Figure 1 Method of Green

The general scheme of this method involves the following steps:

1) selection and refinement of biological products;

2) establishment of a nutrient medium for keratinocytes on a special mattress;

3) formation of a layer in a few days;

4) assess the current status;

5) detachment of the finished layer of the mattress and putting it on the gauze with paraffin;

6) transplanted to the wound and treatment.

The general formulation of the problem

As can be seen from the general scheme, restoration of skin transplants grown layers of keratinocytes is a complex technology. One reason for the failure of the epithelial layers of organ transplants is the lack of willingness to plastic layers or late transplantation. Optimum is a layer consisting of 8-12 layers of cells [2]. When transplanting immature layer, the surface of culture flask overgrown cells evenly.

Figure 2 Stages of keratinocyte growth

In different parts of the reservoir has different number of layers of cells. The effectiveness of such transplants is low. Transplantation of overripe layer disturbed nutrition keratinocytes of the basal layer. There are also other factors affecting the quality of the material during growth: a violation of temperature, abnormal concentration dispazy etc. Regular monitoring of the viability of cultivated layer, in a timely manner will help to see possible deviations and to take a decision to keep the quality of the reservoir or a fragment seeding mattress new culture. The problem is that the untimely release of mattresses obviously defective cultures leads to the loss of significant financial resources, and most importantly time, which is often a deciding factor in the treatment of patients with extensive lesions of the skin. Material for transplantation, which can be done in store, often needed in large quantities and at short notice.

Problem Statement

An important step for a successful operation is to determine the viability of the material. Existing approaches to assess the viability of cells in the majority of cells involves the processing of chemicals to identify the varying degrees of intensity of a specific color and allow to assess the number and location of test substances in the cells. The disadvantages of these methods is referred time-consuming and subjective results. In recent years intensive development of immunological methods for analyzing cellular activity. Cells were treated with monoclonal antibodies and counted the number of cells interacted with them. However, the loss of information about the location - allocation of reservoir elements under study. The methods have a common drawback - they require the provision of physical effects on cells. Cells were deprived of their environment are staying extra load, stress and the effects are generally negative. To avoid such consequences, is searching for new methods of determining cell viability using computer technology, namely image processing layers. The idea of ​​using images is not new [3], the novelty will be to build new processing algorithms and classification of images, as well as new ways of making decisions about keeping or replacing the reservoir. The paper was tasked with creating a system of monitoring the viability of cultured keratinocytes, as well as predicting the timing of final maturation, a key element of which is classified images to determine the viability of the current layer of keratinocytes. The inputs to the system are the pictures of mattresses with cells obtained with a microscope. As a result the program will give information about the viability of cultures: the percentage of living cells on a mattress and a proposal for further cultivation or sowing of new crops. After the image size of 512x512 pixels, must be received by its characteristic symptoms, which may be used for classification. As stated in [1], to obtain a sign with an image histogram or standard methods of contouring is not possible, because the brightness of the image cell is almost equal to the brightness of its background. Therefore, the main challenge when implementing a quality control system of crop choice is the mathematical apparatus of the formation of character belonging to the cell is alive or dead.

Solution of the problem

The image layer of keratinocytes has a specific texture.

Figure 3 has been formed a layer of keratinocytes

In addition, when imaging is present influence of the operator, who can deploy a microscope and install it on a different scale. Therefore, classifying characteristics should be invariant to rotation and scale change. Similar problems were considered in [3], the study of fibroblasts. To address the problems associated with the rotation and scale, will be the original image of a monolayer of keratinocytes to convert the log-polar image to eliminate the effects of rotation. The resulting log-polar image will be of tilt-invariant and nearly scale-invariant. The polar form of p (a, r) ​​of the NxN image f (x, y) is computed by (1). Then calculated the log-polar image for a given NxN image (2):

 (1)
where а=0,…,S-1; r=0,…,[N/2]-1; S=R=N.
  (2) 
where i=0,…,S-1; j=0,…,R-1.

Log-polar image is obtained by shifting a number of the original, with the brightness values ​​of the original pixels remain the same, but change the coordinates of their location. Log-polar image textures with different angles of rotation, and scale appear to have only a shift along the rows, when compared with the original, turning the log-polar structure. Next, the resulting log-polar image undergoes the wavelet transform [6], using a pair of quadrature mirror filters (high-and low-frequency). This wavelet transform differs from the standard two-dimensional wavelet packet transform that is adaptive and is invariant to a shift in the series (obtained when you create a log-polar image). As described in [6], the main drawback of two-dimensional discrete wavelet packet transform is dependent on the shift input signal through its bi-directional structure, as well as sensitivity to rotation. Feature of wavelet decomposition is used to calculate the excess number of sets of coefficients to obtain invariance to rotation and scale. First, you need to calculate the maximum number of levels of decomposition, an image NxN (3):
  (3)
For each level, calculated as four periodic images with no shift, according to the formulas for calculating the standard two-dimensional discrete wavelet packet transform, ,,,. To obtain the invariance of the series to shift another 4 are calculated periodic images (4), (5) (6), (7), each with a shift in a row:
(4)
(5)
(6)
(7)
where i=0,..,N/2p+1-1; j=0,..,M/2p+1-1;
g (m) and h (n) - a pair of quadrature mirror filter and high-frequency, respectively;
- Set the gray levels of images obtained by the log-polar transformation.
To improve the efficiency of the wavelet transform are chosen certain ranges decomposed image for decomposition in the future, instead of decomposing each image. The best basis representation is obtained by an efficient recursive selection process that determines the best decomposition of the image based on local minimization of the value function of information. The function of the price information in this paper, selected Shannon entropy (8):
(8)
Wavelet coefficients generated by the rotation invariant and nearly invariant to scale. However, a large number of wavelet coefficients is not suitable for stable texture classification. Reduction factors are calculated the energy signature for each sub-zone. Thus, the number of signatures equal to the number of energy sub-zones generated by the adaptive wavelet packet transform shift of the series. The number of signatures can vary depending on the mass of useful information contained in them. Thus, as a result of the log-polar and wavelet packet transform that is invariant to rotation and scale, obtained pre-set amount of energy signatures that characterize the image. We plan to use two energy signatures - the mean value (9) and standard deviation (10).
(9)
(10)
 To solve the problem of texture classification of images of a monolayer of fibroblasts selected classifier Mahalanobis (11) and a method of comparing Euclidean distances (12).
(11)
(12)
where x-classified vector, vi - mean vector for the reference class.

Effectiveness of the proposed log-polar wavelet signatures for texture classification of a monolayer of cells, as well as classification methods tested were verified by other experts in the study of fibroblasts.

Conclusion

The analysis methods were selected tentative assessment of the viability of keratinocytes. Promising methods of processing and image classification layers of culture that allow them to assess the viability of no physical effects on cells, namely the log-polar wavelet transforms laid and to create an image invariance to rotation and scale, as well as the Mahalanobis classifier and a method of comparing the Euclidean distance for texture classification. In the future we plan to conduct computer experiments with the standard images of keratinocytes, in order to test the effectiveness of selected methods.