APPLICATION EXPERT SYSTEMS OF SHELL "TECHEXP"

VLADIMIR P. BURDAYEV

UANet/Relcom

E-mail: burdaev@hiuv.kharkov.ua

http://www.uacm.cit-ua.net

phone:(7-05700) 11-80-21

310176 UKRAINE KHARKOV

KHARKOV ADVANCED TRAINING INSTITUTE

FOR DOCTORS MINISTRY

OF HEALTH OF UKRAINE

KORCHAGINSTEV 58

Abstract - The system "TECHEXP" is intended for creation of expert systems, which will solve problems of interpretation and diagnostics in the following problem domains: medicine, biology, chemistry and others. The working expert systems:"CHDR", "INFARCT", "DCAMG", "HEPATITIS", the learing systems:"PEDIATRIST", "PATHPHYS" and intellectual systems of clusteranalysis "ISCAD".

Expert system, Knowledge base, Production rules, Frames, Cluster of rules

I."TECHEXP"

The expert system shell "TECHEXP" includes: an editor of knowledge base, 'blackboard', an interpreter of rules, block of explanation, subsystem of generation and acquisition of knowledge, regim of consultation. The method used in the system is decomposition the knowledge base on the clusters and a combining of productions and frames properties for knowledge representation; an advancement and verification of logical inference hypotheesis.The whole knowledge base is broken down into clusters.Each of them is activised by means of metarules. The knowledge is divided into static and dynamic. System of "TECHEXP" allows to carry out as forward as backward inference. Besides, an expert can offer his own strategy or use strategy of system. Subsystem of explanation allows to find out why the rule was fulfilled and how the fact was obtained in any moment of the consultation. Facts and rules of knowledge base are supplied with certainty factors early introduced by experts. Certainty factors can be calculated during the consultaion by means of different methods. This system allows to obtain graphically distribution of knowledge base on the clusters, routed inference and get a protocol. The system supports user-friendly interface. Technical characteristics are the following: system is used on personal computers IBM PC; programming system is TURBO PASCAL 7.0, PxEngine 2.0, Paradox 3.5, NetWare Novell 3.11; loading module size is 4Mb.

II. "CHDR"

The "CHDR" (Coronary Heart Disease Risk) expert system is designed for diagnostics and for estimation the level of the coronary heart desease risk and for generation the individual preventive recommendations. The risk factors is arranged by the Delphy expert procedure. The calculation of the relative importance coefficient of CHD risk-factors was fulfilled by the couple comparison procedure. The knowledge base is represented by 13 clusters in accordance with selected risk-factors. The cluster is the homogeneous group of rules for description the one risk-factor. The rules of knowledge base are represented in the form of productions and frames. The certainty coefficients of rules are calculated by specific procedures: bisection test, test of the polar estimations, test of the mark estimations.

III. "INFARCT"

The prototype expert system "INFARCT" helps to doctors in diagnosis of infarction patients and to evaluate their state and give prognosing of development of following complications of myocardial infarction. They are fibrillation, acute leftventricle incompctence, chronic heart incompctence, arrhythmia, thromboembolia, ruptures of myocardium, repeated infarct. Bayesian method used in solving the problem of prognosing favourable or unfavourable outcomes of a trasmural (or macrofocal) myocardial infarction based on the data covering the first 3 days of the hospital stay for patients admitted to it not later than 2 days after the onset of mayocardial infarction and who did not develop during this period extremely serious complications are described. System pulls out two hypothesis: complicated and smooth (without complicated) course myocardial infarction in a patient. Then in process of consultaion with a doctor it calculates a chance to its hypotheses on the basis of symptoms of a patient and his anamnesis. If we receive a first hypothesis, then begin analysis of calculate of the aposteriori probability, which serve basis for the diagnostic degree development corresponding complication of myocardial infarct. The knowledge are presentation in the form by the combination rules and frames. Besides, these rules describe the conditions, at which hypothesis can be received or rejected. The frames are used for collection of information reference for concrete hypothesis. The logic inference is based on Bayesian method of receiving of decision.

IV. "DCAMG"

Expert system "DCAMG" (Diagnostics of CAncer of Mammary Glands) is destined for early diagnostics of tumors of mammary glands. Diagnostics is based on knowledges of expert oncologist, which are grouped on following clusters: thermo graphy, anamnesis, physical examination, echotomography.The system allows to classify following tumors with much accurary: lipoma, fibroadenoma, fibrocystic mastopathia diffusive, fibrocystic mastopathia local, DFA (diffusive), DFA (local), mastitis and give recomendations for its treatment. Volume of base of knowledges consists of 14яяя rules of productions. Time of one performance of consultation consists 3я-45 minutes.

V. "HEPATITIS"

Expert system "HEPATITIS" is destined for diagnostics of acute and chronic diseases of liver. This system gives a possibility: to find the cause of liver's disease to limite it and to get therapeutic effect; to precribe medicaments for treating of liver`s diseases; carry out statictical valuation of therapeutic remegies to the patients. In problem domain four clusters are distinguished: clinical data, labratory data, morphologic data, scintigraphic and roentgenologic data, which give an apportunity to put a detail diagnosis of liver`s disease: The first cluster of knowledge users data of anamnesis and complaints of a patient about liver; results of palpation and percussion of liver and also examination of this patient. The second cluster of knowledge gives an opportunity to determine a type of biliousness ana also a character of liver`s disease on the base of labratory-chemical changes of liver`s function, distruction liver`s cll, holestasis, activity of mesenchyme and other examinations. The third cluster gives an opportunity to determine indications for morphologic diagnostics. For a example it is given an information on that it is needed to do biopsy of liver or it is not needed. It is given also indications or contraindi cations for laparoscopy. The fourth cluster of knowledge is used very seldom, because first three clusters give an opportinity to give detail diagnosis. But if it is necessary the system recommends to carry out an additional technical investigations: scintigraphy of liver, sonography of liver, roentgenology of liver for definition of diagnosis.

VI. "PEDIATRIST"

Controling and learing expert system "PEDIATRIST" is destined for testing of professional knowledges of a doctorpediatrist and for determining of his qualificational stage. Expert valuation of doctor knowledge is produced on following directions: gastroenterology, hematology, pulmonology, cardiorheumatology, nephrology, neonatology, physiopathology of newborns. The scale of twelve` mark is used for valuation of knowledges. The system allows to determine qualificational stage of doctor on the following scale: doctor-intern, doctor-stager, doctor-specialist (II category), leading doctor (I category), doctor of higher qualification (higher category). Qualificational stage is formed on the base of following factors: valuation of qualimetry of knowledges, working characteristics of a doctor, results of work of a doctor and experience of work. Time of one performance of consultation consists of 9я minutes (time for answer on the question - 45 seconds).

VII. "PATHPHYS"

Controling and learing expert system "PATHPHYS" (PATHological PHYSiology) is destined for valuation of knowledges of students and their learing on following devisions of pathological physiology: circulation, white blood, red blood. The scale of ten`s marks is used for valuation of knowledges of students regime: learing and testing. In regime of learing a user prompt of the system for selection of right answers to given question. System proposes also some recognized and situational problems. For the problem of recognition a user may select right answers, belonging to one or another alternatives which are patin the question. For situational problem it is necessary to fulfill analyse of selecting tests of pathology of blood.

VII. "ISCAD"

The "ISCAD" is Intellectual System of Cluster Analysis Data. It is intended for solving problems of diagnostics and automated classification of multidimensional objects using expert`s knowledge. The system allows to select both non-fuzzy and fuzzy clusters in interactive, intellectual (based on the knowledge of experts), automatic modes. The knowledge base on cluster analysis embraces usage of a number of pattern algorithms: iteration algorithm, c-means algorithm, algorithm ISODATA, agglomerative hierarchial algorithm and others. The process of clustering is provided with graphic pictures. The use of knowledge base in cluster analysis gives the user a certain degree of confidence in the classification algorithm offered by the system would be correct. For representation of domain of knowledge the frames and the rules of the production type are used. The deductive inference is used as an algorithm of an ordered goal choice.

P.S.

I send my manuscript for establish contacts and collaborate with ALL (university, hospital, ...).