Abstract: The manual analysis of the karyogram is a complex and time-consuming operation, as it requires meticulous attention to details and well-trained personnel. Routine Q-band laboratory images show chromosomes that are randomly rotated, blurred or corrupted by overlapping and dye stains. We address here the problem of robust automatic classification, which is still an open issue. The proposed method starts with an improved estimation of the chromosome medial axis, along which an established set of features is then extracted. The following novel polarization stage estimates the chromosome orientation and makes this feature set independent on the reading direction along the axis. Feature rescaling and normalizing techniques take full advantage of the results of the polarization step, reducing the intra-class and increasing the inter-class variances. After a standard neural network based classification, a novel class reassignment algorithm is employed to maximize the probability of correct classification, by exploiting the constrained composition of the human karyotype.
An average 94% of correct classification was achieved by the proposed method on 5474 chromosomes, whose images were acquired during laboratory routine and comprise karyotypes belonging to slightly different prometaphase stages. In order to provide the scientific community with a public dataset, all the data we used are publicly available for download.
Abstract: A novel system for the vascular tree identification and the quantitative estimation of arteriolar venular ratio clinical index in retinal fundus images is presented. The system is composed of a module for automatic vascular tracking, an interactive editing interface to correct errors and set the required parameters of analysis, and a module for the computation of clinical indexes. The system was organized as a client-server structure to allow clinicians and researchers from all over the world to work remotely. The system was evaluated by three graders analyzing 30 fundus images. The evaluation of the Pearson's correlation coefficient and p-value of a paired t-test for each pair of graders demonstrates the high reproducibility of the measures provided by the system.
Abstract: 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.
Abstract: The analysis of blood vessels in images of retinal fundus is an important non-invasive procedure that allows early diagnosis and the effective monitoring of therapies in retinopathy. In order to derive a quantitative evaluation of the clinical features, such as vessel diameter and tortuosity, an accurate segmentation of the vessel network has to be performed. A new system for the automatic extraction of the vascular structure in retinal images is proposed. It is based on a sparse tracking technique via a multi-directional graph search approach. We consider the image as a weighted unoriented graph with arches connecting adjacent pixels and assume that vessels are minimum cost paths connecting remote nodes. An initial seed-finding algorithm based on fast 1- dimensional multi-scale matched filters is run over a regular grid. Simultaneous best-first search graph explorations start from each seed: when two search frontiers meet, the computed shortest path is recorded and exploited for a new search starting from it. New paths are found by iterating the procedure, until the entire vessel network is reconstructed. Lastly, in order to cover the unexplored region with lowcontrast vessels, a custom fixing procedure is run. 20 images have been used to test the algorithm, comparing the results with ground-truth manual segmentation. The method provides an average sensitivity of 96.2%.
Abstract: Ocular fundus images provide important information about retinal degeneration, which may be related to acute pathologies or to early signs of systemic diseases. An automatic and quantitative assessment of vessel morphological features, such as diameters and tortuosity, can improve clinical diagnosis and evaluation of retinopathy. We propose a new method to accurately evaluate vessel diameters and centerline starting from an estimated network of vessel axes. The algorithm extracts points laying on the vessel borders by means of an efficient mono-dimensional matched filtering approach. The orientation of the filter kernel is chosen according to the information provided by the network and the appropriate scale is computed by means of an initial diameter estimation performed on the vessels cross section profiles before the filtering process. An adaptive correction step is then run to fix non consistent diameters, in order to obtain a regular and continuous vessel morphology. Vessel border refinement finally yields an accurate representation of the vascular structure. Average calibers were evaluated, for a set of 739 vessel segments, both manually by an expert and automatically by the proposed method and results show high correlation (Ï = 0.97).
Abstract: 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. The first step in every automatic procedure, is the segmentation of the chromosomes, as either single entities or in clusters, in the image. The better the segmentation step, the easier the subsequent disentanglement. We propose for the segmentation step a region based level set algorithm that is able to address the variability in the image background due to the presence of hyper- or hypo-fluorescent regions in the image. We compare its performance with other algorithms proposed in the literature for the segmentation of chromosomes, over a set of 11 manually annotated images. We show the superiority of the proposed approach both in terms of pixel sensitivity, and in terms of number of separate clusters with respect to the manual segmentation. The images used in the paper are available for public download.
Abstract: The manual analysis of the karyogram is a complex, wearing and time-consuming operation. It requires a very meticulous attention to details and calls for well-trained personnel. Even though existing commercial software packages provide a reasonable support to cytogenetists, they very often require human intervention to correct challenging situations. We developed a robust automatic classification system conceived to cope with routine images in which chromosomes are randomly rotated, possibly blurred or also corrupted by overlapping or by dye stains. It consists in a sequence of modules comprising robust feature extraction based on medial axis, chromosome polarization, feature pre-processing, and Neural Network classification followed by a class reassigning algorithm.We show the effectiveness of the proposed method on data comprising karyotypes belonging to slightly different stage of the prometaphase. This dataset contains 119 karyotypes (5474 chromosomes), 70 of which were used for training and validation and 49 for the final testing. In this latter set of images, the system achieved a classification accuracy, as compared to manual ground truth, of 95.6%.
Abstract: 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 hypo-fluorescent regions in the image. Then a greedy approach is used to identify and resolve touching and overlapping chromosomes, based on geometric evidence and image information. We show the effectiveness of the proposed method on routine data: 90% of the overlaps and 92% of the adjacencies are resolved, resulting in a correct segmentation of 96% of the chromosomes.
Abstract: This thesis deals with the automatic analysis of colour fundus images and with its applications in the evaluation of hypertensive and diabetic retinopathy. In particular, it focuses on the project of an adaptive optics fundus camera for the retinal fundus acquisition and on the software implementation for the analysis and the quantitative evaluation of the retinopathies.
Both hypertension and diabetes affect, although with different time courses, the microcirculation: retinopathy is one of the consequences of such circulation damage. The retina and retinal vessels are very sensitive to changes in the microvascular circulation, and it has been demonstrated that single features of hypertensive retinopathy have a strong prognostic value for stroke, carotid stiffness, and coronary disease.
At the same time, retinopathy is a social burden, with heavy direct and indirect costs, since visual loss reduces the capacity of working and carrying on with an independent life. Even if other organs are affected by microcirculation damage, the retina have the complete advantage over other them of being easily available for non invasive examination, therefore suggesting a cost-effective way of monitoring the progression of the systemic disease associated with the retinopathy. Detection of retinopathy at its onset is the most critical issue to avoid blindness, particularly in diabetic retinopathy which does not recede with treatment, at the present state of pharmacology.
Unfortunately these first stages are almost asymptomatic. It has been demon-strated that a screening program could save most of the population at risk from developing sight-threatening retinopathy. In the western world there are not enough resources, in terms of time and available expert ophthalmologists, for carrying on an extensive screening. Thus, a reliable automatic tool for evaluating retinopathies is strongly needed.
Current fundus camerae are able to compensate defocus aberration, introduced by the eye, by means of mobile system of internal optics, but they cannot correct astigmatism or higher order aberrations. The first chapters of the current thesis will describe the attempt of realization of an acquisition system prototype which is able to compensate the latter category of aberration, thus allowing to obtain retinal fundus images of high quality, even when patients with high aberration are con-cerned. The retinal fundus images are then digitalized with high resolution and processed by the software described in the second part of the thesis. The algorithms that composes this software are designed for geometrically characterize the vessel structures of the retina, aiming at an automatic diagnosis of the pathologies.
The base concept of the acquisition instrument developed in our laboratories is to introduce, along the optical path of a fundus camera, a DM (Deformable Mirror) adaptive optics. The main problem of the conventional fundus camera is that the information through which the DM shape is adapted is provided by a complex and expensive wave-front analyzer. The proposed solution consists in projecting a known pattern on the retinal fundus and then acquiring the resulting image with a dedicated sensor. By analyzing real-time this image with custom algorithms it will be possible to obtain information about how the DM has to be opportunely deformed in order to reduce the aberration introduced by the eye This method will allows the acquisition of high resolution images without distortions even in the case of patients affected by aberration that are not compensable by the currently available fundus camerae. Although the simulation tests on the optic bench have provided encouraging preliminary results, the development of a fundus camera suitable for a clinic/diagnostic employment has encountered technical limitations, mostly regarding the issues of the prototyping an instrument in a not-proofed environment. The incidence of spurious reflexes, introduced in the optical system by the surroundings, on the signal-to-noise ratio will be subject of research in the near future.
Retinal fundus images, collected by gathering both datasets built by us and datasets already used by other teams for research purposes, will be analyzed with custom algorithms, which are able to extract important clinical parameters for the diagnosis of the retinopathies, with particular attention to the hypertensive and the diabetic ones. In the framework of this thesis the features of the vascular apparatus will be taken into consideration: the vessel network identification (veins and arteries) will allow the assessment of its main geometrical characteristics (length, direction, caliber, bifurcation, etc.). From these findings, specific indexes of diagnostic relevance will be computed, providing the clinicians information regarding the patientâs retinopathy degree, whose significance is crucial for the diagnosis of retinal pathologies. The algorithms presented in this thesis allow to conceive a tool to be used both for mass-screening of hypertensive and diabetic retinopathies, and also for monitoring the progression of the diseases. Its usefulness will be threefold. Firstly, it shall provide a diagnostic tool to aid the clinical practice. Secondly, it will provide quantitative details of the retinal vessel network, thus providing a tool for clinical research. Finally, it will allow pharmaceutical research to obtain a quantitative and reproducible assessment of disease evolution during pharmacological treatment.
The proposed software for the identification of retinal vascular structures is currently subject of clinical evaluation at the Department of Ophthalmology and Visual Sciences, University of Wisconsin, USA, whose Fundus Photograph Reading Center has expressly declared its will to acquire the system as the standard for the determination of diagnostic features, namely the CRAE, CRVE, and AVR indexes as indicators of generalized arteriolar narrowing.
The recent collaboration with the Department of Twin Research & Genetic Epidemiology, of the Kingâs College London Division of Genetics and Molecular Medicine, St Thomas' Hospital, UK, has brought to the agreement that more than 3500 retinal fundus images (plus 10 new image per week) provided by them have to be processed and analyzed by our software for the estimation of the clinical indexes.
Both results achieved in the experimentations and the fruitful international collaboration established with relevant clinical and research groups, make us quite confident about the quality of the developed methodologies and the potential success of their employment, with the auspice that future improvements can broaden its usability.
