup and disk. This calculated CDR is used for screening of Glaucoma. If CDR is greater than 0.3(globally accepted value), the fundus image under test is said to be glaucomatic else it is healthy. RDR= Rim area in infero-temporal region/ Disk Area . After analysing a database of 50 images, value of 0.4 is decided as threshold for classifying images as glaucomatic or not. If RDR is less than equal to 0.4, fundus image is considered to be glaucomatic else it is glaucoma free. Both these parameters are calculated to train the SVM classifier for detecting glaucoma so as to increase the reliability and robustness of the system.
2.3. Texture and Detection
HOS
features
Based
This method for glaucoma detection using a combination of texture and higher order spectra (HOS) features from digital fundus images. Support vector machine, sequential minimal optimization, naĂŻve Bayesian, and random-forest classifiers are used to perform supervised classification. Our results demonstrate that the texture and HOS features after z-score normalization and feature selection, and when combined with a random-forest classifier, performs better than the other classifiers and correctly identifies the glaucoma images with an accuracy of more than 91% [2].
Page 1065
Imperial Journal of Interdisciplinary Research (IJIR) Vol-3, Issue-2, 2017 ISSN: 2454-1362, http://www.onlinejournal.in 2.3.1. Image Pre-processing The preprocessing step consists of image contrast improvement using histogram equalization and radon transform was performed for HOS feature extraction. Histogram equalization increases the dynamic range of the histogram of an image and assigns the intensity values of pixels in the input image such that the output image contains a uniform distribution of intensities. As a result, the contrast of the image is increased. Radon transform is used to detect features within an image. Radon transform transforms lines through an image to points in the radon domain, where each point in this domain is transformed to a straight line in the image. This radon transformation is used before extracting the HOS parameters from the image. 2.3.2. Feature Extraction In this technique, removed two sorts of features:1) HOS parameters and 2) texture descriptors. Brief clarifications of these components are given in the accompanying. Higher Order Spectra evokes both amplitude and phase data of a given signal. It offers great noise immunity and yields great outcomes, even for weak and noisy signals. HOS comprise of moment and cumulant spectra and can be utilized for both deterministic signals and random processes [10]. We got the components from the third-order statistics of the signal, in particular, the bispectrum. The bispectrum is given by, B(f1, f2) = E[X(f1 )X(f2 )X ∗ (f1 + f2 )],where X(f) is the Fourier transform of the signal x(nT), and E[.] stands for the expectation operation. Features are calculated by integrating the bispectrum along the dashed line with slope = a. Frequencies are normalized by the Nyquist frequency. These bispectral invariants [11] contain information about the shape of the waveform within the window and are invariant to shift and amplification and robust to time-scale changes. In this study, we used these bispectral invariant features for every 20◦. Bispectral entropies have been derived from bispectrum plots to find the rhythmic nature of the heart rate variability and electroencephalogram signals [12], [13]. The normalization in the equations ensures that entropy is calculated for a parameter that lies between 0 and 1 (as required of a probability and, hence, the entropies (Ent1, Ent2, and Ent3) computed are also between 0 and 1. 2.3.3. Texture Features Texture descriptors provide measures of properties, such as smoothness, coarseness, and regularity, which indicate a mutual relationship among intensity values of neighboring pixels
Imperial Journal of Interdisciplinary Research (IJIR)
repeated over an area larger than the size of the relationship. Such properties can be used as features for pattern recognition. Co-Occurrence Matrix: A gray-level co-occurrence matrix (GLCM) depicts how often different combinations of pixel brightness values (gray levels) occur in an image. It is a second order measure because it measures the relationship between neighborhood pixels. 2.3.4. Classifiers used The classifiers were chosen based upon their effectiveness in capturing the discriminative properties of these features, the impact of the ranking of features and efficiency, and the efficacy of the classification results. Four classifiers were employed for supervised learning and testing: SVM [21], SMO, random forest, and NB. Hardware consistency was maintained during the evaluation of these classifiers.
3. Comparative Study This section compares each glaucoma detection methods. The CDR based method performs by using k-means and fuzzy c means clustering. Texture and HOS features based glaucoma detection using naïve Bayesian, SVM, and random forest classifiers for better accuracy. SVM classifier is used for the Cup Disk and Rim Disk Ratio based glaucoma detection. Table 1 shows the evaluation summary of these methods. Table 1. Comparison of different Glaucoma Detection Methods
Method CDR
Pre-processing Technique ROI extraction
Texture and HOS features
Z-score normalization
Cup Disk and Rim Disk Ratio
ROI extraction
Classifier
Accuracy
K means Fuzzy cmeans NaïveBayesian SVM Randomforest SVM classifier
95%
91%
90 %
4. Conclusion Glaucoma permanently ceases eye sight .There are no such noticeable symptoms that able to detect the glaucoma in the early stage. Many automated detection techniques have been proposed for the glaucoma detection. Most of the techniques have used fundus images and OCT images. Current Page 1066
Imperial Journal of Interdisciplinary Research (IJIR) Vol-3, Issue-2, 2017 ISSN: 2454-1362, http://www.onlinejournal.in methods for the automatic detection of glaucoma are based on segmenting the disc and cup which is the crucial step. Early detection helps in recovery of glaucoma patients. The first method uses Morphological techniques to extract two major features for detection of Glaucoma i.e. Area ratio of NRR in ISNT quadrants, Cup to Disc Ratio and achieves an average accuracy of 97.5 percentage. An efficient technique to detect glaucoma is proposed in cup disk and rim disk ratio method. In order to classify the input fundus image as glaucomatic or healthy, two different ocular parameters, cup to disk ratio and rim to disk ratio are considered. Adaptive image threshold technique is used to segment optic disk and optic cup which makes the proposed method independent of image quality and invariant to noise. In Texture and HOS method, the features are used for classification. The random-forest classifier, combined with z-score normalization and featureselection methods, performed the best among the four classifiers with a classification accuracy of more than 91 percentage. The glaucoma detection technique by using CDR is more accurate than other methods.
5. References [1] Joshi GD, Sivaswamy J,Krishnadas SR “Optic disk and cup segmentation from monocular color retinal images for glaucoma assessment”, IEEE Trans. Med. Imag., vol. 30, no. 6, pp. 1192–1205,Jun. 2011. [2] U. Rajendra Acharya, Sumeet Dua, Xian Du, Vinitha Sree S, and Chua Kuang Chua,” Automated Diagnosis of Glaucoma Using Texture and Higher Order Spectra Features” IEEE Transactions on Information Technology in Biomedicine, vol. 15, no. 3, may 2011.
[7] D. Wong, J. Liu, J. Lim, X. Jia, F. Yin, H. Li, and T. Wong,”Level- set based automatic cup-to-disc ratio determination using retinal fundus images in ARGALI”, Engineering in Medicine and Biology Society, pp. 22662269, 2008. [8] S. Kavitha, S. Karthikeyanand Dr. Duraiswamy, “Early Detection of Glaucoma in Retinal Images Using Cup to Disc Ratio,” Computing Communication and Networking Technologies (ICCCNT),pp. 1-5, July 2010. [9] Chalinee Burana-Anusorn1, Waree Kongprawechnon1, Toshiaki Kondo1, Sunisa Sintuwong2and KanokvateTungpimolrut3 Thammasat “Image Processing Techniques for Glaucoma Detection Using the Cupto- Disc Ratio,” International Journal of Science and Technology, Vol. 18, No. 1, January-March 2013. [10] C. L. Nikias and A. P. Petropulu,”Higher-Order Spectra Analysis: A Nonlinear Signal Processing Framework”. Englewood Cliffs, NJ: Prentice Hall, 1993. [11] V. Chandran and S. L. Elgar, “Pattern recognition using invariants defined from higher order spectra onedimensional inputs,” IEEE Trans. Signal Process., vol. 41, no. 1, pp. 205–212, Jan. 1993. [12] K. C. Chua, V. Chandran, U. R. Acharya, and C. M. Lim, “Cardiac state diagnosis using higher order spectra of heart rate variability,” J. Med. Eng. Technol., vol. 32, no. 2, pp. 145–155, Mar. 2008. [13] C. L. Nikias and A. P. Petropulu,”Higher-Order Spectra Analysis: A Nonlinear Signal Processing Framework”, Englewood Cliffs, NJ: Prentice Hall, 1993.
[3] Hafsah Ahmad, Abubakar Yamin, Aqsa Shakeel, Syed Omer Gillani, Umer Ansari,” Detection of Glaucoma Using Retinal Fundus Images”, 2014 International Conference on Robotics and Emerging Allied Technologies in Engineering (iCREATE) Islamabad, Pakistan, April 22-24, 2014. [4] Ayushi Agarwal, Shradha Gulia, Somal Chaudhary, Malay Kishore Dutta, Carlos M. Travieso, Jesús B. Alonso-Hernández,” A Novel Approach to Detect Glaucoma in Retinal Fundus Images using Cup-Disk and Rim-Disk Ratio”, 2015 International Work Conference on Bio-inspired Intelligence (IWOBI). [5] J. Liu, D. Wong, J. Lim, H. Li, N. Tan, Z. Zhang, T. Wong and R.Lavanya, ARGALI, “An Automatic Cup-toDisc Ratio Measurement System for Glaucoma Analysis Using Level-set Image Processing”, Proc. IFMBE, pp. 559-562, 2009. [6] J. Liu, D. Wong, J. Lim, X. Jia, F. Yin, H. Li, and T. Wong,” Optic cup and disk extraction from retinal fundus images for determination of cup-to-disc ratio”, Industrial Electronics and Applications, pp. 1828-1832, June 2008.
Imperial Journal of Interdisciplinary Research (IJIR)
Page 1067