Pedro Alemany

Automatic optic cup segmentation algorithm for retinal fundus images based on random forest classifier

Glaucoma is an eye disease that constitutes the second cause of blindness over the world. Although it cannot be cured, its progression may be prevented if it is early detected. Expert ophthalmologists use as a sign of suffering from the disease, the evaluation of the relationship between optic disc and cup areas in retinal fundus images and, therefore, image processing techniques applied to glaucoma has become an emerging research line. This paper presents a novel technique for the detection of the optic cup in retinal fundus images, which may be included in a glaucoma computer aided diagnosis tool. The method, based on a color space related to human perception and adapted to surrounding conditions, JCh from CIECAM 02 (International Commission on Illumination Color Appearance Model), utilizes a random forest classifier to obtain cup edge pixels. As vessels tend to bend in the edge of the cup, the classifier does not consider all the pixels in the image. In fact, only those belonging to vessels and possessing the highest curvature among their neighbors are taken into account. Another prior knowledge used in the proposed method is the fact that cup area usually posses a bright yellow color. Therefore the feature vector serving as an input for the classifier is made with the curvature, the color of the candidate pixel and its location relative to the OD center. Finally, a basic post processing is performed to join the selected pixels with a smooth curve. The method has been tested in a publicly available database, GlaucomaRepo, from where we used 35 images for training and 55 for test. Five numerical parameters were calculated and a comparison against three color spaces was performed. The results obtained indicate the effectiveness of the approach.

Automatic cup-to-disc ratio estimation using active contours and color clustering in fundus images for glaucoma diagnosis

In this paper we propose a new automatic technique for the segmentation of the Optic Disc (OD) and optic nerve head (cup) regions in retinographies for glaucoma diagnosis. It provides an estimation of the Cup-to-Disc Ratio, the main clinical indicator of the disease. OD is detected combining intensity-based, multi-tolerance and morphological methods along with the active contour technique. Cup region is obtained with a new human perception adapted version of the well-known K-means algorithm in the uniform CIE L*a*b* color space with CIE94 color difference. For comparisons, the accurate cup border obtained is rounded and soften with two different techniques: ellipse fitting and mathematical morphology along with Gaussian Smoothing. The proposed method with both rounding steps has been tested in a database of 55 images and compared with the ground truth provided by an expert ophthalmologist. Both, OD and cup region, were satisfactory localized, achieving a mean error of 0.14 for ellipse fitting and 0.13 for morphology. The algorithm proposed seems to be a robust and reliable tool worthy to be included in any CAD system for glaucoma screening programs.

Automatic detection of optic disc from retinal fundus images using dynamic programming

Automatic detection of optic disc (OD) in fundus images is used to determine potential clinical parameters for diagnosis of retinopathic diseases. Due to the presence of vascular-tree blood vessels, the detection of OD area is a complicated task for a computer-aided diagnosis (CAD) system, desirable by ophthalmologists. In this paper, a novel system for the detection of OD area has been developed. It consists of four major steps: preprocessing with a color space transformation and contrast normalization; segmentation of the vascular-tree through radial projection (RP) and weighted-derivative of Gaussian (WDOG) techniques; feature preserving removal of the detected vessels by a fast marching inpainting algorithm, detection of candidate to OD pixels via dynamic programming and OD area location using ellipse fitting methods. The proposed technique has been tested on 129 retinal images from to public and widely used datasets, DRIVE and DIARETB1. Experiments on this dataset indicate that this algorithm is computationally fast and able to achieve 92.5 % of accuracy for OD detection.

Automatic recognition of severity level for diagnosis of diabetic retinopathy using deep visual features

Diabetic retinopathy (DR) is leading cause of blindness among diabetic patients. Recognition of severity level is required by ophthalmologists to early detect and diagnose the DR. However, it is a challenging task for both medical experts and computer-aided diagnosis systems due to requiring extensive domain expert knowledge. In this article, a novel automatic recognition system for the five severity level of diabetic retinopathy (SLDR) is developed without performing any pre- and post-processing steps on retinal fundus images through learning of deep visual features (DVFs). These DVF features are extracted from each image by using color dense in scale-invariant and gradient location-orientation histogram techniques. To learn these DVF features, a semi-supervised multilayer deep-learning algorithm is utilized along with a new compressed layer and fine-tuning steps. This SLDR system was evaluated and compared with state-of-the-art techniques using the measures of sensitivity (SE), specificity (SP) and area under the receiving operating curves (AUC). On 750 fundus images (150 per category), the SE of 92.18%, SP of 94.50% and AUC of 0.924 values were obtained on average. These results demonstrate that the SLDR system is appropriate for early detection of DR and provide an effective treatment for prediction type of diabetes.

Detección automática de microaneurismas en retinografías

PURPOSE We present the development of a tool for the automatic detection of microaneurysms and its clinical evaluation. The intention of this tool is to facilitate the diagnosis of diabetic retinopathy in general screening programs. METHOD The designed and developed tool consists of three stages of processing: 1) Obtaining of the basic image of eye with the retinal camera, inverted image on the green channel, and a high-pass filter of the image. This phase enhances the microaneurysms. 2) Detection of the candidates for microaneurysms, by means of an adaptive prediction filter and regions growth. 3) Selection, among the candidates, of whom microaneurysms must be considered to fulfil the criteria of circular shape, high intensity in the inverted green channel and contrasts with respect to the surrounding pixels. RESULTS We selected to 20 retinal photographs of good quality and dimensions 600x600 pixels from patients with nonproliferative diabetic retinopathy. The ophthalmologists detected 297 microaneurysms in these images. The tool for automatic detection correctly located 252 microaneurysms, with a mean sensitivity of 89% and a false positives rate of 93%. CONCLUSIONS The results obtained seem to indicate that the tool developed will be very useful for its potential use in screening programs in primary care centres. On the other hand, more work is needed on the algorithm to decrease the rate of false positives.PURPOSE We present the development of a tool for the automatic detection of microaneurysms and its clinical evaluation. The intention of this tool is to facilitate the diagnosis of diabetic retinopathy in general screening programs. METHOD The designed and developed tool consists of three stages of processing: 1) Obtaining of the basic image of eye with the retinal camera, inverted image on the green channel, and a high-pass filter of the image. This phase enhances the microaneurysms. 2) Detection of the candidates for microaneurysms, by means of an adaptive prediction filter and regions growth. 3) Selection, among the candidates, of whom microaneurysms must be considered to fulfil the criteria of circular shape, high intensity in the inverted green channel and contrasts with respect to the surrounding pixels. RESULTS We selected to 20 retinal photographs of good quality and dimensions 600x600 pixels from patients with nonproliferative diabetic retinopathy. The ophthalmologists detected 297 microaneurysms in these images. The tool for automatic detection correctly located 252 microaneurysms, with a mean sensitivity of 89% and a false positives rate of 93%. CONCLUSIONS The results obtained seem to indicate that the tool developed will be very useful for its potential use in screening programs in primary care centres. On the other hand, more work is needed on the algorithm to decrease the rate of false positives.

Automatic Tool for Optic Disc and Cup Detection on Retinal Fundus Images

The aging of the population is a matter of concern due to its association with various diseases in humans that limit their quality of life. Among them, glaucoma is one of the leading causes of blindness in the world. To its early diagnose, retinal fundus images are visually inspected by experts. In recent years, image-based computer aided diagnosis systems have been proposed. Automatic segmentation of Optic Disc (OD) and cup areas are their first and most difficult tasks. In this paper, a computerized technique aimed to their extraction from the original images is presented. The tool is related to human perception due to the use of an advanced color metric, CIE94 within a uniform color space, CIE L*a*b* to compute pixels’ color gradients [1]. Based on this information, a classifier assigns a probability value to each of the pixels, meaning its suitability for being part of the Optic Disc and Cup border. The tool has been tested on 200 images from different public databases achieving an accuracy value of 96.63%. This quality level makes the proposed color-based image processing system capable to assist the physicians in glaucoma screening programs.

Automatic CDR Estimation for Early Glaucoma Diagnosis

Glaucoma is a degenerative disease that constitutes the second cause of blindness in developed countries. Although it cannot be cured, its progression can be prevented through early diagnosis. In this paper, we propose a new algorithm for automatic glaucoma diagnosis based on retinal colour images. We focus on capturing the inherent colour changes of optic disc (OD) and cup borders by computing several colour derivatives in CIE L∗a∗b∗ colour space with CIE94 colour distance. In addition, we consider spatial information retaining these colour derivatives and the original CIE L∗a∗b∗ values of the pixel and adding other characteristics such as its distance to the OD centre. The proposed strategy is robust due to a simple structure that does not need neither initial segmentation nor removal of the vascular tree or detection of vessel bends. The method has been extensively validated with two datasets (one public and one private), each one comprising 60 images of high variability of appearances. Achieved class-wise-averaged accuracy of 95.02% and 81.19% demonstrates that this automated approach could support physicians in the diagnosis of glaucoma in its early stage, and therefore, it could be seen as an opportunity for developing low-cost solutions for mass screening programs.