Maia Mikeladze

The task of medical data analysis is considered in order to identify frequent symptoms and symptom sets for each disease. For data analysis, the association rule mining method is used, which allows to assess the specificity of the symptoms found and the confidence of the diagnostic rules based on the identified associations. The proposed method was used to process the clinical data of a number of endocrinological diseases. Based on the obtained association rules, a knowledge base of the medical intelligent system designed for diagnosing endocrinological diseases was formed.

An intellectual system for diagnosing diseases, using medical knowledge and capable of learning, is considered. As a knowledge model, a semantic network is used, also presented in the form of a matrix of relations. Both the network and the matrix define the relationship between symptoms and their causes. The weights are assigned to the arcs of the network and the corresponding matrix. These coefficients are formed in the learning process. To make a decision on establishing a diagnosis, the formation of a linear function and a block for choosing a maximum, as well as the theory of fuzzy sets and the Jager method are used. The use of this model is shown on the example of solving the problem of diagnosing primary headaches.

The task of medical data analysis as one of the data mining tasks is considered. As methods of solving it, methods of machine learning are offered - a logical method of classification based on a conceptual approach and feedforward neural networks. To train the network, a learning algorithm with error correction is used. Based on the results of the application of these methods, the intelligent system for diagnosing and predicting primary headaches are constructed. This intelligent system is used to support decision-making by a doctor.

A huge amount of heterogeneous data generated by modern medicine, require the development of new technologies for data analysis in order to discover certain patterns and hidden knowledge. Methods of data mining provide new opportunities for solving of this actual task.

In the paper features of the medical data and the main tasks associated with their analysis and processing are considered. A brief overview of data mining methods an algorithms, most often used in medicine, are given; their advantages and disadvantages are considered. Data mining applications for solving various problems of medical diagnosis and prediction, as well as the most actual directions of development of data mining technologies in medicine are given.

The problem of data processing in the tasks of medical diagnostics is considered as one of the problems of data mining. As a data mining method, a logical recognition method based on a conceptual approach is proposed, as well as a quantitative feature binarization algorithm that ensures the effectiveness of the classifier rule in terms of both diagnostic and differentiating abilities.

On the basis of the conceptual approach the methods of solving the non-formalized tasks of diagnosis the complex systems state are proposed. Diagnosing of a system state is considered as a problem of classification and recognition, and creation of the generalized description of classes is carried out by training on examples. The proposed methods provide reference of new states to a certain class (extrapolation) both for quantitative, and for qualitative features. On the basis of these methods non- formalized task of differential diagnostics of migraine and a tension headache was solved.

The question of the knowledge organization in intelligent information systems for solving unformalized problems of diagnosing is considered. The method of diagnosing the state of complex systems which works with both quantitative and qualitative data is offered. The descriptions of classes of states resulting from the application of this method are high-informative predicates – a disjunction of conjunctions, each of which is characterized by the certainty factor. It allows to use production model with certainty factors for representation of heuristic knowledge which effectively works in the conditions of fuzziness, uncertainty and incompleteness of data and knowledge.

The problem of an assessment of an ecological state of environment is considered. It is shown that information received as a result of diagnostics of an ecological state of environment, is insufficient for its adequate assessment. Information received on the basis of forecasting and the causal analysis of an ecological state of the environment is necessary for the solution of a task also. The model of intellectual system using methods of mathematical logic, Boolean algebra and the discriminant analysis is offered.

On the basis of the conceptual approach to the methods of solving the non-formalized tasks of diagnosis of the complex systems are proposed. Diagnostics of the condition of a system is considered as a problem of classification and recognition and creation of the generalized description of classe is carried out by training on examples. The proposed methods provide reference of new states to a certain class (extrapolation) both for quantitative and qualitative values of features.On the basis of these methods non-formalized task of differential diagnostics of migraine and a tension headache was solved.

In the description of the patient, like any multi-system, we are confronting not only with abundance, but also with a large variety of factors (attributes, symptoms). In such a situation, the challenge becomes: how all this variety of features should be reduced to one type without information losing. Proposed algorithm helps to achieve this objective. The algorithm of data unification that allowing reducing all variety of exact and vague factors to one type without the loss of information is offered for that problem solution.

The description of the knowledge-based system, which enables to reveal intuitive knowledge of an expert, that is hard to formalize and to present it in the form of productive rules thus giving opportunity for estimating the state of economic system as «satisfactory» or «unsatisfactory» is given in the present report. The submission of properties of differentiation for essential values of factors provides high precision of sifting «unwanted» state of the system by means of these rules. Furthermore, there is a procedure of clarifying insufficient data on the basis of hierarchy of concept, the connection of «general concept-specification» and fuzzy sets.

Doctoral Thesis Referee

Master Theses Supervisor

Doctoral Thesis Supervisor/Co-supervisor

Scientific editor of monographs in foreign languages

Scientific editor of a monograph in Georgian

Editor-in-Chief of a peer-reviewed or professional journal / proceedings

Review of a scientific professional journal / proceedings

Member of the editorial board of a peer-reviewed scientific or professional journal / proceedings

Participation in a project / grant funded by an international organization

Participation in a project / grant funded from the state budget

Patent authorship

Membership of the Georgian National Academy of Science or Georgian Academy of Agricultural Sciences

Membership of an international professional organization

Membership of the Conference Organizing / Program Committee

National Award / Sectoral Award, Order, Medal, etc.

Honorary title

Monograph

Handbook

Research articles in high impact factor and local Scientific Journals

Publication in Scientific Conference Proceedings Indexed in Web of Science and Scopus