INTRODUCTION

At present, medical laboratories growing culture of cells, it is necessary to predict the timing of receipt ripe for transplant material. This need is often associated with complex preparations patients to operations, the specifics of cultured cells, not admitting perezrevanie or maintaining its viability outside the incubation time is extremely limited. Designation method. is for morphological studies of tissue samples obtained from endoscopical saddle and puncture biopsy, operations. Diagnostic value method. Makes svetoopticheskom level to diagnose all kinds of dystrophic, inflammatory, compensatory-adaptive and tumor processes.
The presentation of the results of the study. ¬ ment study results presented in a text protocol containing a description of processing technology material, microscopic images, the conclusion (formulated in accordance with international classification of diseases), additional comments (contains additional diagnostic meaningful updates the conclusion) and recommendations. When using methods possible bakterioskopicheskih verification of a number of infectious disease vectors (helikobakterioz, tuberculosis, etc.).

Objects in the medical images have great complexity and multifactorial, which makes high demands on the reliability, accuracy and reliability of research results. Using computer technology and mathematical methods in this industry not only expedite the processing of material, but also improve the accuracy of the results of the study.
The main reason for the lack of automation in histology is the high variability and poor contrast majority histological structures. However, the rapid development of digital and analog technology in the recently opened new opportunities for developers. For example, the increase in speed computer technology allows the use of complex, critical at the time of algorithms, and thanks to the emergence of colour television high-resolution sensors can receive and process color. This new technical possibilities allow greatly expanded range of research, open new ways to solve problems related to image analysis. This work is devoted to one of these tasks - segmentation of the image histological preparations.
The purpose of work - to develop a segmentation algorithms to determine the objects on the histological slabokontrastnyh colour images and half to meet the challenges of diagnosing diseases. To achieve this goal required: ? histological classified images of objects in the geometric, topological, optical characteristics; ? develop algorithms segmentation cells; ? develop techniques histological colour images of objects;

Thus, before being subjected to analysis, image must undergo a training phase, which consists in carrying out operations improve visual quality (increased contrast, the removal of fuzziness, underlining boundaries, filtering) operations and the formation of graphic drug (segmentation, allocation of units) image.

Рис.1 - Procedures and methods of recognition of images
(121 КБ (124 664 байт))

Changes in contrast

Smoothing noise

Images at the stage of digitization are exposed to the additive and pulse noise. Additive noise represents some random message that added to a useful outlet system, in this case additive noise caused by grain films. Pulse noise, unlike additive, characterized the impact on the useful signal in only a few random locations (the value of the resulting signal in these locations Takes random value). Pulse noise is typical for digital transmission systems and storage of images.

Underline borders

Methods smoothing images can remove noise very effectively. A significant drawback is smoothing algorithms smaz images (that is, reduction Definition of contour elements), while the value smaza proportional to the size of masks used for smoothing. For a clear image analysis, especially when calculating geometric structural elements, it is very important to remove smaz with the contours of objects in the image, that is strengthened the difference between the gradations of brightness contour of the object and neighboring elements background. In this case, image processing techniques underscores contours. Usually emphasizing the borders is carried out using high spatial filtering. Specifications filters set in a mask, in which the average value must be equal to zero. Another way to underscore the borders is the so-called static differentiation. The method of brightness value of each element is divided into a statistical evaluation of RMS.

Median filtering

Median filtering applies to nonlinear techniques of image processing and has the following advantages over linear filtration (classical procedures smoothing): retains the sharp differences in (the border); effectively smooths pulse noise; does not change the brightness of the background. Median filtering is implemented by some movement aperture (masks) along discrete images and replacement values central element masks median (mean orderly sequence) building blocks inside the aperture. In general, aperture can have many different forms, but in practice more often all applicable square aperture.

Segmentation images

Under the segmentation of the image refers to the process of splitting at its constituent parts, having meaningful sense: objects, their boundaries or other informative Samples characteristic geometric features, etc. In the case of automation techniques for obtaining images segmentation should be seen as basic start-up phase of analysis, is to construct a formal description of the image quality of which largely determines the success of solutions task recognition and interpretation facilities.

Methods of allocating units

There rarely are confronted with the task of finding perimeters, curvature, form factors, specific surface facilities, etc. All of the tasks so or otherwise associated with analysis of contour elements objectes.Metody highlight the contours of the image can be divided into the following main classes:
• high-frequency filtering techniques;
• methods of spatial differentiation;
• methods of functional approximation.
Common to all these methods is to consider the border area as a sudden image brightness function f (i, j); same distinguishes them entered the notion of a mathematical model search algorithm border and boundary points.

The method of image first appeared as a ready-to-use technical means in 1963, along with the development QTM (KTM -- Quantitative Mikroskopa Television), subsequently became part of the company "MEKOC". This device was intended for use in metallurgical laboratories - especially for the quantitative monitoring of the purity of steel and other mikrostrukturnyh measurements, but soon made it clear usefulness this device and other areas.One of the first applications in biology was the measurement of the size of airspaces in the lungs (which is required for quantitative description of the degree of pulmonary lesions) and for counting the number of grains of silver in avtoradiografii.
Since then advanced image analysis technology has found its application in almost all scientific and technical fields of natural science, ranging from anatomy with zoologists, and has expanded its capacity to ensure that include mathematical processing functions such as filtering and image enhancement.

The MECOS-C2 automated microscopy system (AMS) family is intended for medical, scientific, sanitary, educational, veterinary, agricultural, manufacturing laboratories. MECOS-C2 AMS complete set includes the following main parts.
1) Microscope trinocular. Above 100 models of all microscopy types of main manufacturers can be used.
2) Digital videocamera. Above 50 models of leading manufacturers for all analysis types are applied.
3) Microscope motorization equipment for specimen relocation and focusing under computer control. Equipment of MECOS production (MECOS-MS2) or other manufacturers can be used.
4) PC or notebook of different types.
5) Special software of MECOS production (MECOS-C2soft) including:
- "virtual microscopy" platform for information services;
- group of functional program - techniques for analyses of different types production;
- "reference virtual slides" program for quality control and personal training.
6) Devices for specimen preparation of different manufacturers (Options).
Some program - techniques can work without automation equipment using usual manual microscope.

An example is the medical complex computerized image analysis "DiaMorf" used in hospitals and research institutes. Specialized complexes "DiaMorf" provide automatic entry microscopic images, objects selected photo (cell nuclei, plots different colour or brightness). Is developed tools to take measurements in the picture: linear dimensions, perimeter, area, optical parameters, the situation objects. Statistical processing subsystem holds a mathematical measurement results with automatic construction of a wide range of histograms, graphs, tables.
Software set in automatic mode perform the following functions both quantitative and qualitative image analysis: According to group objects: the number of objects, total perimeter, the total area, total integral optical density.

VideoTesT - Master (Morphology) is designed for entry, conversion and image analysis in medicine and biology. The analysis processed statistically.
VideoTesT - Master (Morphology) includes six pre-analysis techniques:
"Measurement", "Counting and Measurements", "Volume Share", "Eritrotsitometriya", "PodschetTrombotsitov", "NCR". It is also an opportunity to work in a free mode ( "Netmetodiki").
The program also allows you to create new methods of analysis for various user tasks. Using the methodology entire sequence of operations over the image is automatically.
VideoTesT - Master (Morphology) contains six pre-techniques:
1. The technique "Measurement" is used to measure morphological and brightness settings objects (nuclei of cells, etc.) in the preparations.
2. The technique "Counting and Measurements" is used to measure morphological and brightness settings objects (nuclei of cells, etc.) in the preparations for and automatic classification of measured objects.
3. The technique "Share volume" is used to assess the areal ratios different colour (or brightness) phases in histological preparations.
4. The technique "Eritrotsitometriya" is intended to measure erythrocyte in stained smears and constructing histograms distribution of erythrocytes size.
5. The technique "Counting of platelets" is used to assess the content of platelets on erythrocytes in stained smears.
6. The technique "NCR" is intended to calculate the nuclear-cell relations.
The program also allows to operate in free mode, without any methodology. Using a version of the work, you can enter the picture, and then process it with using the full range of functions available in the program (the conversion, editing, measurement, etc.).
IMAGES AND DATA TRANSFER THROUGH BUFER Images and data can be transferred to other applications, Windows (MS Word, MS Excel, etc.) through the buffer temporary storage.

The work is dedicated to the problem of segmentation of the image histological preparations. Its goal is the development of algorithms to identify histological objects in the image preparation, retaining geometric and optical properties of the object. A classification of objects to determine segmentation algorithm. A half-utonsheniya algorithm that takes into account the characteristics of images histological preparations.
На основе методов morphology developed mathematical algorithms segmentation receptacles and fibers with small and large optical increase, as well as identification algorithm vessels and fibers with a large increase, using the results of segmentation algorithm. Segmentation algorithms developed polygonal objects (cell nuclei cells, cross-sectional vessels and fibers) methods of mathematical morphology and association areas, as well as an algorithm to determine the cells binary images obtained as a result of the threshold segmentation.
To perform segmentation histological objects on the colour images developed coordinate system describe the color of PHS. An image analysis system Bioscan, which implemented the above algorithms. Obtained in dissertation work results are intended for implementation in automated systems analysis and histological preparations can be used in traditional processing and analysis of histological objects.

1. Allocation of histological digital images of objects on the difficult substantial variability and low contrast. Therefore, in problems segmentation solutions for private tasks apply a variety of methods. Universal approaches to the selection algorithm for segmentation of arbitrary images histological objects are unknown. Consequently, there is a need to build a coherent histological classification of objects and algorithms segmentation, as well as the development of algorithms segmentation more general and universal character to highlight a broad class of objects.
2. A histological classification of objects to determine the method of segmentation histological images of objects. The basis of classification based on optical and geometric characteristics histological facilities as well as optical characteristics of their environment - background. For each class of objects determined the best method of segmentation. This classification determines the universal approach to the selection algorithm for allocation of histological objects of a certain class, while maintaining the high quality of results.
3. A half-algorithm utonsheniya slabokontrastnyh images of objects on histological objects. The main distinction is an algorithm that the operation utonsheniya process begins with an image of points with his entourage for maximum brightness characteristics. This feature allows you to process and receive high-quality results on images with a complex background to the changing brightness characteristics, especially for images histological preparations.
4. An algorithm segmentation and tracking of vessels or optical fiber at high magnification. The proposed algorithm based on an analysis areas from the utonsheniya changes brightness, gives a qualitative result, suitable for further processing. Moreover, for this method of segmentation is proposed tracking method, based on the analysis also highlighted areas. These features allow to increase speed and quality processing.
5. An algorithm morphological segmentation histological polygonal objects. The algorithm identifies objects such as cells, blood vessels and fibre in cross section on the images histological preparations with the background, characteristics which are changing brightness, and texture is not expressed.
6. An algorithm segmentation method of cell areas. He focused on image processing with a complex background, which has changed brightness characteristics and the texture is present, consisting of false objects and artifacts. The method of combining areas significantly slower morphological segmentation, but it allows you to define objects, even when differences in levels of brightness are the same as that of the surrounding background. Lack of stadium "Zasevaniya, growth and division of areas leads to win in speed compared to traditional algorithms growth areas.
7. Proposed to describe the color system is intended to preserve the coloration of mathematical methods when working on the images morphology histological preparations. In presenting the image in this coordinate system the bulk of the processing takes place on one axis coordinates, showing images of half the properties. This helps improve the quality and expedite the processing of colour images histological preparations. Obtained in the dissertation work results are intended to implement automated systems in the histological analysis of drugs and may used in traditional processing and analysis of histological.

1. Atlas on histology, cytology and Embryology / SL Kuznetsov, N. Mushkambarov, VL Goryachkina. -- Moscow: MIA, 2002. -- 131 c.
2. Quick algorithms in digital image processing / Ed. Hwang GS -- Moscow: Radio and communication, 1984. -- With 221.
3. DiaMorf Lens - sophisticated software to automatically morphometry and counting objects in digital images [electronic resource]. The regime of access: http://www.diamorph.ru/aboutprog.html.
4. The logic simulation and testing of digital devices / Skobtsov YA, Skobtsov VY -- Donetsk: IAMM NASU, DonNTU, 2005. -- With 436.
5. Martynenko TV Segmentation and classification of colour images histological sections / / Obchislyuvalna техніка is автоматизація: Zb. Sciences. DonNTU Ave, випуск 107. -- Donetsk: DNTU. -- 2006. -- S. 104-110.