Abstract—We present a method for automated segmentation of the vasculature in retinal images. The method produces segmentations by classifying each image pixel as vessel or nonvessel, based on the pixel’s feature vector. Feature vectors are composed of the pixel’s intensity and two-dimensional Gabor wavelet transform responses taken at multiple scales. The Gabor wavelet is capable of tuning to specific frequencies, thus allowing noise filtering and vessel enhancement in a single step. We use a Bayesian classifier with class-conditional probability density functions (likelihoods) described as Gaussian mixtures, yielding a fast classification, while being able to model complex decision surfaces. The probability distributions are estimated based on a training set of labeled pixels obtained from manual segmentations. The method’s performance is evaluated on publicly available DRIVE (Staal et al., 2004) and STARE (Hoover et al., 2000) databases of manually labeled images. On the DRIVE database, it achieves an area under the receiver operating characteristic curve of 0.9614, being slightly superior than that presented by state-of-the-art approaches. We are making our implementation available as open source MATLAB scripts for researchers interested in implementation details, evaluation, or development of methods. Index Terms—Fundus, Gabor, pattern classification, retina, vessel segmentation, wavelet.

Motivated by the goals of improving detection of low-contrast and narrow vessels and eliminating false detections at non-vascular structures, a new technique is presented for extracting vessels in retinal images. The core of the technique is a new likelihood ratio test that combines matchedfilter responses, confidence measures and vessel boundary measures. Matched filter responses are derived in scale-space to extract vessels of widely varying widths. A vessel confidence measure is defined as a projection of a vector formed from a normalized pixel neighborhood onto a normalized ideal vessel profile. Vessel boundary measures and associated confidences are computed at potential vessel boundaries. Combined, these responses form a 6-dimensional measurement vector at each pixel. A training technique is used to develop a mapping of this vector to a likelihood ratio that measures the "vesselness" at each pixel. Results comparing this vesselness measure to matched filters alone and to measures based on the Hessian of intensities show substantial improvements both qualitatively and quantitatively. The Hessian can be used in place of the matched filter to obtain similar but less-substantial improvements or to steer the matched filter by preselecting kernel orientations. Finally, the new vesselness likelihood ratio is embedded into a vessel tracing framework, resulting in an e#cient and e#ective vessel centerline extraction algorithm.

Motivated by the goals of improving detection of low-contrast and narrow vessels and eliminating false detections at non-vascular structures, a new technique is presented for extracting vessels in retinal images. The core of the technique is a new likelihood ratio test that combines matched filter responses, confidence measures and vessel boundary measures. Matched filter responses are derived in scale-space to extract vessels of widely varying widths. A vessel confidence measure is defined as a projection of a vector formed from a normalized pixel neighborhood onto a normalized ideal vessel profile. Vessel boundary measures and associated confidences are computed at potential vessel boundaries. Combined, these responses form a 6-dimensional measurement vector at each pixel. A learning technique is applied to map this vector to a likelihood ratio that measures the "vesselness" at each pixel. Results comparing this vesselness measure to matched filters alone and to measures based on the intensity Hessian show substantial improvements both qualitatively and quantitatively. When the Hessian is used in place of the matched filter, similar but less-substantial improvements are obtained. Finally, the new vesselness likelihood ratio is embedded into a vessel tracing framework, resulting in an e#cient and e#ective vessel extraction algorithm.

Retinopathy of prematurity is a disease that a#ects the eyes of many babies who are prematurely born. If the retinopathy is not detected in the days following birth, blindness may occur. Studies have demonstrated that by observing the blood vessels within the retina, the disease can be quantified at an early stage, and early treatment can save the baby's eyes. We have developed a new tool to assess retinopathy of prematurity. Our technique captures the image of the retina to extract and quantify both tortuosity and dilation of blood vessels. Our approach demonstrates a 80% sensitivity and 92% specificity in the prediction of retinopathy compared to experts, shows significantly reduced diagnosis time, and features clinical integration via speech recognition and glare detection.

We present a method for automated segmentation of the vasculature in retinal images. The method produces segmentations by classifying each image pixel as vessel or nonvessel, based on the pixel's feature vector. Feature vectors are composed of the pixel's intensity and continuous twodimensional Morlet wavelet transform responses taken at multiple scales. The Morlet wavelet is capable of tuning to specific frequencies, thus allowing noise filtering and vessel enhancement in a single step. We use a Bayesian classifier with class-conditional probability density functions (likelihoods) described as gaussian mixtures, yielding a fast classification, while being able to model complex decision surfaces. The probability distributions are estimated based on a training set of labeled pixels obtained from manual segmentations. The method's performance is evaluated on publicly available DRIVE [34] and STARE [16] databases of manually labeled non-mydriatic images. On the DRIVE database, it achieves an area under the receiver operating characteristic (ROC) curve of 0.9598 and an accuracy of 0.9467 versus 0.9473 for a second human observer.

This report presents a method of detecting branching structure, such as blood vessels from retinal images, using a Gaussian Intensity model. Features are modelled with a Gaussian function parameterised by position, orientation and variance within some spatial window. Multiple features are modelled using a superposition of Gaussian models. A non-parametric classier (k-means) is used to cluster components corresponding to each feature. Two dierent groups of images are used to test the methodology: articial images and images of the human retina. i Contents 1

Surgical planning of liver tumor resections requires detailed three-dimensional (3D) understanding of the complex arrangement of vasculature, liver segments and tumors. Knowledge about location and sizes of liver segments is important for choosing an optimal surgical resection approach and predicting postoperative residual liver capacity. The aim of this work is to facilitate such surgical planning process by developing a robust method for portal vein tree segmentation. The work also investigates the impact of vessel segmentation on the approximation of liver segment volumes. For segment approximation, smaller portal vein branches are of importance. Small branches, however, are difficult to segment due to noise and partial volume effects. Our vessel segmentation is based on the original gray-values and on the result of a vessel enhancement filter. Validation of the developed portal vein segmentation method in computer generated phantoms shows that, compared to a conventional approach, more vessel branches can be segmented. Experiments with in vivo acquired liver CT data sets confirmed this result. The outcome of a Nearest Neighbor liver segment approximation method applied to phantom data demonstrates, that the proposed vessel segmentation approach translates into a more accurate segment partitioning.

This paper presents a method of personal authentication process using digital retinal image matching. The process composed of four modules: reference point’s detection, blood vessel segmentation and derivation of corresponding binary image skeleton of one pixel width, feature points extraction and finally matching similarities among these feature points of different images. The Fovea center and the Optic disc are used as reference points for compensating the unwanted rotational and translational effects. The maximum principal curvature of the Hessian matrix of the intensity image is used along with some image filtering to segment the blood vessel structure. Then the skeleton of the binary image and corresponding blood vessel intersection points are extracted using two proposed algorithms. Finally the matching process is done by proximity analysis of the intersection points of different retinal images. The whole process is then tested on several retinal images of different persons and the tested images were classified correctly. Keywords Biometric personal authentication, fovea center detection, optic disc detection, Retina blood vessel skeleton generation, Blood vessel intersection point detection, blood vessel segmentation 1.