DonNTU   Masters' portal

Abstract

Contents

Introduction

Now in medicine special value gets the direction connected with decrease in perinatal mortality. The period beginning with the 28th week of pre-natal development when the mass of a fruit reaches 1000 and more, and proceeding to the 8th day (168 h) life of the newborn is called as the perinatal period. At all the relative short duration the perinatal period is the major stage in human life, as mortality during this period same as mortality at the age of the person from 8 days to 40 years, and danger of heavy neurologic violations during this period even exceeds that in the next decades human lives.

The most effective way in decrease in death lies in development of programs of forecasting. Complexity of their development consists in need of the scientific analysis of a large number of clinical and laboratory indicators which are in difficult dependence from each other and not always give in to a quantitative assessment. Therefore not less important is the problem of determination of risk factors as the analysis of all available information, as a rule, causes essential difficulties during the developing and realization of methods of forecasting at creation of analytical system.

1. Actuality of theme

The last decades XX and the beginning of the XXI century were marked by an accurate tendency of decrease in the main indicators, both in a demography, and in obstetric aid practically around the world. Against decrease in intensity of an increase in population indicators of maternal, perinatal, infantile and child mortality increase.

Possibility of forecasting of perinatal risk on the eve of childbirth will allow to reduce in due time maternal incidence and mortality of newborns, and also to render adequate therapy with involvement of highly skilled obstetricians-gynecologists.

2. The purpose and objectives of the study

The purpose of this work is to receive useful information from a set of parameters (risk factors) and to develop analytical system for forecasting of perinatal risk.

For development of analytical system of forecasting of perinatal risk it is necessary to solve the following problems:

  1. Development of structure of system.
  2. Development of methods of forecasting on the basis of GMDH.
  3. Realization of methods of forecasting on the basis of GMDH.

Object of research: process of design and development of methods of extraction of knowledge for medical analytical systems.

Article of research: methods of development of extraction of knowledge for analytical system of forecasting of perinatal risk.

3. Scientific novelty

Scientific novelty consists in application of a method of the group accounting of arguments for selection of optimum regression model, and also the accompanying problem of selection of the risk factors influencing perinatal risk at women is solved.

4. Development of algorithm of selection of optimal regressive model by means of method of group account of arguments

In most cases, in medical tasks, the result of forecasting depends on a large number of unequal factors on the importance which besides can be interconnected. This fact considerably complicates a stage of selection of data, excepting opportunity to use the most part of known methods.

However allocation of risk factors is not the only task as it is necessary to estimate a role of each of them. It follows from this that the importance of each factor on risk of development of various obstetric complications will be various [4]. Nevertheless, selection of risk factors is one of the most important stages of creation of predicting model and substantially defines its quality. Thus, at creation of optimum model selection of factors of perinatal risk is feasible also, from parameters originally offered by doctors, thus we will consider the variables interconnected among themselves.

Full search of regression models, even within the set basic function, at rather big set of input parameters in practice it isn't possible to realize. For enough complex challenges of modeling multirow algorithms of the method of the group accounting of arguments (GMDH) [3] are applied. The multirow algorithm of GMDH excludes from search some models thanks to existence porogov.soby variables.

Previously in multirow (threshold) algorithm of MGUA on an entrance some vector of entrance variables x = x1, x2,..., xn. On the first row of selection "private descriptions" (1) – (3), uniting entrance variables on two are formed:

Formula 1 (1)
Formula 2 (2)

...

Formula 3 (3)

Some number of models of the most satisfying gets out of them to external criterion of selection. In our case as such criterion there will be a mean square mistake (4) on test data.

Формула 4 (4)

where M – quantity of training examples, F – the received result, Y – the valid result.

On the second row "private descriptions" second row are formed:

Formula 5 (5)

...

Formula 6 (6)

...

Formula 7 (7)

A quantity of the best also gets out of them for use in the following, the third row, etc. For each row there is the best (by criterion of selection) model (fig. 1). Ranks of selection are increased while the assessment of criterion decreases ("the rule останова"). On the last row the best model will be optimum. Coefficients in regression models pay off the method of the smallest squares (MSS).

multirowed GMDH

Figure 1 – Multirowed GMDH

Conclusion

The actual problem of selection of optimum regression model, for forecasting of perinatal risk, including a problem of selection of risk factors was considered. Further it is planned to realize the considered mathematical apparatus and to test on the real medical data provided by staff of the center of motherhood and the childhood. Also development and deployment of system of forecasting of perinatal risk is planned.

This master's work is not completed yet. Final completion: December 2013. The full text of the work and materials on the topic can be obtained from the author or his head after this date.

References

  1. Радзинский В.Е. Акушерский риск. Максимум информации минимум опасности для матери и младенца. / В.Е. Радзинский, С.А. Князев, И.Н. Костин. – Изд.: Эксмо. – 2009 г. – С. 285
  2. Т.А. Васяева Применение метода группового учета аргументов для отбора оптимальной регрессионной модели прогнозирования потери крови при родах// Т.А. Васяева/ Вестник Херсонского национального технического университета. – 2012. – № 1(44). – С. 374.
  3. Ивахненко А.Г. Самоорганизация прогнозирующих моделей / Ивахненко А.Г., Мюллер И.А. – К.: Техника, 1985. – 223 с.
  4. Т.А. Васяева Анализ методов отбора факторов риска развития патологий в акушерстве и гинекологии / Т.А. Васяева, Д.Е. Иванов, И.В. Соков, А.С. Сокова // Збірка матеріалів ІІ Всеукраїнської науково-технічної конференції студентів, аспірантів та молодих вчених.ІУС КМ-2011 11–13 квітня 2011р., Донецьк: ДонНТУ, 2011. – № 1. – С. 209–212.