No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Automatic detection of hard and soft exudates in fundus images using color histogram thresholding
Abstract
Diabetic retinopathy is considered as the root cause of vision loss for diabetic patients. However, if the symptoms are identified earlier and a proper treatment is provided through regular screenings, blindness can be avoided. In order to lessen the cost of these screenings, modern image processing techniques are used to voluntarily detect the existence of abnormalities in the retinal images acquired during the screenings. Exudates are a major indicator of diabetic retinopathy that can possibly be quantified automatically. This paper focuses on automatic detection of diabetic retinopathy exudates in color fundus retinal images. A series of experiments on classification of hard and soft exudates is performed with the use of image processing techniques. Initially the color fundus retinal images are subjected to preprocessing for CIELab color space conversion and Fundus region detection using binarization and mathematical morphology respectively. Subsequently nonlinear diffusion segmentation is employed to encapsulate the variation in exudates and lesion boundary criteria pixels. To prevent the optic disc from interfering with exudates detection, the optic disc is detected and localized with the aid of region props and color histogram. Exudates are detected with the aid of thresholding color histogram, which is used to classify the hard and soft exudates pixel from the color fundus retinal image. Experimental evaluation on the publicly available dataset DIARETDB1 demonstrates the improved performance of the proposed method for automatic detection of Exudates. These automatically detected exudates are validated by comparing with expert ophthalmologists' hand-drawn ground-truths. Sensitivity, Specificity and Accuracy are used to evaluate overall performance. The overall sensitivity, specificity and accuracy of the proposed method are 89.78%, 99.12% and 99.07%, respectively.
... A method to detect exudates automatically from colour fundus retinal image using colour histogram thresholding is presented in [10]. The preprocessing of colour fundus retinal images is done using colour space conversion and fundus region detection. ...
... Although the techniques presented in [6][7][8][9][10][11] have given algorithms to identify exudates, they do not mention how soft and hard exudates are separated. In comparison to previous techniques published, our proposed algorithm clearly indicates how soft and hard exudates are separated. ...
... As RGB to HSI conversion is straightforward, it is not required to convert the RGB image to grey-scale as RGB to grey-scale conversion is not unique. To eliminate the optic disc, it is assumed that the component with the largest circular shape of a fundus image is the optic disc [8,10]. The second stage involves classification of exudates as hard exudates using fuzzy logic. ...
Diabetic retinopathy (DR), that affects the blood vessels of the human retina, is considered to be the most serious complication prevalent among diabetic patients. If detected successfully at an early stage, the ophthalmologist would be able to treat the patients by advanced laser treatment to prevent total blindness. In this study, the authors propose a technique based on morphological image processing and fuzzy logic to detect hard exudates from DR retinal images. At the initial stage, the exudates are identified using mathematical morphology that includes elimination of the optic disc. Subsequently, hard exudates are extracted using an adaptive fuzzy logic algorithm that uses values in the RGB colour space of retinal image to form fuzzy sets and membership functions. The fuzzy output for all the pixels in every exudate is calculated for a given input set corresponding to red, green and blue channels of a pixel in an exudate. This fuzzy output is computed for hard exudates according to the proportion of the area of the hard exudates. By comparing the results with hand-drawn ground truths, the authors obtained sensitivity and specificity of detecting hard exudates as 75.43 and 99.99%, respectively.
... In the past few decades, numerous researchers have devoted themselves to solving the segmentation of diabetic retinopathy. In early years, the researchers focused on traditional image processing methods, such as morphological operations and threshold segmentation [7][8][9]. Limited by the heavy dependence of the design level, the traditional methods of lesion segmentation are relatively infeasible in real-world application. ...
... Antal and Hajdu [11] adopted an ensemble learning strategy to integrate a series of image preprocessing approaches to improve final segmentation of microaneurysms. Kavitha and Duraiswamy [8] extracted exudate features using a multilayer threshold method, but this model has requirements for the input image quality. In conclusion, the traditional methods of lesion segmentation are relatively inefficient with poor generalization. ...
Diabetic retinopathy is one of the main causes of blindness in human eyes, and lesion segmentation is an important basic work for the diagnosis of diabetic retinopathy. Due to the small lesion areas scattered in fundus images, it is laborious to segment the lesion of diabetic retinopathy effectively with the existing U-Net model. In this paper, we proposed a new lesion segmentation model named FFU-Net (Feature Fusion U-Net) that enhances U-Net from the following points. Firstly, the pooling layer in the network is replaced with a convolutional layer to reduce spatial loss of the fundus image. Then, we integrate multiscale feature fusion (MSFF) block into the encoders which helps the network to learn multiscale features efficiently and enrich the information carried with skip connection and lower-resolution decoder by fusing contextual channel attention (CCA) models. Finally, in order to solve the problems of data imbalance and misclassification, we present a Balanced Focal Loss function. In the experiments on benchmark dataset IDRID, we make an ablation study to verify the effectiveness of each component and compare FFU-Net against several state-of-the-art models. In comparison with baseline U-Net, FFU-Net improves the segmentation performance by 11.97%, 10.68%, and 5.79% on metrics SEN, IOU, and DICE, respectively. The quantitative and qualitative results demonstrate the superiority of our FFU-Net in the task of lesion segmentation of diabetic retinopathy.
... From some of the research investigations we can identify the severe stages of diabetic retinopathy in the form of cotton wool spots which causes nerve fiber layer blocking and blown up [5,6], along with the other pathological parameters like optic disc, optic cup, Blood Vessels, Micro aneurysms Hemorrhages detection and lesions like hard and soft exudates [7,8]. Figure 1 shows pathological features of retinal image of typical diabetic retinopathy [9]. This paper focus on the major contributions and review towards the detection methods pertaining to Nonproliferative DR to analyze the severity of disease in fundus images. ...
... This paper focus on the major contributions and review towards the detection methods pertaining to Nonproliferative DR to analyze the severity of disease in fundus images. Retinopathy image [9]. ...
Diabetic retinopathy is caused by individuals who are suffering from long-run diabetes for more than 10 to 15 years. And it also can be varied from years of having diabetes. Generally, it is treated as a case which leads to damages to blood cell These blood vessels are further extended and represented as blood leaking and fluids on theretina to form different anatomical structures like microaneurysms,haemorrhages, hard exudates, cotton wool spots. Vigilant treatment and monitoring of the eyes couldreduce at least 90% of blindness in new cases. Beneficial preventing visual impairment and blindness can be achieved by early diagnosis with regular screening and treatment.This work focus on the review of automatic detection of Non-proliferative type of diabetic retinopathy with various methods used to detect the symptoms of the disease
in early stages to avoid blindness with the most recent developments and methods using new software version.
... This method fails to get rid of some artifacts. Kavitha et al [12] used a method based on color histogram thresholding for identification of hard and soft exudate pathologies in retinopathy images. The overall sensitivity, specificity and accuracy obtained by this approach were 89.78%, 99.12% and 99.07%, respectively. ...
... Therfore, a preprocessing step of the fundus photographs are needed before detecting hard exudates. Initially, the original color fundus retinal images of red, green and blue (RGB) space are transformed to CIELab color space [12] (figure 3.b). This color space is a the second of two systems adopted by CIE in 1976 as models that better showed uniform color spacing in their values. ...
Diabetic retinopathy is a severe and widely spread eye disease. Early diagnosis and timely treatment of these clinical signs such as hard exudates could efficiently prevent blindness. The presence of exudates within the macular region is a main hallmark of diabetic macular edema and allows its detection with high sensitivity. In this paper, we combine the k-means clustering algorithm and mathematical morphology to detect hard exudates (HEs) in retinal images of several diabetic patients. This method is tested on a set of 50 ophthalmologic images with variable brightness, color, and forms of HEs. The algorithm obtained a sensitivity of 95.92%, predictive value of 92.28% and accuracy of 99.70% using a lesion-based criterion.
... In the SEs detection, the method proposed by Kavitha et al. [13] was employed, where the histogram information of the color image was used to detect HEs and SEs from the retinal image. Then the HEs image detected in the study [12] is subtracted from the resulting image of this method leaving it with only SEs. ...
... When light hits the retina, a series of chemical and electrical reactions begin, culminating in the transmission of nerve impulses. These signals go via the optic nerve and are processed in the various visual areas of the brain [7]. The retinal, which functions similarly to the film of a camera, is the substrate upon which the visual world is recorded when the optic nerve transmits visual information from the retina to the brain. ...
Recent advances in digital image analysis have improved diagnostic procedures and given healthcare experts vital information. Previously, ophthalmologists analysed medical images, which were limited by their non-systematic search pattern, image noise, and illness complexity. CAD systems were developed by combining image data and clinical information. This technology detects lesions, assesses disease extent, and supports diagnostic decisions to improve healthcare systems. diabetic retinopathy has increased mortality. Using a computer-aided decision support system can help ophthalmologists detect diabetic retinopathy early and reduce blindness. Retinal image screening classifies and grades retinal images effectively. Decision assistance systems were developed using many methods. Using a variety of samples and experts, a computer-aided diagnostic system was developed and evaluated to categorise retinal images using SVM. It can be concluded from the results that the accuracy with which retinal pictures are categorised as normal or abnormal increases by using several criteria and selecting the SVM classifier. There is no correlation between the size of the input space and the computational complexity of SVM. Therefore, it boosts classification precision across the board by decreasing erroneous acceptance and false rejection rates. The SVM classifier has been found to have both a lower false acceptance rate and a lower false rejection rate than competing classifiers. After putting the suggested system through its paces, its accuracy was determined to be 97.65 percent. The proposed technology can act as a third party observer in clinical decision-making. Images of lesions had an average sensitivity of 96.89 percent, a specificity of 98.76 percent, and an accuracy of 98.1 percent. The experimental findings prove that the proposed technique has better sensitivity, specificity, accuracy, and predictive values than the alternatives. The proposed method produces grading outcomes that are comparable to those obtained using other methods.
... In their experiment, the best BPN performance showed 98.45% accuracy. Kavitha and Duraiswamy [6] did some research on automatic detection of hard and soft exudates in fundus images, using color histogram thresholding to classify exudates. Their experiments showed 99.07% ...
... In their experiment, the best BPN performance showed 98.45% accuracy. Kavitha and Duraiswamy [37] did some research on automatic detection of hard and soft exudates in fundus images, using color histogram thresholding to classify exudates. eir experiments showed 99.07% ...
Diabetic retinopathy (DR) is a complication of long-standing diabetes, which is hard to detect in its early stage because it only shows a few symptoms. Nowadays, the diagnosis of DR usually requires taking digital fundus images, as well as images using optical coherence tomography (OCT). Since OCT equipment is very expensive, it will benefit both the patients and the ophthalmologists if an accurate diagnosis can be made, based solely on reading digital fundus images. In the paper, we present a novel algorithm based on deep convolutional neural network (DCNN). Unlike the traditional DCNN approach, we replace the commonly used max-pooling layers with fractional max-pooling. Two of these DCNNs with a different number of layers are trained to derive more discriminative features for classification. After combining features from metadata of the image and DCNNs, we train a support vector machine (SVM) classifier to learn the underlying boundary of distributions of each class. For the experiments, we used the publicly available DR detection database provided by Kaggle. We used 34,124 training images and 1,000 validation images to build our model and tested with 53,572 testing images. The proposed DR classifier classifies the stages of DR into five categories, labeled with an integer ranging between zero and four. The experimental results show that the proposed method can achieve a recognition rate up to 86.17%, which is higher than previously reported in the literature. In addition to designing a machine learning algorithm, we also develop an app called “Deep Retina.” Equipped with a handheld ophthalmoscope, the average person can take fundus images by themselves and obtain an immediate result, calculated by our algorithm. It is beneficial for home care, remote medical care, and self-examination.
... Li et al [10] (2009) proposed an algorithm that consists on the division of the fundus image into 64 sub-images, followed by the application of the region growth algorithm in order to segment the exudates. Kavitha et al [11] (2011) used a method based on the histogram thresholding of the of the fundus images for the detection of exudates. Amel Feroui [12] (2012) proposed a method for the detection of exudates in fundus eye images based on the cooperation between the Fuzzy classification algorithms (Fuzzy C-Means) and the morphological operators. ...
Diabetes might be the cause of lots of visual troubles (when it is poorly controlled); it induces the thickening and hardening of the retina’s blood vessels and prevents the accomplishment of its function suitably. Among the complications of diabetes on the eye, there is the diabetic retinopathy, which is considered as a major cause of blindness in the industrialized countries. This paper presents a system of fundus eye image analysis, whose purpose is the extraction of lesions related to the non-proliferative diabetic retinopathy (micro-aneurysm, hemorrhages and exudates). The detection of these lesions is essentially based on the detection of the nonpathological structures in the retina (blood vessels, optic disc and fovea) and their elimination in order to have just the pathological structures in the result.
... The major factor for inhomogeneity, in terms of luminosity and contrast lies in the same image or between images. In space conversion was given in [12]. The variation is correlated with the skin pigmentation and iris color of the different person. ...
The principal target of preprocessing is to get more appropriate resultant image than its original for further additional analysis. Enhancement of retinal images creates several challenges. The main obstacle is to develop a technique to accommodate the wide variation in contrast inside the image. Necessity of preprocessing methods are for image normalization and to increase the contrast for achieving accurate analysis. This work examined literature in the prior process of digital imaging, in the field of the analysis of fundus image to extract normal and pathologic retinal traits within the context of diabetic retinopathy (DR).
... Automatic detection of hard and soft exudates using histogram thresholding 4 is described. The fundus image is pre-processed using CIELab color space and then used mathematical morphology for the detection of hard and soft exudates. ...
This paper aims to find an efficient hybrid method to diagnose diabetic retinopathy, which is an anomaly in the human eyes that occur due to the decrease of insulin content in the blood. Damages to the blood vessels in the light-sensitive tissue of the eye is its root cause. The symptoms of diabetic retinopathy are hemorrhages, exudates and micro-aneurysms. Eventually it will lead to total blindness. This erratic disorder is developed in people having both type-1 and type-2 diabetes. The longer period of time you have uncontrolled blood sugar levels, it is more likely that this condition of diabetic retinopathy may arise. Since the number of diabetic retinopathy patients are high in number, the significance of automating the diagnostic process is much more relevant. In order to diagnose this disease automatically, a hybrid and efficient system has been developed to interpret and analyse the 2D fundus images. Grayscale conversion and Contrast Level Adaptive Histogram Enhancement (CLAHE) has been performed as a pre-processing step to improve the quality of the input image which will further aid in blood vessel extraction and exudates determination in a better way. The pre-processed image is further manipulated with the Kirsch’s template for the blood vessel extraction. Subsequently, the features of the images are extracted from the morphologically processed images through a multi-level Maximally Stable Extremal Regions (MSER) to precisely extract and identify the exudates from the eye. The determination of exudates helps the ophthalmologist to diagnose the diabetic retinopathy and further proceed with respective treatment.
... Li et al [10] (2009) proposed an algorithm that consists on the division of the fundus image into 64 sub-images, followed by the application of the region growth algorithm in order to segment the exudates. Kavitha et al [11] (2011) used a method based on the histogram thresholding of the of the fundus images for the detection of exudates. Amel Feroui [12] (2012) proposed a method for the detection of exudates in fundus eye images based on the cooperation between the Fuzzy classification algorithms (Fuzzy C-Means) and the morphological operators. ...
Diabetes might be the cause of lots of visual troubles (when it is poorly controlled); it induces the thickening and hardening of the retina's blood vessels and prevents the accomplishment of its function suitably. Among the complications of diabetes on the eye, there is the diabetic retinopathy, which is considered as a major cause of blindness in the industrialized countries. This paper presents a system of fundus eye image analysis, whose purpose is the extraction of lesions related to the non-proliferative diabetic retinopathy (micro-aneurysm, hemorrhages and exudates). The detection of these lesions is essentially based on the detection of the non-pathological structures in the retina (blood vessels, optic disc and fovea) and their elimination in order to have just the pathological structures in the result.
... Many attempts have been made towards automated detection of DR based on features obtained from HEs in CF images [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25]. However, very limited efforts are made towards automated quantification of HEs. ...
Background
Hard exudates (HEs) are the classical sign of diabetic retinopathy (DR) which is one of the leading causes of blindness, especially in developing countries. Accordingly, disease screening involves examining HEs qualitatively using fundus camera. However, for monitoring the treatment response, quantification of HEs becomes crucial and hence clinicians now seek to measure the area of HEs in the digital colour fundus (CF) photographs. Against this backdrop, we proposed an algorithm to quantify HEs using CF images and compare with previously reported technique using ImageJ.
Methods
CF photographs of 30 eyes (20 patients) with diabetic macular edema were obtained. A robust semi-automated algorithm was developed to quantify area covered by HEs. In particular, the proposed algorithm, a two pronged methodology, involved performing top-hat filtering, second order statistical filtering, and thresholding of the colour fundus images. Subsequently, two masked observers performed HEs measurements using previously reported ImageJ-based protocol and compared with those obtained through proposed method. Intra and inter-observer grading was performed for determining percentage area of HEs identified by the individual algorithm.
Results
Of the 30 subjects, 21 were males and 9 were females with a mean age of the 50.25 ± 7.80 years (range 33–66 years). The correlation between the two measurements of semi-automated and ImageJ were 0.99 and 0.99 respectively. Previously reported method detected only 0–30% of the HEs area in 9 images, 30–60% in 12 images and 60–90% in remaining images, and more than 90% in none. In contrast, proposed method, detected 60–90% of the HEs area in 13 images and 90–100% in remaining 17 images.
Conclusion
Proposed method semi-automated algorithm achieved acceptable accuracy, qualitatively and quantitatively, on a heterogeneous dataset. Further, quantitative analysis performed based on intra- and inter-observer grading showed that proposed methodology detects HEs more accurately than previously reported ImageJ-based technique. In particular, we proposed algorithm detect faint HEs also as opposed to the earlier method.
... In the process, AHE [19] is used and then procedure divides the image into regions and then the technique i.e. histogram equalization is performed on each region. The distribution of intensity values displays hidden things of an image more observable. ...
Diabetic Retinopathy (DR) is the problem for long-standing diabetes and the major cause for visual loss is because of variations in blood vessels of the retina. Vision loss is due to DR is preventable with regular checkups at the earlier stages. The presence of exudates is the basic signs of DR. Hence, exudates detection becomes a major problem-solving task, in which retinal images plays a vital role. In this paper, a technique is proposed depending on morphological closing operations which are in image processing is used to identify exudates from retinal image. At the starting phase, the retinal image must be preprocessed, where at first the image which is in RGB color format is altered to the image of HSI color format and on Intensity-band of HSI color format average filter is performed to reduce noise and subsequently contrast-limited adaptive histogram equalization is performed to normalize the image. In subsequent phase, exudates are identified using morphological closing operation and then Nilblack's thresholding is applied. On comparing the results, the proposed technique got better performance than the existing technique.
... The detection of exudates using morphological approach is very extensive; couple of the important morphological techniques is discussed here. Kavitha et al. proposed an automatic detection of hard and soft exudates [35] using histogram thresholding. They illustrated the fundus images using CIE Lab color space and then used mathematical morphology for the detection of hard and soft exudates. ...
Diabetic Retinopathy is a progressive eye diseases that causes changes in the blood vessels of the retina which may cause blindness if not prevented and treated at early stage. Diabetic retinopathy is one of the complications caused by diabetes and it appears in the retina, which is the tissue responsible for the vision in the eye. The early detection and diagnosis is essential to save the vision of diabetic patients. The indications of diabetic retinopathy on the surface of the retina are microaneurysms, hemorrhages and exudates. The identification of exudates by opthalmolgists normally requires dilation of pupil's eye using a chemical solution which consumes lot of time and affects patients. The presented work investigates and presents a morphology based techniques for the detection of diabetic retinopathy through exudates from color fundus images which includes the elimination of optic disc and the detected exudates are classified using image processing methods.
... Automatic detection of hard and soft exudates using histogram thresholding is described by Kavitha and Duraiswmy [8]. They preprocessed the fundus images using CIELab colorspace and then used mathematical morphology for the detection of hard and soft exudates. ...
Retinopathy is a diabetic-related complication and main cause of vision loss of the diabetic patients. Exudates are the major sign of diabetic retinopathy. This paper presents a morphology-based method for the detection of diabetic retinopathy through exudates from color fundus images. We applied our approach on fundus images and obtained satisfactory results, which are compared with the ophthalmologists' hand-drawn ground truths.
... These systems can be used to grade the DR [6,7] or to detect different symptoms of DR. There are symptoms that appear in the color images such as microaneurysms and hemorrhages that occur as round small dots or blots [8] and exudates that appear as yellowish color [9,10]. But there is another symptom that does not occur in the color images such as the retinal ischemia that appears only in the fundus fluorescein angiograms (FFAs). ...
When the blood vessels stop supplying any region in the retina of the eye, this region is called capillary non-perfusion (CNP). With increasing and spreading of these regions across the retina, the patient can go blind. These regions appear only in the fundus fluorescein angiograms (FFAs). In this paper, an algorithm to automate the segmentation and classification of these regions is presented. The segmentation algorithm consists of three main steps: pre-processing, vessels extraction and CNP segmentation. After that, the automatic classification is applied to determine the severity level of each image. In the segmented algorithm, the CNPs are extracted by using the region growing algorithm. The algorithm is tested on 88 FFA images and evaluated by using two different ground truth images. The severity level is classified based on the percentage of the CNPs in each image. © 2015, WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE, All rights Reserved.
... S. Kavitha et al [8] presents an automatic detection system of diabetic retinopathy exudates in color fundus retinal images. Initially the color fundus retinal images are subjected to preprocessing for CIELab color space conversion and Fundus region detection using binarization and mathematical morphology respectively. ...
... These systems could reduce the workload of ophthalmologists and aid in the early diagnosis of the disease to make the treatment more efficient [3,4]. There are several automatic systems which detect different symptoms of DR which appear in the color images such as microaneurysms and hemorrhages that occur as round small dots or blots [5], exudates that appear as yellowish color [6,7] or to grade the DR [8,9]. But there is another symptom that does not occur in the color images such as the retinal ischemia that needs another type of images to be detected. ...
The regions in the retina of the eye, where the blood supply stops, are called capillary non-perfusion
(CNP). These regions appear only in the fundus fluorescein angiograms (FFAs). If these regions spread across the retina the patient can go blind. In this paper, an algorithm to automate the segmentation of these regions is presented. This algorithm consists of three steps: pre-processing, vessels extraction and CNP segmentation. At the pre-processing stage, the region of interest (ROI) is determined, the contrast of the image is enhanced and the noise is removed. Then, the blood vessels are extracted by using the mathematical morphology in the next stage. In the final stage, the CNPs are segmented by using the region growing algorithm. The algorithm is tested on 88 FFA images and evaluated by using the ground truth images. The obtained results achieved an average sensitivity of 96.4%, specificity of 99.7%, accuracy of 97.7%, and ischemic regions detection of 97%.
... Promising results were reported. It is proposed in [15] to first subject the color fundus retinal images to CIELab color space conversion and fundus region detection using binarization and mathematical morphology. Then nonlinear diffusion segmentation is applied for encapsulation of the variation in exudates and lesion boundary pixels. ...
The chronic hyperglycemia of diabetes is associated with long-term damage, dysfunction, and failure of different organs, especially the eyes, kidneys, nerves, heart, and blood vessels. The regular examination of diabetic patients can potentially reduce the risk of vision impairment and in the last instance blindness. Early diabetic retinopathy detection enables application of laser therapy treatment in order to prevent or delay loss of vision. The diagnostics and detection of diabetic retinopathy is performed by specialized ophthalmologists manually and represents expensive procedure. Automatic exudates detection and retina images classification would be helpful for reducing diabetic retinopathy screening costs and encouraging regular examinations. We proposed the automated algorithm that applies mathematical modeling which enables light intensity levels emphasis, easier exudates detection, efficient and correct classification of retina images. The proposed algorithm is robust to various appearance changes of retinal fundus images which are usually processed in clinical environments.
... In [16] fundus region detection is proposed, using binarization and mathematical morphology. Then nonlinear diffusion segmentation is applied for encapsulation of the variation in exudates and lesion boundary pixels. ...
SUMMARY Diabetic retinopathy is the progressive pathological alterations in the retinal microvasculature that very often causes blindness. Because of its clinical significance, it will be helpful to have regular cost-effective eye screening for diabetic patients by developing algorithms to perform retinal image analysis, fundus image enhancement, and monitoring. The two cost-effective algorithms are proposed for exudates detection and optic disk extraction aimed for retinal images classification and diagnosis assistance. They represent the effort made to offer a cost-effective algorithm for optic disk identification, which will enable easier exudates extraction, exudates detection and retinal images classification aimed to assist ophthalmologists while making diagnoses. The proposed algorithms apply mathematical modeling, which enables light intensity levels emphasis, easier optic disk and exudates detection, efficient and correct classification of retinal images. The algorithm is robust to various appearance changes of retinal fundus images and shows very promising results. Fundus images are classified into those that are healthy and those affected by diabetes, based on the detected optic disk and exudates. The obtained results indicate that the proposed algorithm successfully and correctly classifies more than 98% of the observed retinal images because of the changes in the appearance of retinal fundus images typically encountered in clinical environments. Copyright © 2014 John Wiley & Sons, Ltd.
Diabetic Retinopathy is a major problem for diabetic patients and is caused by changes in the blood vessels and abnormalities in the macular region. This disease will lead to vision loss if the diabetes is not controlled. This disease can be encountered by the primary signs of Microaneuryusm, Haemorrhages and Exudates. In this paper, a methodology is proposed for Hardware-based Detection of Exudates using RAPIDS CUDA Segmentation algorithm. The Input retinal image is pre-processed by using the Gray scale conversion method and Contrast Limited Adaptive Histogram technique. The pre-processed retinal image is further segmented by using K-mean Segmentation technique and the exudates are detected from the segmented output region. In this proposed approach RAPIDS CUDA Machine Learning algorithm runs completely on the GPU architecture which has increased the speed of the execution with less execution time.
WHO projects that diabetes will be the 7th major cause leading death in 2030. Diabetic Retinopathy caused by leakage of blood or fluid from the retinal blood vessels and it will damage the retina. For detection and extraction of non-proliferative diabetic retinopathy lesion we have invent the new wavelet filter. The proposed filter give the good extraction result as compare to exiting wavelet filter. In proposed algorithm, we have extract the microaneurysms, hemorrhages, exudates and retinal blood vessels. After extraction of lesions, grading is done by using feed forward neural network. The proposed algorithm achieves sensitivity of 98%, specificity of 92% and accuracy of 98%.
Diabetic retinopathy (DR) is a disease related to eye correlated with long-standing diabetes. It is a leading cause of blindness among working adults. Detection of this condition in the early stage is critical for good prognosis. Present day detection of DR normally requires digital fundus image or images generated using optical coherence tomography (OCT). As OCT are high-priced, diagnosis of DR using fundus image will benefit for the patient and the ophthalmologists. Manual inspection of morphological changes in blood vessels, microaneurysms, exudates, hemorrhages, and macula are time consuming and tedious tasks. So, designing a computer-aided system helps in analyzing the morphological changes and identifying the DR. This chapter reviews the applications of machine learning and deep learning algorithms for detection of nonproliferative diabetic retinopathy by analyzing fundus images.
Diabetic retinopathy, otherwise called diabetic eye illness, is a therapeutic condition in which damage occurs to the retina because of diabetes and is a main source of visual deficiency. As exudates (exudates are mass of cells and fluid that has seeped out of blood vessels or an organ) are among early clinical indications of diabetic retinopathy, their location would be a basic resource for the mass screening errand and fill in as an essential. A procedure is proposed which depends on morphological image processing and fuzzy logic to recognize hard exudates from diabetic retinopathy retinal image in this dissertation. At the underlying stage, the exudates are distinguished utilizing mathematical morphology that incorporates image preprocessing utilizing HSV colour model and elimination of optic disc. The hard exudates are separated utilizing an adaptive fuzzy logic algorithm that utilizations values in the RGB colour space of retinal image to form fuzzy sets and membership function. The fuzzy output for all the pixels in every exudate is calculated for a given input set corresponding to red, green and blue channels of a pixel in exudates. Since, digital image is formed from combination of pixels, during image acquisition process, the quality of the image diminishes from the point they are captured. To get a quality image, image quality metrics are applied on the proposed algorithm. Then, fuzzy output is computed for hard exudates according to the proportion of the hard exudates detected. By comparing the results with hand-drawn ground truths, it has been obtained that the sensitivity and specificity of detecting hard exudates are 81.75% and 99.99%, respectively.
Diabetics is a severe disease which affects the different parts of the human body. Due to variations of sugar level the blood vessels present in the retina get damaged. This disease is known as Diabetic Retinopathy (DR). Early identification of diabetic retinopathy will prevent vision loss. Exudates, Haemorrhages, Micro aneurysms are the primary symptoms of Diabetic Retinopathy. To diagnose the abnormalities, the ophthalmologists use fundus images which are captured after applyingdrugs inside the patient’s eye. These drugsmay causeinconvenience to the patient’s eyes. So we developed an automatedsystem to detect retinal lesions with help of non-dilated retinal fundus images. In our proposed system pre-processed images are segmented by new technology called region growing method. In this approach after pre-processing step, certain features are extracted by using Gabor Wavelet method. Adaptive boosting classifier is proposed to investigate the abnormality level of the disease. This result is compared with various segmentation techniques. This method achieved sensitivity of 100% and specificity of 98.8% with accuracy of 98.4%.
The presence of exudates on the retina is the most characteristic symptom of diabetic retinopathy. As exudates are among early clinical signs of DR, their detection would be an essential asset to the mass screening task and serve as an important step towards automatic grading and monitoring of the disease. Reliable identification and classification of exudates are of inherent interest in an automated diabetic retinopathy screening system. Here we review the numerous early studies that used for automatic exudates detection with the aim of providing decision support in addition to reducing the workload of an ophthalmologist.
This paper addresses a novel segmentation algorithm for detecting one of the diabetic retinopathy pathologies, called “exudates”. Exudates segmentation is ordinarily examined from retinal fundus images by various image processing techniques. Instead of carefully picking up the specific exudates features on the retinal images as has been done by the other works, our scheme is to observe global information of the retinal images. The global information, as well as spatial information, is extracted by maximum entropy-based thresholding. The proposed algorithm determines a reasonable threshold value for separating exudates areas, which are usually sparse and brighter, from the rest of images. This approach also ensures and minimizes the illumination variance effects of different images since it takes into account the global information. In addition to the proposed algorithm, luminance channel of the retinal images is exploited for pre-processing stage. After the optical disc which has similar characteristic to the exudates is separated, the pathological areas are subsequently acquired. Evaluations on the E-OPHTHA-EX retinal fundus images database show the advantages of the proposed approach, with the accuracy 99.4 percent, specificity 99.6 percent, and sensitivity 16.9 percent.
Blindness is one of the most serious complications among diabetic patients due to diabetic retinopathy. Primary cause of diabetic retinopathy is abnormality of retinal blood vessels which becomes fragile and often raptured, due to which lipid deposits gets accumulated in the intra retinal space and are visible as yellowish colour in varying shapes sizes and locations. This paper presents a method which uses 2D Otsu thresholding to detect the presence of exudates in digital retinal fundus image. The proposed method uses various techniques, like Hough transform, image thresholding etc. and exhibits a sensitivity of 97.60%, specificity of 98.15% and accuracy of 97.85%.
WHO predicts that in year 2012 there are about 347 million people worldwide have diabetes, more than 80% of diabetes deaths occur in different countries. WHO projects that diabetes will be the 7th major cause leading death in 2030. Diabetic Retinopathy caused by leakage of blood or fluid from the retinal blood vessels and it will damage the retina. Non-proliferative diabetic retinopathy (NPDR) is an early stage of diabetic retinopathy and it is categorized into three stages they are mild, moderate and sever NPDR. The characteristic of the Mild; is specified by the presence of minimum microaneurysm, Moderate; specifies the presence of hemorrhages, microaneurysms, and hard exudates where as Severe; determine on the blockage of vessels, depriving several areas of the retina. With their blood supply. These areas of the retina send signals to the body to grow new blood vessels for nourishment. The proposed algorithm tested on online databases like STARE, DRIVE, DiarectDB0, DiarectDB1 and SASWADE (the database collected during the research work). The statistical techniques were applied on NPDR lesion and calculate the mean, variance, standard deviation, & correlation for classification. K-means clustering have been applied on the dataset with extracted features 95% of correct classification have been achieved.
The World Diabetes Foundation has predicted that more than 439 million people in 2030 will suffer from diabetes. Long-term diabetics can lead to the damage of retinal blood vessels, known as diabetic retinopathy, the leading cause of blindness in developing countries. One of the clinical features of diabetic retinopathy is exudate. Exudates have similar characteristic with optic disc. Therefore, in this research work, removal of optic disc is conducted to reduce false positive of exudates detection. The optic disc detection is done by finding the small area of the optic disc which is enlarged to obtain its total area. Green channel that contains useful information for exudates detection is filtered based on high pass filter. Afterwards, segmentation of exudates is conducted by using thresholding and morphological operations. Final result of exudates is validated with ground truth images by measuring accuracy, sensitivity and specificity. The results show that proposed approach for exudates detection achieves accuracy, sensitivity and specificity of 99.99%, 90.15% and 99.99%, respectively. This result indicates that the proposed method successfully detects exudates and is useful to assist the ophthalmologists in analysing retinal fundus image especially for exudates detection to diagnose diabetic retinopathy.
Diabetic retinopathy is a chronic progressive eye disease associated to a group of eye problems as a complication of diabetes. This disease may cause severe vision loss or even blindness. Specialists analyze fundus images in order to diagnostic it and to give specific treatments. Fundus images are photographs taken of the retina using a retinal camera, this is a noninvasive medical procedure that provides a way to analyze the retina in patients with diabetes. The correct classification of these images depends on the ability and experience of specialists, and also the quality of the images. In this paper we present a method for diabetic retinopathy detection. This method is divided into two stages: in the first one, we have used local binary patterns (LBP) to extract local features, while in the second stage, we have applied artificial neural networks, random forest and support vector machines for the detection task. Preliminary results show that random forest was the best classifier with 97.46% of accuracy, using a data set of 71 images.
A fundus camera provides digitised data in the form of a fundus image that can be effectively used for the computerised automated detection of diabetic retinopathy. A completely automated screening system for the disease can largely reduces the burden of the specialist and saves cost. Noise and other disturbances that occur during image acquisition may lead to false detection of the disease and this is overcome by various image processing techniques. Following this different features are extracted which serves as the guideline to identify and grade the severity of the disease. Based on the extracted features classification of the retinal image as normal or abnormal is brought about. In literature various techniques for feature extraction and different types of classifiers have been used to improve sensitivity and specificity. FROC analysis and confusion matrix are used to evaluate the system performance. In this paper critical analysis of various algorithms and classifiers is done that have been used for the automated diagnosis.
Exudates and drusens detection and measurement from the retina background makes a significant impact on the diagnosis of retinal pathologies. These diseases usually appear as cotton wall spots, yellowish exudates and drusens (macula degeneration). Information about illness severity can be inferred by the measurement of the sizes of the exudates and drusens and comparing them to the retina background size. In this paper, we have proposed robust algorithm for automatic exudates and drusens detection, segmentation, and measurement on 2D retinal images. The applied methods we suggest for exudates and drusens measuring are mathematic (labeling function) and numerical methods. Numerical methods have more sophisticated calculation steps and can be used to approximate more complicated area of exudates using a poly Area function. In our algorithms, prior to measuring exudates and drusens (AMD), a preprocessing takes place, in which first exudates and drusens detection and segmentation were implemented. For these implemented processes, we applied preprocessing operations, including image filtration, correction of non-uniform illumination, and color contrast enhancement, and then the combined approaches for image segmentation and classification were implemented using: two methods of texture, an adaptive threshold, and morphological operators. Moreover, we introduce methods to eliminate the optic disc completely for exudates detection and measuring. After applying these approaches to a number of images provided from ophthalmologists as well as Drive database, this automated diagnostic algorithm resulted in more accurate yields of exudates and Drusens detection and measurements especially for low intensity and less color contrast images from non-dilated eye pupils. This automated algorithm helps ophthalmologists monitor the progression of diabetic retinopathy.
This paper presents experimental results of a method for image-based classification of diabetic retinopathy. In our study we categorize the disease into two classes: diabetic retinopathy non-proliferative and diabetic retinopathy proliferative. The method reduces the dimensionality of the images and finds features using the statistical method of principal component analysis (PCA). Then, it classifies the images using decision trees, the naive Bayes classifier, neural networks, k-nearest neighbors and support vector machines. Preliminary results show that the naive Bayes classifier obtains the best results with 73.4% of accuracy, and 68.4% for F-measure, using a data set of 151 images and testing with different resolutions.
In this paper we present experimental results of an automated method for image-based classification of diabetic retinopathy. The method is divided into three stages: image processing, feature extraction and image classification. In the first stage we have used two image processing techniques in order to enhance their features. Then, the second stage reduces the dimensionality of the images and finds features using the statistical method of principal component analysis. Finally, in the third stage the images are classified using machine learning algorithms, particularly, the naive Bayes classifier, neural networks, k-nearest neighbors and support vector machines. In our experimental study we classify two types of retinopathy: non-proliferative and proliferative. Preliminary results show that k-nearest neighbors obtained the best result with 68.7% using f-measure as metric, for a data set of 151 images with different resolutions.
This paper proposes robust methods for segmentation of clinically significant features of fundus images, from the point of view of detection and gradation of Diabetic Retinopathy as well as Maculopathy. After pre-processing to remove intra- and inter- image illumination variances, the optic disk and fovea are detected and exudates, haemorrhages and microaneurysms are segmented out using modified morphological techniques. The retina is then divided into circular zones concentric around fovea and the pattern and extent of abnormalities in these zones is used to classify the abnormal images into different grades of non-proliferative diabetic retinopathy (NPDR) and maculopathy.
The tool introduced in this paper allows to automatically decide
in an image or in a video sequence which regions are important and which
ones are not. For this purpose, fuzzy logic has been used to modelize
human subjective knowledge about the way to allocate priorities to
regions. The resulting classification can be used in a wide range of
applications going from image coding to image understanding
With more and more audio-visual information being available over the Web, the problem of multi-media indexing, i.e. efficient ways of searching and locating multimedia data in response to queries, has become increasingly important. Efficient content-based access, however, needs a good description and representation of multimedia information. In this paper, we present descriptors for color and shape information in images that we proposed for MPEG-7. In particular, we describe two ways of capturing color content in images called Color Histogram and Region Color. These descriptors are suitable for a wide variety of applications requiring image-to-image matching and object-to-image matching. We also present descriptors for capturing shape information on objects and within images called affine structure and location hash tree. These shape descriptors enable a pose-invariant localization and recognition of 2d patterns. The utility of color and shape descriptors is demonstrated through experiments on the MPEG-7 dataset.
To evaluate a quantitative system to measure the early lesions of diabetic retinopathy seen in stereoscopic fundus photographs.
Using a quantitative classification system, photographs of 4657 eyes (7 stereo pairs of 35-mm slides per eye) were scored for 16 diabetic lesions. A single severity level (identical to the ETDRS Interim Scale) was calculated for each eye. The reliability of this technique, and its reproducibility by independent examiners, was evaluated for individual lesions and severity levels using percent agreement, kappa, and weighted kappa statistics.
This quantitative technique demonstrated an "almost perfect" agreement (weighted kappa > or = 0.810) on all but one lesion by independent observers. For the severity levels, there was a 95.7% perfect agreement (kappa = 0.9428). The reproducibility of agreement over time was "almost perfect" on all but four lesions; with 88% perfect agreement (kappa = 0.8394) for severity levels.
When used to evaluate the early lesions of diabetic retinopathy, the Vanderbilt Classification System is highly reliable between graders and over time. This system can gather quantitative data and evaluate incremental changes in an accurate, reproducible manner.
To describe whether quantitative assessment of early changes in the morphology of retinopathy lesions can predict development of vision-threatening diabetic maculopathy.
We used a nested case-control study, and we studied 11 type 2 diabetes patients who had developed visual loss secondary to diabetic maculopathy. For each diabetes patient, we also studied three matched control patients who had been followed for a comparable period of time without developing visual loss. Fundus photographs describing the early development of retinopathy were digitized and subjected to a full manual quantitative grading on a computer monitor. Differences in the early development of retinal morphology were compared between the two groups. The outcome parameters were changes in the number and area of haemorrhages and exudates in different regions of the fundus, and the weighted distance of these lesions from the fovea and the optic disc.
In patients who developed visual loss secondary to diabetic maculopathy there was significant early progression in the total area and number of haemorrhages and exudates. The haemorrhages had progressed in all retinal areas except the area around the optic disc and the temporal vascular arcades. The exudates had progressed temporally from the fovea and in the retinal periphery.
The results suggest that a quantitative description of the regional development of early diabetic retinopathy may help in identifying patients who will later develop vision-threatening maculopathy.