ISSN 1060992X, Optical Memory and Neural Networks (Information Optics), 2014, Vol. 23, No. 4, pp. 278–286. © Allerton Press, Inc., 2014.

Computer Systems for Geometrical Analysis of Blood Vessels Diagnostic Images N. Yu. Ilyasova Image Processing Systems Institute of the RAS, Samara, Russian Federation; Samara Aerospace University named after academician S.P. Korolyov (SSAU) email: [email protected] Received September 16, 2014; in final form, November 11, 2014

Abstract—In this paper, we give a review of existing systems for computer analysis of geometrical char acteristics of blood vessel diagnostic images. We present a “OphthalmOffice” system for analysis of diagnostic images of eye fundus vessels and a “CardiOffice” system for analysis of coronary vessels. We also present comparative characteristics of their computing efficiencies. Keywords: human vascular system, image, processing DOI: 10.3103/S1060992X14040110

Computer analysis of images became the basic tool of medical diagnostic systems allowing one to increase diagnostics quality significantly. Information technologies are most widely introduced in ophthalmology and cardiology. In the present paper, we analyze images of vascular systems of eye fundus and coronary vessels. In both cases, the blood vessel images carry important diagnostic information. Basing on examination of blood vessels, a doctor not only concludes about the state of the organ, but also identifies general systemic diseases such as insular diabetes, polycythemia, anemia and primary hypertension [1]. The goal of our study is as follows: developing of a methodology allowing one to get from the vessel images important diagnostic information, developing and examination of methods, algorithms and information tech nologies of modeling, processing and analysis of blood vessel images, as well as creation of computer systems for medical purposes on the basis of the obtained results. Such systems have to provide solving of early and dif ferentiated vascular disease detection problems with the aid of quantitation of morphological characteristics. For this purpose, we solved the following problems: 1. We analyzed the uptodate state of the problem of vascular pathologies detection, revealed the main steps of blood vessel images processing and determined informative characteristics for clinical recognition. 2. We worked out a methodology of extracting on the blood vessel images important diagnostic informa tion, which includes mathematical models, methods and computer systems. 3. We developed a mathematical model of blood vessels imaging allowing one to provide a formal descrip tion of geometrical parameters and to construct diagnostic criteria. We also developed methods and algorithms for enhancing accuracy in estimation of geometric parameters. 4. On the basis of quantitation of changes of vascular system states and methods of estimation of the corre sponding diagnostic criteria, we proposed a set of integral indicators of the vascular system state. We analyzed their efficiency and proposed methods of extracting important diagnostic information about the state of an individual vascular system. 5. We developed information technology of analysis of images of eye fundus vessels including methods of generation of the space of important signs for the effective early detection and prophylaxis of development of the diabetic retinopathy in patients with insular diabetes. 6. We proposed information technology of restoration of spatial structure of coronary vessels from a small number of unmatched angiographic projections that was oriented towards estimation of local spatial geomet rical characteristics of vascular system and diagnostic parameters. 7. We worked out guidelines, algorithms and software for diagnostic systems supporting ophthalmologists and cardiologists in decisionmaking process. Review of systems for computer analysis of blood vessel images. The promising research direction is the development of software with different auxiliary algorithms for analysis of images of coronary and retinal ves sels including the vessel separation, determination of the important image elements (like optic disks), the mea 278

COMPUTER SYSTEMS FOR GEOMETRICAL ANALYSIS

279

surements of vessel diameters, vessel tortuosity angles and other diagnostic signs of vascular systems. All these types of software can quicken the study of relations between changes in the anatomy of the human vascular sys tems and different diseases. The SIRIUS system (System for the Integration of Retinal Images Understanding Services) [2] is a com puter system for analysis of retina images on the basis of the web application that provides a possibility of col laborative work for ophthalmologists and other experts in the field. The SIRIUS system consists of web inter face, an application server that gives access to different services and tools for image processing. The system allows exchange of images and provides different computeraided methods for reducing errors when analyzing retina images. The system service module for analysis of vessel microcirculation includes semiautomatic algo rithm for calculation of arteriovenous ratio (AVR). The weakness of this system is in the absence of the possi bility to measure different diagnostic signs of vessel pathological changes when detecting pathologies on the retina images. A longstanding competition known as ROC (Retinopathy Online Challenge) [3] is focused on determina tion of the best algorithms for different aspects of the diabetic retinopathy detecting. It allows one to cooperate when developing methods for creating a highperformance computer system for solving monitoring and diag nostic problems [4]. The project RIVERS (Retinal Image Vessel Extraction and Registration System) [5, 6] also makes efforts in the same direction. The software application VAMPIRE (Vascular Assessment and Mea surement Platform for Images of the Retina) [7] allows one to carry out semiautomatic estimation of retina vessels and their characteristics. This computer system serves for efficient and reliable detecting of the most important retinal elements (optic disc, vasculature) and estimation of the key parameters used frequently in investigative studies: vessel diameter, tortuosity, and vessel branching angles. LiveVessel [8] is semiautomatic software for segmentation of vascular trees in 2D color retina images. A lot of large companies and research centers developed their own computer systems for automatic pro cessing and analysis of medical images of special and generalpurposes. Each of the systems has its own dis tinctive features, merits and demerits. The system “Gamma Multivox” [9] consisting of a hardware/software complex for data inputting, process ing, and storage was developed in Lomonosov Moscow State University. The complex includes automated workstations designed for multimodal operation with 2D/3D medical images of lungs, bronchus cavities, and different anatomical units (teeth, paranasal sinuses, contrasted vessels inside soft tissues and so on), in partic ular, with coronary vessel images. The main part of the workstation is a specialpurpose software that allows a doctor to see images obtained with the aid of medical diagnostic devices, which use different physical registra tion methods. The system “Gamma Multivox” allows one to visualize and to process 2D/3D images depend ing on their modality type. It makes it possible to work with several 2D/3D images and several series of images, as well as synthesis, visualization and processing of 3D vessel images. There are a lot of working modes for visu alization and processing of images. They are the autoview of several series of images, the mapping of section lines; examination of different orthogonal sections of an array of objects (multiplanar reconstruction) in the modes of maximal and minimal rate specifying an arbitrary slice thickness of the object under study; providing of projections with maximal image intensities; isometric image projecting with the given transmittance curve depending on density; generation of different cuts on the image that makes it possible to see the inner structure of a 3D array (trimming, selection of opasificated vessels, and so on); accurate measuring of volumes of seg mented vessels; the possibility to examine the contrasted vessels by means of the erection method with analysis of stenotic areas; the availability of ready superstructures of imagining of three dimensional objects for visual ization of different anatomical units (contrasted vessels inside soft tissues). The hardware/software complex “Gamma Multivox” is also a system for transmitting and archiving of images (Picture Archivingand Commu nication System—PACS) that can be integrated with Medical information system. A shortcoming of the com plex is rather steep requirements (comparing with other analogous systems) to computer facilities and opera tional software and lack of possibility to obtain full geometrical data relating to vessels and morphological char acteristics. The Delante company offers an alternative diagnostic software for analysis of medical images including vessels [10]. They developed two highend applications for diagnostic workstations IQVIEW and Myrian. The IQVIEW program can integrate with any image transmitting and archiving system (PACS). It unites all func tions of a fullscale workstation for twodimensional images reading and a workstation for 3D postprocessing and realizes preprocessing of any data from threedimensional images. The 3D technology they offer provides the possibility to work on the most types of systems with standard graphics adapters and with minimum requirements to system resources. For example, notebook will do. The analogous requirements to system resources characterize the diagnostic software Myrian. One of it modulus is the Myrian® XPVessel applica tion that simultaneously supports two modalities: vessel visualization by computed tomography method and by nuclear magnetic resonance method. This is the most suitable offer of the company for vessel monitoring. OPTICAL MEMORY AND NEURAL NETWORKS (INFORMATION OPTICS)

Vol. 23

No. 4

2014

280

ILYASOVA

It allows one to create automatically multiplane reformatting (MPR) of images, maximum intensity and threedimensional projections (MIP) for quick estimation and planning of complex surgery. The Myrian® XPVessel distinguishing features are automatic extraction of the central line, smoothed presenta tion, measuring of the vessel diameter and length, curved planar reconstruction (CPR), and axial visualization (in transversal sections). Moreover, this application creates a diagram of the relation between the vessel diam eter and its length, as well as shows the vessel crosssections. Instant visualization and measuring of the mini mal and maximal vessel diameters in the viewport of the transversal section are possible. In addition, the appli cation provides an automatic segmentation of vessels as well as calculation of their central lines. All these pos sibilities combined with other advantages of the system make it suitable for vessel images postprocessing. However, it does not allow one to calculate vessel diagnostic signs for revealing their deformations of different kinds and the pathology degree. With regard of high necessity of Russian medical treatment facilities in available angiographic systems the scientificresearch manufacturing cooperative “Electron” [11] offers a device with inbuilt software for carry ing out the most demanded vessel examinations, such as examinations of brain vessels, chest blood vessels, extremity blood vessels, cardiac cavity and other angiography examinations of vessels of all anatomical regions. The diagnostic quality of images is provided by a set of special filters for testing of different types and a com plete set of organoautomatic programs for all anatomical regions. The first Russian domestic digital angio graph has inbuilt software that includes a package for vessel examination. The Russian system is the cheapest one among the other available angiography systems, however as with other analogues systems, it is strictly con nected with its own equipment that is with angiograph. Also it cannot provide diagnostic analysis of the vessel shape on the basis of a full set of geometrical data. From all the present systems the best one is the AlluraXper system [12], produced by Philips company. It allows one to carry out all the range of interventional cardiology and vessel procedures. Large range of vision and flat detector of high resolution go with diagnostic and threedimensional interventional tools, and all thee are all combined with fixed Xray system in one advanced hybrid surgery room. The system provides high fineness and sharpness of visualization of small details and objects during cardiovascular interventions. The Allura 3D system ensures threedimensional visualization of the brain, cerebral, peripheral vasculature and bone structures, performs threedimensional visualization basing on informa tion obtained from a single rotational angiography run of coronary artery minimizing effects of perspective distortions. The inbuilt algorithm Xres eliminates noise in real time and provides image quality enhance ment. In the hybrid angiographic laboratory this system realizes the function of the vessel visualization prior and after the procedure for data analysis. The builtin application StentBoost provides enhancement of stent visualization in coronary artery, and allows the realtime control of the stent turning and position ing during the coronary angiography. The facilities of the AlluraXper system that are compatible with the DICOM 3.0 standard provide suitable federation with other systems and guarantee availability of infor mational images when it is necessary. The optimized federation of the system combined with the large number of builtin options, which allows one to construct correct threedimensional vessel models, to eliminate noise in real time and so on, is one of many advantages of the system developed by Philips. How ever, unfortunately, the system is highly specialized and it costs a lot. One more uptodate computer system of image postprocessing that is not connected tightly with the equipment, was proposed by the Mediso company (Medical Imaging Systems, Budapest, Hungary). Their InterView Fusion product [13] is the system for visualization, processing and analysis of medical vessel images of different origin. It was developed basing on the latest achievements in the field of algorithms and software tools for image processing. The InterView Fusion system allows one to download and superpose images with the aid of specialized viewer and automatic registration algorithms, and to carry out static measurements over the regions and volumes of the object of interest. The system is designed for processing of images of different types. It does not have special tools for analysis of vessel images, but only the general set of the supplied packages. The system General Operator Processor (GOP) from the company ContextVision [14] is the computer system for postprocessing of different medical images of organs, bones, soft tissues, including vessels. It is based on imitation of the human visual system, and uses a hierarchical approach when identifying ele ments of images at different levels of abstraction. The GOP method allows one to identify the examined object structures analyzing each pixel with account of its environment. On the image, the fixed tonal gra dation of grey corresponds to each pixel or each element of the image. GOP provides the possibility to use pixels as lowlevel indicators, such as an edge or a line. In the same time, they can be used as highlevel indicators, showing textures, treatment regions, bounds of objects, separate objects and relations between them. Adaptive filtration and restoration of an intensely noisy image constitute the main field of applica tion of this system. The contrast enhancements, definition of edges, and high noise immunity provide the OPTICAL MEMORY AND NEURAL NETWORKS (INFORMATION OPTICS)

Vol. 23

No. 4

2014

The estimation of tortuosity and branching angles (possible other morphological signs)

Classification of regions of interest, automation of diagnostic stages, analysis of subclinical morphological changes of vessels

Analysis and synthesis of 3D images

Links with the facility

+ – + + + + + + + + +

Diameters estimation (AVR) (the possibilityto create a diagram of diameters along the trace)

– – – + + – + + + + +

281

Segmentation and quantitative estimation of regions of interest, central line extraction

SIRIUS LIVEVESSEL STARE Gamma Multivox Myrian® XPVessel VAMPIRE ELECTRON Allura Xper (Philips) InterView Fusion GOP (ContextVision) SYNGO.VIA.

Tools for preprocessing, marking and measuring of images

Computer systems

Image registration, visualization, accounting and sharing of information

The system ability

COMPUTER SYSTEMS FOR GEOMETRICAL ANALYSIS

– + + + + + – + – – +

+ – – + + + – – – – –

– – – – – + – – – – –

– – + – – – – – – – –

– – – + + – – + – – +

– – – + –+ – + + – – –

Fig. 1. The abilities of computer systems for vessel analysis.

improvement of the quality of image of minor vessel structures. As a result, small details including the smallest vessels are seen more clearly. The system works with standard tools, and it is supported by multi processor systems, so one does not need to bother about technical supply. The processing is completely optimized for special anatomical structures; it also fully integrated into operational process and works automatically. However, it suggests only the ways to improve images of vascular systems. Siemens presents its own postprocessing system SYNGO.VIA. When with the aid of this postpro cessing system one views images of eye fundus vessels, it gathers all possible information about the patient and automatically loads the image into a suitable application preparing it for processing with regard to the disease. For example, it would select the optimal phase for reconstruction; pick out and mark the vessels, and so on. The system also supplies the doctor with the minimal set of tools for analysis necessary in the case of vessel diseases. The system imposes low requirements on the equipment and it integrates with other systems. However, since it is a multipurpose one, it has not sufficient number of tools necessary when processing images of eye fundus vessels, and has no elements for automatic diagnostics of such images. In Fig. 1, we show the general properties and capabilities of the abovementioned systems. Today in many countries in public screening centers, they heavily use quantitative approaches of esti mations of vessel images when revealing vascular pathologies with the aid of automated systems of pattern recognition (Goldbaum, Taylor, Abramoff, Kelvin, PerezRovira, Stewart). However, our examination of the currently existing software complexes for analysis of vascular system images showed that most of them do not have applied software for measuring of the full set of diagnostic signs and diagnosis. They have only tools for image registration, recordkeeping diagnostic information about patients and most frequently used tools for preprocessing, the image quality improvement and marking the images. Therefore, the urgent problem is developing of a system for analysis of subclinical morphological changes allowing one to automatize diagnostic stages and to carry out quantitative monitoring of pathologic vessel changes. OPTICAL MEMORY AND NEURAL NETWORKS (INFORMATION OPTICS)

Vol. 23

No. 4

2014

282

ILYASOVA

Diagnostic software complex “OphthalmOffice” for analysis of the vessel system of eye fundus. We developed a method for diagnostics of early stages of diabetic retinopathy and a methodology of quantita tive estimation of elements of pathomorphology picture of eye fundus physicians use when defining a degree of retina pathology. Our method allows one to achieve higher measuring accuracy of diagnostic signs accustomed to the majority of physicians that can correspond to the physician own experience and the results of known clinical investigations [15]. The software complex is intended for the eye fundus vessels tracing, localization and analysis of the region of the optic nerve disk, calculation of the basic set of geometrical characteristics of the eye fundus vessels, as well as calculation of diagnostic signs, classification of vessels and carrying out planimetric stud ies that include dynamic analysis of images. The functional specification of this complex includes entering and preprocessing of eye fundus images, quantitative estimation of the basic types of microvessel changes in diseases: the mean diameter alteration, the diameter irregularity, the vessel tortuosity, and so on. It also provides the objective control of dynamics of changes of the pathological region sizes on subsequent images, estimation of the degree of pathology, automatic entering of database on patients, images and measurements. As input data, it uses images in BMP, JPG, GIF, TIF, PNG formats. This allows the user to obtain input data from practically any available device for digital registration of eye fundus images. The diagnostic complex divided logically into the following systems: the system that guarantees data storing and generates reports; the system of estimation of eye fundus and optic nerve disk diagnostic signs; the sys tem for planimetric analysis; the system for classification and diagnostic research. The system for classification and diagnostic examination provides the tools of the correlation and dis criminant analyses for generating the space of more informative signs. For adjusting the classifier, it pro vides the tools for generating the optimal sampling with regard to the criterion of the efficiency of dividing into groups of pathologies [16]. For the learning sampling filtering with the purpose of eliminating of invalid data and receiving the normative values of signs for different groups of pathologies it provides the tools of the cluster analysis. The system makes it possible for a user to govern the study process. The system of data analysis serves for diagnostic decisionmaking, as well as for generating standard values of signs for each type of vessel pathologies. Thus, data mining allows the user to determine the pathology degree, the standard values of the signs for each degree of the disease pathology, as well as the probability of the disease development. A characteristic feature of the diagnostic complex is using of elements of expert systems, such as diagnostic sign database, the correlation, discriminant and cluster analyses with screenings of invalid data, prediction of the pathology degree based on expert judgments. When developing this complex: 1. We worked out a new generalized mathematical model of two classes of blood vessels images, namely eye fundus vessels and coronary vessels. This model allows one to formalize the description of geometrical characteristics and to determine diagnostic signs of eye fundus vessels and coronary vessels [17, 18]. 2. We proposed a complex of methods and algorithms for estimation of the basic set of vessel geomet rical characteristics and analyzed their efficiency for different vessel types [19, 20]. 3. For vessel tracing and estimation of local directions we worked out a new method of the modified local fan transform based on analysis of the luminance radial functions in the region of the sliding sector inside the scanning window (up to now a modification of the ray transform of the luminance function was used) [21]. 4. We developed an information technology for generation of the space of effective signs for analysis of eye fundus vessels images. Then basing on the discriminant analysis we selected the most informative sets of diagnostic signs allowing us to diminish errors up to 2.5% when sorting the vessels by groups normal and different stages of pathologies (insular diabetes) [18]. 5. We solved the problem of selection of vessel segment central lines basing on a modified wavelet trans form by using undirected 2D wavelets formed from 1D ones according the proposed model of vessels (other authors did not take into account vessel models, and this led to selection of different artifacts together with the vessels). Our method showed a high accuracy when selecting vessel central lines, high noise immunity (due to the wavelet smoothing properties) and invariance with regard to the vessel orien tation and the vessel image luminance. The algorithm steady work is certified when the ratio noise/signal is less then 0.9 and at that, the values of the vessel trace standard deviation is not larger then 0.6 pixels (Fig. 2) [22]. The system of coronary vessels analysis. The proposed system is intended for solving numerical simu lation problems, estimation of geometrical characteristics and visualization of threedimensional struc OPTICAL MEMORY AND NEURAL NETWORKS (INFORMATION OPTICS)

Vol. 23

No. 4

2014

COMPUTER SYSTEMS FOR GEOMETRICAL ANALYSIS

283

Standard deviation 4

Standard deviation 2.5 2.0

10 20 30 40

3

1.5

2

1.0 1

0.5 0 0.01

0 0.1

1

100 10 Ratio noise/signal

1

2

3

4

5

5

7

8

9

10 Scale

Fig. 2. The dependence of the standard deviation of the central line: on the value of the ratio noise/signal (left), on the noise level for different scales (σnoise changes from 10 to 40).

ture of the vessel tree basing on the data of the angiographic study of the coronary vessels [23]. The func tional specification of the complex includes: (1) data reading of xray angiographic studies of coronary vessels in DICOM v.3.0 format; (2) modeling of the filming geometrical characteristics, neutralization of geometrical distortions and synchronization of projections, (3) numerical simulation and estimation of geometrical parameters of the threedimensional structure of the coronary vessels, (4) visualization of the spatial structure of the coronary vessels. The methods and algorithms of processing and analyzing of coronary vascular systems images developed under the investigation process made the basis of four subsystems of the program complex “CardiOffice”: (1) the subsystem for input data processing. It reads data files in the DICOM format and transforms this data into the intrinsic format of the system; (2) the preprocessing subsystem. It accomplishes the time synchronization of the projections and geo metrical compensation of the projections images. This subsystem includes some standard algorithms for image refinement; (3) the modeling subsystem. It consists of two modules: the module for geometrical binding of projec tions and the tracing module. The subsystem realizes the modeling of the filming geometrical character istics, performs data simulation and estimation of geometrical characteristics of the threedimensional structure of the coronary vessels; (4) the visualization subsystem. They realize the visualization of the spatial structure of coronary vessels using the tools of the OpenGL graphic library (Fig. 3). The data accumulation procedure for the following processing was carried out in hospitals Tel Hashomer and HADASA (Jerusalem) with the aid of special CARM equipment. The obtained cinean giography data included information about the cineangiography geometry, the projection films constitut ing of frames that are angiographic images of coronary vessels made in fixed instants of time and under a certain angle. They all were stored in a uniform medicine information storage format DICOM. In average, each separate document in DICOM format contains 4–6 films with images of the left coronary artery and 2–3 films with images of the right one. Each frame is an image in BMP (GreyScale) format and its size is 512 × 512 pixels. The cineangiography of the right and left coronary arteries are done in series. At first, one composed all projections of the right heart (the filming time of one projection is about 3–5 seconds), then the doctor changes the angle of the cineangiography and repeats the procedure. We tested the effi ciency of the developed diagnostic software complexes using rather large number of test and onsite images. We showed that our results guaranteed 95% of their reliability and stability of their repeatability. When developing the system we used the information technology of restoration of threedimensional structure of coronary vessels with the aid of a small number of angiographic projections. It allows one to estimate spatial geometrical characteristics of the coronary vessel system and to formulate diagnostic signs under the conditions of the unmatched projections. We solved the problem of restoration of spatial struc ture of vessels using a small number of planar projections under the view. This problem belongs to the class of illposed problems and therefore it is extremely complex. The problem complexity increases since the OPTICAL MEMORY AND NEURAL NETWORKS (INFORMATION OPTICS)

Vol. 23

No. 4

2014

284

ILYASOVA

Fig. 3. The restoration results: the initial projections of coronary vessels, graphical interface of the visualization sub system.

object under investigation is a dynamic one, and therefore due to the nonsimultaneous registration we, in fact, study projections that are unmatched in time. Besides, there is noise on the Xray angiographic projections due to use of low energy Xradiation. We propose the information technology of automatic restoration of the threedimensional structure of the vascular system and estimation of its geometrical characteristics. The main method it uses is the simul taneous analysis of vessel images on the projections and the parallel restoration of the geometrical structure of the vascular tree. The specific feature of this technology is using of a small number of angiographic pro jections. It allows one to restore spatial structures of objects that are most closely correspond to given projec tions from available noisy central projections of a dynamic object obtained in different instances of time. Results of our investigation. In the course of our studies, we proved that 1. when identifying vessel pathologies from digital images of eye fundus, it is effective to use quantita tive estimates of the overall set of geometrical characteristics as integral indicators of the vessel state; 2. the developed new generalized mathematical model of images of eye fundus vessels and coronary vessels allows one to formalize the description of the basic geometrical characteristics, to estimate them and to formulate diagnostic signs; OPTICAL MEMORY AND NEURAL NETWORKS (INFORMATION OPTICS)

Vol. 23

No. 4

2014

COMPUTER SYSTEMS FOR GEOMETRICAL ANALYSIS

285

3. to solve the problem of the calculation accuracy improvement, it is worthwhile to use a complex of methods and algorithms of local diameters estimations including methods of indirect measuring of the parameters, as well as approximating methods based on using of different models of parametric approxi mation of the luminance profile according to the type of the vessel under analysis; 4. the modified method of the local fan transform based on analysis of the luminance radial functions of the region of the sliding sector inside the scanning window is efficient when estimating local directions of vessels, and for analysis of their tortuosity, branching and ending; 5. the new method of the modified wavelet transform based on using 2D wavelets formed from 1D ones by continuation along the second coordinate according to the vessel model is an efficient tool for solving problems of selection of the vessel central line; 6. our diagnostic software complex based on the information technology of the eye fundus images anal ysis includes the algorithms of generation and extracting signs important for diagnostic with the aid of the discriminant analysis, allows one to increase the efficiency of classification of the vessels dividing them into groups, which are “norm” and four stages of the diabetic retinopathy; 7. the system of coronary vessel analysis based on the information technology of restoration of spatial structure of coronary vessels from a small number of available unmatched angiographic projections allows one to estimate spatial geometric characteristics of the coronary vessels and to formulate diagnostic signs. ACKNOWLEDGMENTS The work was supported by the Ministry of Education and Science of the Russian Federation; RFBR grants 120100237a, 140100369a, 140797040p; the 2013–2014 Program “Bioinformatics, mod ern information technologies and mathematical methods in medicine” of the basic research of DNIT of RAS. REFERENCES 1. Ilyasova, N.Yu., Methods for digital analysis of human vascular system. Literature review, Computer Optics, 2013, vol. 37, no. 4, pp. 517–541, ISSN 01342452 (in Russian). 2. Ortega, M., Sirius: a webbased system for retinal image analysis, Ortega, M., Barreira, N., Novo, J., Penedo, M.G., PoseReino, A., and GómezUlla, F., Eds., International Journal of Medical Informatics, 2010, vol. 79, pp. 722–732. 3. Niemeijer, M., Retinopathy online challenge: automatic detection of microaneurysms in digital color fundus photographs, Niemeijer, M., van Ginneken, B., Cree, M.J., et al., Eds., IEEE Transactions on Medical Imaging, 2009, vol. 29, pp. 185195. 4. Li, Q., Colour Retinal Image Segmentation for ComputerAided Fundus Diagnosis Department of Computing, Qin Li, Ed., The Hong Kong Polytechnic University, 2010. 5. Stewart, C.V., RIVERS: Retinal Image Vessel Extraction and Registration System [Electronical Resource] / Stewart, C.V. and Roysam, B., URL: http://cgivision.cs.rpi.edu/cgi/RIVERS/index.php.in. 6. Tsai, C.L., Automated retinal image analysis over the internet, Tsai, C.L., Madore, B., Leotta, M.J., Sofka, M., Yang, G., Majerovics, A., Tanenbaum, H.L., Stewart, C.V., and Roysam, B., Eds., IEEE Transactions on Infor mation Technology in Biomedicine, 2008, vol. 12, pp. 480–487. 7. PerezRovira, A., VAMPIRE: vessel assessment and measurement platform for images of the REtina, Perez Rovira, A., MacGillivray, T., Trucco, E., et al., Eds., Engineering in Medicine and Biology Society, EMBC, 2011 Annual International Conference of the IEEE, 2011, pp. 3391–3394. 8. Kelvin, P., Livevessel: extending livewire for simultaneous extraction of optimal medial and boundary paths in vascular images, Kelvin, P., Ghassan, H., and Rafeef, A., Eds., Proceedings of the 10th International Conference on Medical Image Computing and ComputerAssisted Intervention, Brisbane, Australia: SpringerVerlag, 2007. 9. “Gamma Multivox” [electronic resource]: http://www.gammamed.ru/MultivoxGammaD1.html, http://www.multivox.ru/multivox2d.shtml. 10. “Myrian® XPVessel”—http://www.eukon.it/site/download/Myrian_XPVessels.pdf. 11. “Electron”—http://electronxray.com/equipment/rentgenohirurgicheskie_apparaty/angiograf/angiograf oko_2_wp/. 12. “AlluraXper” [electronic resource]: http://www.healthcare.philips.com/ru_ru/products/interventional_xray/ Product/interventional_cardiology/imaging_systems/intcardio_fd20.wpd, http://www.healthcare.philips.com/ ru_ru/products/ interventional_xray/Product/interventional_cardiology/imaging_systems/intcardio_fd10.wpd. 13. “InterViewFusion”—http://www.mediso.hu/products.php?fid=1,10,6&pid=67. 14. “General Operator Processor” ContextVision [electronic resource]—http://www.contextvision.com/our technology/gopimageenhancement/. OPTICAL MEMORY AND NEURAL NETWORKS (INFORMATION OPTICS)

Vol. 23

No. 4

2014

286

ILYASOVA

15. Ilyasova, N.Yu., Diagnostic computer complex for vascular fundus image analysis, Biotehnosfera, 2014, vol. 33, no. 3, pp. 20–24 (in Russian). 16. Ilyasova, N.Yu., Formation features for improving the quality of medical diagnosis based on the discriminant analysis methods, Ilyasova, N.Yu., Kupriyanov, A.V., and Paringer, R.A., Eds., Computer Optics, 2014, vol. 38, no. 4, pp. 751–756, ISSN 01342452 (in Russian). 17. Ilyasova, N.Yu., Measuring Biomechanical Characteristics of Blood Vessels for Early Diagnostics of Vascular Reti nal Pathologies, Ilyasova, N.Yu., Ananin, M.A., Gavrilova, N.A., and Kupriyanov, A.V., Eds., Lecture Notes in Computer Science. Medical Image Computing and Computer Assisted Intervention, MICCAI 2004, Proceedings of 7th International, Conference SaintMalo, France, 2004, September, vol. 3217, issue 1, part 2, part II, pp. 251– 258. 18. Ilyasova, N.Yu., Estimation of geometric features of the blood vessels spatial structure, Computer Optics, 2014, vol. 38, no. 3, pp. 529–538, ISSN 01342452 (in Russian). 19. Ilyasova, N.Yu., Methods for formation of features of treelike structures on fundus images, Ilyasova, N.Yu. and Yatul’chik, V.V., Pattern Recognition and Image Analysis, MAIK “Nauka/Interperiodica,” 2006, vol. 16, issue 1, pp. 124–127. 22. Anan’in, M.A., Estimating directions of optic disk blood vessels in retinal images, in Pattern Recognition and Image Analysis, Anan’in, M.A., Ilyasova, N.Yu., and Kupriyanov, A.V., Eds., MAIK “Nauka/Interperiodica”, 2007, vol. 17, issue 4, pp. 523–526. 21. Kupriyanov, A.V., Geometrical parameters estimation of the retina images for blood vessels pathology diagnos tics, Kupriyanov, A.V., Ilyasova, N.Yu., and Ananin, M.A., Eds., Proceedings of 15th European Signal Processing Conference September 3–7 2007, EUSIPCO 2007, Poznan, Poland, 2007, pp. 1251–1254. 22. Ilyasova, N., A method of the wavelet transformation for estimation of geometrical parameters upon the diag nostic images, Ilyasova, N., Korepanov, A.O., and Kupriyanov, A., Eds., Optical Memory & Neural Networks, 2009, vol. 18, no. 4, pp. 343–348. 23. Ilyasova, N.Yu., Computer technology for the spatial reconstruction of the coronary vesels structure from angi gographic projections, Ilyasova, N.Yu., Kazanskiy, N.L., Korepanov, A.O., Kupriyanov, A.V., Ustinov, A.V., and Khramov, A.G., Eds., Computer Optics, 2009, vol. 33, no. 3, pp. 281–318 (in Russian).

OPTICAL MEMORY AND NEURAL NETWORKS (INFORMATION OPTICS)

Vol. 23

No. 4

2014