"Development of specialized computer systems for diagnosing diseases of urinary system"

Introduction The development of new diagnostic methods, as well as the empowerment of existing methods are an urgent task in medicine. The emergence of new diagnostic and therapeutic technologies require the use of artificial intelligence methods for data processing and interpretation with the possibility of accumulation, storage and reuse of medical data [1]. One of the most effective tools in this area are expert systems. They allow you to automate the process of decision making on the inspection of patients, raising the level of the users' qualifications to the level of experienced experts [3]. Therefore, it is necessary that expert systems have the possibility of flexible tasking, applicable to all areas of biology and medicine, had a large information capacity and noise immunity, did not need long time to develop [1]. The continuous development of computer technology expands the potential of such systems in connection with what is necessary to constantly upgrade knowledge in this field.

Expert systems for a long time used in medicine to diagnose diseases. Each system in this case has a limited scope, in connection with the initial focus of development. The application of these expert systems in areas for which they were not originally intended, it is difficult and often impossible, including the fact that they are limited built-in way O [3]. As a consequence, there are needs in the development of expert systems, which are focused and can diagnose certain diseases of human systems. In particular, urinary tract disorders.

Formulation of the problem. Diseases of the urinary tract are extremely widespread and pose a serious public health problem worldwide. Causes of urinary tract disorders, very much. Common causes include: infectious diseases, alcohol and tobacco, raw water, defective products and spiced food, chill.

The problem developed an expert system for medical diagnostics is to determine the possible diagnoses based on patient's knowledge about diseases of urinary system and data to the survey, which included the analysis of urine, blood, ultrasound of the kidneys and bladder.

This expert system is based on the conceptualization of the domain, the explicit representation of which is called ontology. We give a general description of the formal model ontology for medical diagnostics [5].

The processes occurring in a patient, can be divided into external (observed) and internal, which are the subject of diagnosis. The observed processes are called features, and internal - diseases. Signs have values that are the result of observation of these signs, and the values of features can change over time. The values of the attributes are qualitative (scalar). Signs are a subclass of observations. Another subclass of observations are anatomical and physiological characteristics of the patient's organism. The latter also have scalar values, in this paper is that these values can not change over time. The last subclass of observation are the events that occurred with the patient, who also have scalar values.

General model of the ontology is the basis of an expert system, which will handle the input data.

The method of solving the problem. Stage of processing the input information directly doctor. The result of his work are the patient data made to the database, which is developed in an environment Microsoft Access. The doctor makes the data from the on-screen form shown in Fig. 1.

The system of organization of work is at the client level, so it requires constant physician participation in the introduction and adjustments made, and edit data, in turn, allows to avoid errors in the input data.

As a result of the EC receive a diagnosis, that is a specific type of disease, if it is, of course, it was found, otherwise the doctor is invited to appoint a patient more tests.

The output can be diagnosed:

a) Pyelonephritis;

b) Urolithiasis;

a) Hydronephrosis;

g) Renal failure, etc.

The process of ES is iterative, so at first be limited to a maximum number of input variables, using only the most significant. If the hypothesis about the reliability of peer review is not confirmed, then you need to use additional input signals. This approach ensures that no redundant information in the database, which increases the reliability of the ES. Number of levels for each input value depends on the coefficient value of this quantity and the magnitude of the error of laboratory equipment. Note that the number of input variables and their levels are rising, due to the improvement of laboratory equipment and increasing the database, thereby improving the reliability of expert assessment.

Natural way to solve this problem is the enumeration of all possible values of the output data (specific diseases). For each disease is performed solving the direct problem - construction of all possible variants of causal relationships (based on information from the knowledge base, formed by experts, the values of anatomical and physiological characteristics of the patient and the values that have occurred to him the events) and search through them to those who meet all the observed values of the patient's symptoms.

The result of solving the problem will be one of two things:

• a message that the patient is healthy, indicating the reasons for the observed values of the trait;

• or a few mutually exclusive diagnoses, each of which represents one disease, which is ill patient, together with the reasons the values of attributes.

The analysis of the literature showed that the most efficient algorithm for solving the problem of diagnosis is an algorithm that uses fuzzy terms [5].

Algorithm for solving this problem is quite trivial. The main work is in the proper organization of the knowledge base.

The knowledge base is represented as:

(1)

where - Fuzzy term that calculates the value log ; Output is estimated by fuzzy term , M - the number of terms used for the linguistic assessment of output data.

The logical conclusion is based on the well-known algorithm for inference in expert systems, formalized synthesis of which is as follows.

1. Let - Membership function of input ( ) Fuzzy thermo Ie Where - The set of pairs of elements of subsets;

2. -Membership function fuzzy output from thermal ;

(2)

3. degree of membership input ( ) Fuzzy thermo from the knowledge base is defined by the following system of fuzzy logic equations:

; (3)

4. fuzzy subset Corresponding input ( ), Is defined as

(4)

where - Operation of union of fuzzy subsets;

5. clear output value - y, the corresponding input ( ) Is defined by the median:

(5)

where G - the power of fuzzy subsets.

Algorithm itself is divided into several subtasks, one of which is the main (control):

1. Check for signs of each execution of the necessary conditions.

2. If, for some of the signs is not the necessary condition, the data about the patient considered invalid and exit.

3. Test the hypothesis that the patient is healthy.

4. If the hypothesis that the patient is healthy, confirmed

4.1. is considered a result of the diagnosis, in which no disease, as well as his explanation (acquired causes of the observed values of attributes);

4.2. otherwise go through all of the disease from the knowledge base and for each:

4.2.1. verify the necessary conditions for this disease;

4.2.2. if necessary condition is satisfied, then test the hypothesis that the patient is ill with the disease that started at the beginning of observations, and in case of its confirmation to add the disease, together with his explanation (find out why the observed values of attributes) to the set of solutions;

5. Shut down, issuing the results (the diagnosis and causes the characteristic values).

Consider the following example:

Patient following a blood test:

Obschevospalitelnye changes: leukocytosis, accelerated erythrocyte sedimentation rate, leukocyte shift to the left, with strong inflammation - anemia.

1. Check for signs of a general analysis of blood, the implementation of the necessary conditions: the number of formed elements.

2. If this character is not satisfied the necessary condition (the input data should not be less than the minimum threshold and not greater than the maximum threshold), the data about the patient considered invalid and exit.

3. Test the hypothesis that the patient is healthy.

4. If the hypothesis that the patient is healthy, confirmed,

4.1. then derive the result that the patient is healthy;

4.2. otherwise the disease through all of the knowledge base and for each:

4.2.1. check the fulfillment of the necessary conditions (shift orthocytosis left) for this disease;

4.2.2. This condition is satisfied for the disease pyelonephritis. Test the hypothesis that the patient is ill with this disease and if it is added to this disease, along with his explanation (find out why the observed values of attributes) to the set of solutions;

5. Shut down, betraying the result:

Disease: pyelonephritis.

Tags: leukocytosis, accelerated erythrocyte sedimentation rate, leukocyte shift to the left, with strong inflammation - anemia.

This example using the general scheme of the algorithm with a single sign is very generalized. However, it illustrates the work of the expert system.

Summary

In this article, selected and described by the optimal algorithm for expert systems for diagnosis of diseases of urinary system, which satisfies the following requirements [5]:

1) Provides the necessary accuracy of diagnosis.

2) Allows you to analyze the symptoms and the relationship between them over a long period.

Completed description of the database and provides screen form with the database. Describe the knowledge base and a formalized synthesis inference information.

Main sub-program algorithm is shown graphically in the work with exemplary diagnosed according to the analysis of blood.

References

1. Статические и динамические экспертные системы: Учебное пособие / Э.В. Попов, И.Б. Фоминых, Е.Б.Кисель, М.Д. Шапот. — М.: Финансы и статистика, 1996. — 320 с.

2. Автоматизированное проектирование информационно-управляющих систем. Проектирование экспертных систем на основе системного моделирования / Г. Г. Куликов, А.Н. Набатов, А.В. Речкалов и др.; Уфимск. гос. авиац. техн. ун-т. — Уфа, 1999. — 223 с.

3. Куликов Г. Г., Брейкин Т.В., Арьков В.Ю. Интеллектуальные информационные системы: Учебное пособие. — Уфа: УГАТУ, 1999. — 129 с.

4. Тельнов Ю.Ф. Интеллектуальные информационные системы в экономике. — М., 1998. — 174 с.

5. Клещёв А.С., Москаленко Ф.М., Черняховская М.Ю. Модель онтологии предметной области «медицинская диагностика». Часть 2. Формальное описание причинно-следственных связей, причин значений признаков и причин заболеваний // НТИ. Сер. 2. – 2006. – № 2. – С. 19-30.

6. Черняховская М.Ю., Негода В.И., Москаленко Ф.М. Формальное представление знаний о конъюнктивитах. – Владивосток: ИАПУ ДВО РАН, 2008. – 55 с.

7. Клещев А.С., Шалфеева Е.А. Определение структурных свойств онтологий // Изв. РАН. Теория и системы управления. – 2008. – № 2. – С. 69-78.

8. Клещев А.С., Орлов В.А. Компьютерные банки знаний. Многоцелевой банк знаний // Информационные технологии. – 2006. – № 2. – С. 2-8.

9. Список банков знаний Многоцелевого банка знаний. – [Electronic resource]: http://mpkbank2.dvo.ru/mpkbank/index.php?page=banks.

10. Москаленко Ф.М. "Экспертная система медицинской диагностики, основанная на реальной онтологии медицины, для многопроцессорной ЭВМ" // Труды II между- народной конференции «Параллельные вычисления и задачи управления» PACO’2004 памяти Е.Г.Сухова. М.: Институт проблем управления им. В.А. Трапез- никова РАН, 2004. С.999–1084.