Enrico Grisan

Detection of optic disc in retinal images by means of a geometrical model of vessel structure

We present here a new method to identify the position of the optic disc (OD) in retinal fundus images. The method is based on the preliminary detection of the main retinal vessels. All retinal vessels originate from the OD and their path follows a similar directional pattern (parabolic course) in all images. To describe the general direction of retinal vessels at any given position in the image, a geometrical parametric model was proposed, where two of the model parameters are the coordinates of the OD center. Using as experimental data samples of vessel centerline points and corresponding vessel directions, provided by any vessel identification procedure, model parameters were identified by means of a simulated annealing optimization technique. These estimated values provide the coordinates of the center of OD. A Matlab/spl reg/ prototype implementing this method was developed. An evaluation of the proposed procedure was performed using the set of 81 images from the STARE project, containing images from both normal and pathological subjects. The OD position was correctly identified in 79 out of 81 images (98%), even in rather difficult pathological situations.

Luminosity and contrast normalization in retinal images.

Retinal images are routinely acquired and assessed to provide diagnostic evidence for many important diseases, e.g. diabetes or hypertension. Because of the acquisition process, very often these images are non-uniformly illuminated and exhibit local luminosity and contrast variability. This problem may seriously affect the diagnostic process and its outcome, especially if an automatic computer-based procedure is used to derive diagnostic parameters. We propose here a new method to normalize luminosity and contrast in retinal images, both intra- and inter-image. The method is based on the estimation of the luminosity and contrast variability in the background part of the image and the subsequent compensation of this variability in the whole image. The application of this method on 33 fundus images showed an average 19% (max. 45%) reduction of luminosity variability and an average 34% (max. 85%) increment of image contrast, with a remarkable improvement, e.g., over low-pass correction. The proposed image normalization technique will definitely improve automatic fundus images analysis but will also be very useful to eye specialists in their visual examination of retinal images.

A divide et impera strategy for automatic classification of retinal vessels into arteries and veins

The first pathologic alterations of the retina are seen in the vessel network. These modifications affect very differently arteries and veins, and the appearance and entity of the modification differ as the retinopathy becomes milder or more severe. In order to develop an automatic procedure for the diagnosis and grading of retinopathy, it is necessary to be able to discriminate arteries from veins. The problem is complicated by the similarity in the descriptive features of these two structures and by the contrast and luminosity variability of the retina. We developed a new algorithm for classifying the vessels, which exploits the peculiarities of retinal images. By applying a divide et impera approach that partitioned a concentric zone around the optic disc into quadrants, we were able to perform a more robust local classification analysis. The results obtained by the proposed technique were compared with those provided by a manual classification on a validation set of 443 vessels and reached an overall classification error of 12%, which reduces to 7% if only the diagnostically important retinal vessels are considered.

Early origins of adult disease: Low birth weight and vascular remodeling

Cardiovascular diseases (CVD) and diabetes still represent the main cause of mortality and morbidity in the industrialized world. Low birth weight (LBW), caused by intrauterine growth restriction (IUGR), was recently known to be associated with increased rates of CVD and non-insulin dependent diabetes in adult life (Barkers hypothesis). Well-established animal models have shown that environmentally induced IUGR (diet, diabetes, hormone exposure, hypoxia) increases the risk of a variety of diseases later in life with similar phenotypic outcomes in target organs. This suggests that a range of disruptions in fetal and postnatal growth may act through common pathways to regulate the developmental programming and produce a similar adult phenotype. The identification of all involved signaling cascades, underlying the physiopathology of these damages in IUGR fetuses, with their influence on adult health, is still far from satisfactory. The endothelium may be important for long-term remodeling and in the control of elastic properties of the arterial wall. Several clinical and experimental studies showed that IUGR fetuses, neonates, children and adolescents present signs of endothelial dysfunction, valuated by aorta intima media thickness, carotid intima media thickness and stiffness, central pulse wave velocity, brachial artery flow-mediated dilation, laser Doppler skin perfusion and by the measure of arterial blood pressure. In utero identification of high risk fetuses and long-term follow-up are necessary to assess the effects of interventions aimed at preventing pregnancy-induced hypertension, reducing maternal obesity, encouraging a healthy life style and preventing childhood obesity on adult blood pressure and cardiovascular disease in later life.

Automatic Recognition of Corneal Nerve Structures in Images from Confocal Microscopy

PURPOSE To devise a method for automatically tracing corneal nerves in confocal microscopy images. METHODS Images were acquired with a confocal microscope. They were normalized and enhanced in luminosity and contrast. The nerves were recognized by applying a novel tracing algorithm, which includes Gabor filtering to enhance nerve visibility and postprocessing procedures to remove false recognitions and to link sparse segments into continuous structures. A prototype of the algorithm was implemented in commercial software and run on a personal computer. RESULTS A retrospective evaluation of the automatic procedure was performed on a data set containing 90 images, from normal and non-normal subjects. The average percentage of correctly recognized nerves length with respect to total manually traced lengths of visible nerves was 80.4% in normal subjects and 83.8% on non-normal subjects; the average rate of false nerve length recognition (with respect to the total automatically traced length) was 6.5% in normal subjects and 9.1% in non-normal subjects. Correlation coefficients between manual and automatic lengths on the same image were 0.94, 0.95, and 0.86 in all, normal, and non-normal subjects, respectively. A further evaluation was performed on an independent set of 80 normal subject images, resulting in a correlation coefficient of 0.89 between manual and automatic nerve lengths. CONCLUSIONS Automatic and manual length estimations on the same image were very well correlated, indicating that the automatic procedure is capable of correctly reproducing the differences in nerve length between different subjects.

A novel method for the automatic evaluation of retinal vessel tortuosity

Tortuosity is among the first alterations in retinal vessel network to appear in many retinopathies. Automatic evaluation of retinal vessel tortuosity is thus a valuable tool for early detection of vascular suffering. Quite a few techniques for tortuosity measurement and classification have been proposed, but they do not always match the clinical concept of tortuosity. This justifies the need for a new definition, able to express in mathematical terms the tortuosity as perceived by ophthalmologists. We propose here a new algorithm for the evaluation of tortuosity in vessels extracted from digital fundus images. It is based on the partitioning of each vessel in segments of constant-sign curvature and on the combination between the number of such segments and their curvature values. This algorithm has been compared with the other tortuosity measures on a set of 20 vessels from 10 different images. These vessels had been preliminarily ordered by an expert ophthalmologist in order of increasing perceived tortuosity. The proposed algorithm proved to be the best one as regards arterial tortuosity and among the best for vein tortuosity evaluation.

Textureless Macula Swelling Detection With Multiple Retinal Fundus Images

Retinal fundus images acquired with nonmydriatic digital fundus cameras are versatile tools for the diagnosis of various retinal diseases. Because of the ease of use of newer camera models and their relatively low cost, these cameras can be employed by operators with limited training for telemedicine or point-of-care (PoC) applications. We propose a novel technique that uses uncalibrated multiple-view fundus images to analyze the swelling of the macula. This innovation enables the detection and quantitative measurement of swollen areas by remote ophthalmologists. This capability is not available with a single image and prone to error with stereo fundus cameras. We also present automatic algorithms to measure features from the reconstructed image, which are useful in PoC automated diagnosis of early macular edema, e.g., before the appearance of exudation. The technique presented is divided into three parts: first, a preprocessing technique simultaneously enhances the dark microstructures of the macula and equalizes the image; second, all available views are registered using nonmorphological sparse features; finally, a dense pyramidal optical flow is calculated for all the images and statistically combined to build a naive height map of the macula. Results are presented on three sets of synthetic images and two sets of real-world images. These preliminary tests show the ability to infer a minimum swelling of 300 and to correlate the reconstruction with the swollen location.

Automatic Segmentation and Disentangling of Chromosomes in Q-Band Prometaphase Images

Karyotype analysis is a widespread procedure in cytogenetics to assess the possible presence of genetics defects. The procedure is lengthy and repetitive, so that an automatic analysis would greatly help the cytogeneticist routine work. Still, automatic segmentation and full disentangling of chromosomes are open issues. We propose an automatic procedure to obtain the separated chromosomes, which are then ready for a subsequent classification step. The segmentation is carried out by means of a space-variant thresholding scheme, which proved to be successful even in presence of hyper- or hypofluorescent regions in the image. Then, the tree of choices to resolve touching and overlapping chromosomes is recursively explored, choosing the best combination of cuts and overlaps based on geometric evidence and image information. We show the effectiveness of the proposed method on routine data acquired with different microscope-camera setup at different laboratories: from 162 images of 117 cells totaling 6683 chromosomes, 94% of the chromosomes were correctly segmented, solving 90% of the overlaps and 90% of the touchings. In order to provide the scientific community with a public dataset, the data used in this paper are available for public download.

Segmentation of candidate dark lesions in fundus images based on local thresholding and pixel density

Due to its blood microcirculation, the retina is one of the first organs affected by hypertension and diabetes: retinal damages can lead to serious visual loss, that can be avoided by an early diagnosis. The most distinctive sign of diabetic retinopathy or severe hypertensive retinopathy are dark lesions such as haemorrhages and microaneurysms (HM), and bright lesions such as hard exudates (HE) and cotton wool spots (CWS). Automatic detection of their presence in the retina is thus of paramount importance for assessing the presence of retinopathy, and therefore relieve the burden of image examination by retinal experts. The most widespread scheme for automatically detect retinal lesion rely on a initial segmentation and a subsequent refinement stage, usually by means of a supervised classification or based on heuristic rules. The first step is therefore required not to lose any possible lesions, at the same time discarding as much of the normal retina as possible. In this work we propose a simple and effective method to detect and identify haemorrhagic (dark) lesions in retinal images, by using a simple local thresholding followed by an evaluation of a measure of the spatial density of the pixels selected at the first step. We evaluate the algorithm on 6 images presenting dark lesions extracted from a database of 60 annotated images, resulting in a mean detection rate of 94% the lesions present in an image, with good performance in term of false candidate rejection.

An improved system for the automatic estimation of the Arteriolar-to-Venular diameter Ratio (AVR) in retinal images

The Arteriolar-to-Venular diameter Ratio (AVR), a parameter derived from vessel caliber measurements in a specific region of retinal images, is used as a descriptor of generalized arteriolar narrowing, an eye fundus sign often seen in patients affected by hypertensive or diabetic retinopathies. We developed an improved system to compute AVR in a totally automatic way. Images are at first enhanced to highlight the vessel network, which is then traced by a vessel tracking algorithm. From the detected vessel structure, the position of the optic disc is derived and the region inside which the AVR data are to be measured is determined. Vessels within this region are classified as either arteries or veins, their caliber is estimated and the AVR parameter is eventually computed. Improvements with respect to the previous version are related to post-processing algorithms to enhance vessel tracking and a totally new artery/vein discrimination technique. Results provided by the new system have been compared with manually derived AVR values on 20 eye fundus images, resulting in a final correlation coefficient of 0.88.