Abstract
Content
2.1 A method choice for a compression of medical images
2.2 A method choice for an image filtration
5. REVIEW OF RESEARCHES AND DEVELOPMENT
5.1 Review of international sources
5.2 Review of national sources
INTRODUCTION
Many centuries the mankind seeks to create adaptations which could simplify or at all solve an objective for the person. One of such adaptations undoubtedly is the computer which basic principles of work are formulated in far 1945 by the well-known mathematician John von Neumann. It should be noted that the same fundamental principles of operation of the computer are actual and to this day.
Computer technologies in the modern world develop at a frantic pace, new technologies are created, improved existing, and the computer gets into all fields of activity of the person more and more. Not an exception is also the medical branch: computers help to carry out operations for which microscopic accuracy – expression isn't simple, restore, apparently, irrevocably spoiled pictures of bodies and tissues of the person, and thanks to development of the systems of support of decision-making (SSDM) can help the doctor to make the diagnosis at all and to appoint treatment.
1. RELEVANCE OF THE SUBJECT
One of the most important directions of use of computer technologies for modern medicine is processing of images. Traditionally this area shares on 3 components: improvement of quality of the image, allocation of borders of object, and scoping of object on a series of images.
Aspect of quality of the image – a cornerstone of modern medicine. Quite often real images are distorted at compression or on the area of the image interesting the doctor shadows of other bodies are cast that in turn complicates statement of the correct diagnosis. Therefore the problem of cleaning of the image of noise and shadows is actual.
2. PROBLEM DEFINITION
2.1 A method choice for a compression of medical images
The leading role in medico-diagnostic information is occupied by medical images which are in turn characterized by huge volumes of information. If to take into account limitation of space for data storage and the capacity of communication channels – importance of compact submission of information comes to the forefront.
The basic and the most logical way out is application to the initial image of compression algorithms. For a right choice of algorithm it is necessary to consider efficiency and speed for each class of images (color, gradation gray and black-and-white).
For medical images that algorithms of compression without loss of data which in turn provide the only and unambiguous option of recovery of data are applied to compression of data is standard. But unfortunately the coefficient of compression of such algorithms as a rule isn't great, about 3/8. As a result it is possible to draw a conclusion that for achievement of better results it is necessary to use the algorithms based on compression methods with losses of data that will allow to increase compression coefficient much. But also you shouldn't forget that the chosen algorithm has to provide as safety of diagnostic information, and to provide reliability in respect of recovery of data.
As algorithm of compression with losses of data the algorithm is chosen wavelet transformations (WAVELET) which perfectly is suitable for work with images in gradation gray, namely to such look and reduction of the initial image is planned [1].
2.2 A method choice for an image filtration
Obtaining the high-quality digital image which is characterized by the high level of specification and contrast, the good level of brightness is a necessary condition for modern medical diagnostics. Algorithms the carrying-out similar transformations are called as algorithms of improvement of quality of the initial image.
The vast majority of the existing methods of improvement of quality are based on empirical or heuristic approach. Practically all from them can't work in completely autonomous mode, demanding amending and adjustments of some parameters from the user.
For providing the full-fledged autonomous mode of improvement of quality of the image it is decided to use the genetic algorithms (GA) [2].
2.3 Forthcoming plan of works
Practical realization of an objective can be presented in the form of the following scheme:
Figure 1 – Plan of practical implementation of the project
The stage 1 – the entrance image will be transformed to gradation of the gray.
The stage 2 – the image received at the previous stage is compressed according to algorithm wavelet transformations.
The stage 3 – the image received at the previous stage is filtered. Parameters of a filtration are calculated by means of HECTARE.
The stage 4 – the filtered image is restored according to algorithm wavelet transformations and if necessary will back be transformed to the color image. At the exit from this stage we will receive the resultant image.
3. WAWLET TRANSFORMATION
The simplified option wavelet transformations it is possible to present as follows (fig. 2):
Figure 2 – The simplified scheme of wavelet transformations
At the first stage the entrance image will be transformed to space like chromaticity/brightness. For transfer to space of YUV it is necessary to use the following formulas
where Y – the brightness component; U and V – the components which are responsible for color (respectively chromatic red and chromatic blue); R, G and B – the components of color presented according to RGB model where R – a component responsible for red color; G – for the green; B – for the blue.
For the return transformation to the RGB model it is necessary to multiply the vector consisting of Y, U, V values on the return matrix from coefficients at R, G, B respectively and subsequently to subtract a column of free coefficients from the turned-out vector.
At the second stage the image is processed by two filters: high-frequency (VCH) in the lines and low-frequency (LF) on columns with the subsequent thinning.
Filters represent small "windows". Values of brightness and the chromaticity getting to "window" of pixels are multiplied with the set of coefficients of the window and then summarized, and "window" moves for calculation of the following value. Counting of LF and VCH of components are defined according to expressions (4) and (5).
where n = 0, 1 … (N/2 – 1); C (2*n), C (2*n + 1) – counting of the entering digital signal; L (n) – low-frequency coefficients of transformation by the filter; H (n) – high-frequency coefficients of transformation by the filter.
As a result of a filtration instead of one size of the image m*n, it will be received (m/2) 4 images in size * (n/2). Result of a low-frequency filtration across or verticals as a rule is the most informative image which is exposed to further processing. Results of a high-frequency filtration are in most cases rejected.
The image received after processing of LF and VCH by filters is quantized and after coding gets to an output stream, as a result at the exit we receive the massif of certain numerical coefficients.
At the following stage the massif of the received coefficients is quantized also values close to zero are rejected. The new massif is coded according to algorithm therefore at the exit the squeezed image turns out.
4. GENETIC ALGORITHM
Now quickly the new direction in the theory and practice of artificial intelligence – the evolutionary calculations (EC) develops [3]. This term is usually used for the general description of algorithms of search, optimization or training based on some formalized principles of natural evolutionary selection. Features of ideas of evolution and self-organization are that they find confirmation not only for biological systems, the developing many billions of years. These ideas with success are used now when developing many technical and, in particular, program systems.
The Genetic Algorithm (GA) is the heuristic algorithm of search used for the solution of problems of optimization and modeling by casual selection, a combination and a variation of required parameters with use of the mechanisms similar to natural selection in the nature [2].
Genetic algorithms in that look which they represent now, were based at Michigan University (University of Michigan) by the researcher John Holland [4] and originally developed for a problem of optimization as rather effective mechanism of combinatory search of versions of the decision. Unlike many other works, J. Holland's purpose was not only the solution of specific objectives, but research of the phenomenon of adaptation in biological systems and its application in computing systems. Thus the potential decision – an individual is submitted a chromosome – a binary code. Population contains a set of individuals. In the course of evolution four main genetic operators are used: reproduction, selection, crossover and mutation.
The principle of work of GA consists in the following: first – each individual is the solution of specifically put problem; secondly – a set of individuals make themselves population, that is area of potentially possible versions of the solution of an objective; thirdly – search of the optimum solution of an objective is carried out in the course of evolution of population by means of above-mentioned operators of a reproduction, selection, a crossover and mutation.
As well as process of biological evolution genetic algorithms are based on 3kh the fundamental principles:
1) The strongest survives [5]. The principle was formulated by Charles Darwin in 1859 in the book "Origin of Species by Natural Selection". The essence of the principle is that according to Ch. Darwin the individuals who were more adapted for the solution of the task in a certain environment survive and bring healthy posterity, weak on the contrary die out. This principle was described by two first genetic operators: reproduction and selection.
2) The chromosome of the descendant consists of parts of chromosomes of parents. The principle was formulated by Gregor Mendel in 1865. Realization of this principle is the operator of a crossover.
3) In the course of evolution individuals have genes which are absent at parents (mutation). The principle was formulated by Hugo de Friz in 1890 in the book "Mutational Theory". The term a mutation was used by it for the description of sharp changes of properties of descendants or acquisition of the new properties, which are absent at parents thereby increasing variability of population. Process of performance of GA can be broken into 10 stages, which are for descriptive reasons illustrated on the flowchart (fig. 3).
Figure 3 – Algorithm of work of GA
5. REVIEW OF RESEARCHES AND DEVELOPMENT
5.1 Review of international sources
Examples of works of our foreign colleagues in the similar direction are:
1) Sudipto Gyyukho and Boulos Harb article on a subject "Algorithms of approximation for wavelet transformations of the coded data flow" [6];
2) 2) Chao Gongg-T and Ksi Dong-Mey article on the subject "The Analysis of Algorithm Wavelet Transformations and Its Realization in Language C" [7];
3) 3) J. Beylkin, R. Coifman and V. Rokhlin article on a subject "Fast wavelet transformations and numerical methods" [8].
5.2 Review of national sources
Examples of works of our compatriots are:
1) A.M. Gubsky article on the subject "Combination of the wavelet-analysis and Genetic Algorithm for Minimization of Errors of the Global Navigation System". The multilevel algorithm of an assessment of errors of diverse sources of the navigation system on the basis of wavelet-transformation with intellectual control of a wavelet by means of genetic algorithm and an assessment of an error separate sources by optimization of parameters of multilevel wavelet-transformation is offered [9];
2) Soroki A.M., Kovalets P.E., Heydorova I.E. article on the subject "Method of Classification of Speech Signals with Use of the Adaptive Description on the basis of wavelet-transformation and Genetic Algorithm". The method of creation of the description on the basis of adaptive wavelet-function which use allows to increase distinctive ability of the description is presented in article. The multi-stage method of classification of phonemes of Russian with use of the received adaptive wavelet-functions, the considering way of formation of phonemes and allowing to increase the accuracy of classification of phonemes is offered[10];
3) A.V. Nitsenko and T.S. Hashana article on the subject "Application of Complex Continuous Veyvlet-transformation of Morle in Processing of Audiosignals". In article the modified method of a phase voice coder on the basis of complex continuous transformation of Morle is offered. The combination of the specified method, linear interpolation and the KIH-filter allowed to develop algorithm of change of rate of a signal without change of its tone [11].
5.3 Review of local sources
Students and employees of DONNTU conducted the following researches:
1) Article of Yu.A. Skobtsova, V. Yu. Skobtsov, K.M. Nasser Iyad – the Checking crosstalk tests of malfunctions on the basis of evolutionary methods. The problem of creation of the checking tests for "cross malfunctions" of the Crosstalk type, characteristic for deep submicronic design of element base of modern computer systems is considered. At the solution of an objective multiple-valued logical modeling and genetic algorithm of generation of the checking tests for these malfunctions is used [12];
2) Manual – Skobtsov Yu.A. Bases of evolutionary calculations [4].
CONCLUSIONS
Various methods of a filtration of images have in the basis pixel-by-pixel operations, that is consecutive imposing of various filtering masks on each point of the image.
The analysis of the existing methods of a compression of medico-diagnostic images with the subsequent filtration showed that it is expedient to improve and optimize already existing algorithms for the solution of the specific set objective.
The main task of a master's thesis we will consider an application creation for a compression, improvements of quality of the image on the basis of HECTARE, and decompressions of the initial medico-diagnostic image. It is necessary to develop project documentation for correct use of the developed application.
Addition of a master's thesis with additional researches in various, directly connected with the studied subject, areas is possible.
At the time of writing of this paper the master's thesis isn't finished yet. Final end: December, 2015 – January, 2016. The full text of work and materials on a subject can be received directly at the author or his research supervisor after the specified date.
THE LIST OF THE USED SOURCES
1. Wawlet transformation // Internet resource – access mode: https://ru.wikipedia.org/wiki/Вейвлет-преобразование.
2. Genetic Algorithm // Internet resource – access mode: https://ru.wikipedia.org/wiki/Генетический_алгоритм.
3. Evolutionary computation // Internet resource – access mode: https://ru.wikipedia.org/wiki/Эволюционные_алгоритмы.
4. Скобцов Ю.А. Основы эволюционных вычислений // Учебное пособие. – Донецк:ДонНТУ – 2008 – 326с.
5. Ч.Дарвин. Происхождение видов путем естественного обора или сохранение благоприятных рас в борьбе за жизнь // Интернет ресурс – режим доступа: http://charles-darwin.narod.ru/origin-content.html.
6. Sudipto Guha, Boulos Harb. Approximation Algorithms for Wavelet Transform Coding of Data Streams // Интернет ресурс – режим доступа: http://arxiv.org/pdf/cs/0604097.pdf.
7. Zhao Hong-tu, Xi Dong-mei. Analysis on Algorithm of Wavelet Transform And Its Realization in C Language // Интернет ресурс – режим доступа: http://www.academypublisher.com/proc/isecs10w/papers/isecs10wp336.pdf.
8. G. Beylkin, R. Coifman, V. Rokhlin. Fast Wavelet Transforms and Numerical Algorithms I // Интернет ресурс – режим доступа: https://amath.colorado.edu/faculty/beylkin/papers/BE-CO-RO-1991.pdf.
9. А.М. Губський. Комбінація вейвлет-аналізу та генетичного алгоритму для мінімізації похибок глобальної навігаційної системи // Науковий вісник НЛТУ України. – 2011. – С. 355-362.
10. Сорока А.М., Ковалец П.Е., Хейдоров И.Э. Метод классификации речевых сигналов с использованием адаптивного признакового описания на основе вейвлет-преобразования и генетического алгоритма // Речевые технологии. – № 3-4. – 2013. – С. 84-95.
11. А.В. Ниценко , Т.С. Хашан. Применение комплексного непрерывного вейвлет-преобразования Морле в обработке аудиосигналов // Интернет ресурс – режим доступа: http://dspace.nbuv.gov.ua/bitstream/handle/123456789/7671/095-Nitsenko.pdf?sequence=1.
12. Ю.А. Скобцов, В.Ю. Скобцов, К.М. Нассер Ияд. Проверяющие тесты crosstalk неисправностей на основе эволюционных методов. // Вестник «Информатика и моделирование» – 2010. – С. 170-176.