authors: doctor of science, professor a.kolesnikov, candidate of sciences, associate professor...

13
KNOWLEDGE-BASED SYSTEM FOR DECISION MAKING SUPPORT AT DIAGNOSING OF THE ARTERIAL HYPERTENSION Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

Upload: hilary-evans

Post on 18-Dec-2015

218 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

KNOWLEDGE-BASED SYSTEM FOR DECISION MAKING SUPPORT AT DIAGNOSING OF THE ARTERIAL

HYPERTENSION

Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov,S.Rumovskaya

Page 2: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

AH – arterial hypertension; AP – arterial pressure; SA DDAH - The System of Automated

Differential Diagnostics of the Arterial Hypertension

Abbreviations

2

Page 3: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

The task of the arterial hypertension diagnostics is ill-structured, because of:

plurality of alternatives;the cause-effect relation ‘symptoms — the diagnosis’ cannot be written on mathematical language;depending on presence of a symptom the probability of concrete disease can raise or be reduced that is considered in certainty factor;generally qualitative character of the symptoms;in some cases there is absence of the information about cardiovascular diseases.

3

Page 4: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

Stages of the patients’ diagnostics and inspection with an arterial

hypertension

definition of stability and a degree of the AP raising;an exception of a symptomatic arterial hypertension or identification of its form;definition of the common cardiovascular risk;revealing of other risk factors of the cardiovascular diseases and clinical states;definition for the patient’s risk group;diagnostics of defeats of the target organs.

4

Page 5: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

The process of the diagnosing of the AH with usage of SA DDAH

Signs. SF - symbol format, NF – numerical format, AP – arterial pressure, ECG – electrocardiogram

History taking

Physical examination

Laboratory examination

SA DDAH

Confirmationof the AP raising

The start of check-up

Lab A patient

A doctor

Measurement of blood pressure

AP data (NF)

Data input

Processing AP data (the results in SF)

Control over 1 or 2 years

Systolic AH

The beginning of the diagnosing

Questioningpatient

Symptoms, his deseases and in his family (particularly SF)

Anamnesis data input

Processing of the input data (the results in SF)

Examina-tion, palpation of the patient

Input of the results of physical exam

Guidelines under analyses (SF)

The results (particularly SF)

Processing (the results in SF)

Examined

Blood and urine tests - NF echo, ECG, X-ray picture – SF and NF

Data input

The final diagnosis (SF)

Expert knowledge

Tonometer

5

Page 6: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

questions for a history taken and analyzing of the medical report;

questions for fixing the results of doctor’s objective examination (for more information);

questions for fixing the results of special diagnostic procedures, laboratory and tool methods (a necessity in them exists there for an exact ascertainment or diagnosis confirmation)

Questions stored in the IBM6

Page 7: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

Unification of the methods for an arterial hypertension and demonstrating a complete disease picture.

The necessity of the efficiency of decision-making, training students to make diagnoses and working off their skills of the definition of what data is necessary for making diagnosis, and searching for this data (as automated system proves in detail its outputs).

The development of the knowledge field and of the doctor-cardiologist’s base of the facts.

The development of the knowledge base. The development of the mechanism of an explanation of the diagnosis made by the

system. The development of the system laboratory prototype. The system testing.

Tasks, which were solved during the development of the automated system

7

Page 8: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

The field of knowledgeField of knowledge — conditional informal description of the main concepts and interrelations between concepts of the subject domain, revealed from the expert knowledge system, in the form of the diagram, the table or the professional text. The field of knowledge represents the model of problem domain knowledge.

As the elementary pragmatical approach to a creation of the field of knowledge the following sequence of operations was used:

definition input and output data; compilation of the terms dictionary and sets of key concepts; revealing links between concepts; construction of a pyramid of knowledge — a scale of ranks of concepts;

definition of relations (as well as relations between concepts inside each level of a pyramid and between concepts of different levels).

The degree of the knowledge field is: concepts – 415, relations — 15.

8

Page 9: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

The base of the facts

Root

GlobalNamePassMenu8

DDE

Image

KWindow

Pacient

Diagnoz

AnamnezFizikalLabobsledAD

Giperton

RiskUrovenADRecommend

VtorAG

FR

POM

ASK

Hierarchy of the base of facts in the system

Objects of the subclass VtorAG — are the diseases found of the patient:

FR (risk factors);POM (lesion of the target organs);ASK (associative clinical state).

Each object of the subclasses Diagnoz и VtorAG has also features, in which slots contain certainty factors of all components.

Slots of Anamnez, Fizikal, Labobsled and AD contain input data for solving of the task of the diagnosing arterial hypertension.

Objects of the subclass Diagnoz — are the component parts of the diagnosis generating as the result of examing:

Giperton (a stage of the hypertension);Risk (general cardiovascular risk);UrovenAD (the level of the AP);Recommend (recommends about the laboratory examination and the necessity of the following examination);VtorAG (the secondary hypertension).

9

Class

Objects

The size of the system base of the facts makes about 350 variables (pairs ‘object: slot’).

Page 10: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

Certainty factorCertainty factor is calculated on the base of Karnap’s confirmation theory of two evaluations: the belief measure MB(H,E) and disbelief measure MD(H, E).The belief and disbelief measures are fixed in the terms of prior and conditional probability (1), (2):

1, if ( ) 1;

max ( | ), ( ) ( )( , ), if not.

max 1,0 ( )

P H

P H E P H P HMB H E

P H

1, if ( ) 0;

min ( | ), ( ) ( )( , ), if not.

min 1,0 ( )

P H

P H E P H P HMD H E

P H

The function point out the growing up of the belief measure to the hypothesis H, if the fact Е have taken place.

The function point out the growing up of the disbelief measure to the hypothesis H, if the fact Е have taken place.

where: p(H) – prior probability, that hypothesis H is true; p(H|E) – conditional probability, that hypothesis H is true at the fact E having taken place;

The certainty factor, CF(H,E), is calculated in the system, proceeding from values of belief and disbelief measures (3):

(1)

(2)

, 0;1 and , 0;1 .MB H E MD H E

( , ) ( , ) ( , )CF H E MB H E MD H E (3)

And is located between -1 and +1, defining the belief measure to the hypothesis H.

10

Page 11: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

The knowledge baseThe most general view production rule:

11

( ); ; ; ; ,i Q P A B Nwhere:i — the identifier of the production;Q — the application area of rule, namely ‘islands of knowledge’;A→B — a production kernel - if the logical expression accepts value TRUE, then conclusion becomes TRUE, too, that is mean probability of symbol conversion, defined by a kernel, or some operation;P — a condition of applicability of a production kernel, often expressed by a predicate. A condition of applicability and premis of the production’s kernel A are written as one expression and, essentially, make the symbol pattern in our automated system;N — a postcondition executable only after implementation of a kernel of production and represents our changes in a chain of reasoning which should be made in the knowledge base, is not used in our automated system.

(4)

( 2); 1;

' 120 AND 130 AND 80 AND 85’ 2 AND ;

Re 'The control over two years!’.

confirAG Rulelist

SAP SAP DAP DAP god TRUE Unorm TRUE

ccom

(5)

‘Islands of knowledge’

Rulelist1 RulelistA RulelistF RulelistL RulelistOG RulelistReas RulelistKD

Rulelist1 contains rules analyzing AP information. RulelistA contains rules for analyzing the information, gathered on the stage History taking. RulelistF contains rules for analyzing the information, gathered on the stage of the physical examination. RulelistL contains rules for analyzing the information, gathered on the stage of the laboratory examination. RulelistOG contains rules for definition of the cardiovascular degree and the stage of hypertension. RulelistReas contains rules for definition of the symptomatic AH. RulelistKD contains rules for estimation of the diagnosis certainty factor.

Page 12: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

Graphic expression of the production rules by graphs AND-OR

god2= TRUE

Unorm= TRUE

SAP>120SAP<=130 DAP>80 DAP><=85

AND AND

OR OR

Signs. node of graph — node of the type AND (conjunction of the facts signed by arc); SAP — systolic AP, DAP — diastolic AP, god2 — checking the AP in two years, Unorm – normal level of the SH.

Informatively this is the rule: ‘If 120 < ‘systolic pressure’ <130 and 80 < ‘diastolic pressure’ <85 then ‘level of arterial hypertension’ is normal and it is also recommended to check the arterial pressure in two years’.

In the SA DDAH:the reasoning strategy is forward chain;reasoning during the solution of the task of diagnosing arterial hypertension — are ‘wide and deep’;strategies for the resolution of rule conflicts — ‘a selective estimation’ .

Subgraph of the graph of an island of knowledge ‘Analising AP information’

12

Page 13: Authors: Doctor of Science, Professor A.Kolesnikov, Candidate of Sciences, associate professor I.Kirikov, S.Rumovskaya

Laboratory prototype and the results of the system testing

A laboratory prototype — is a system solving all demanded problems, but not steady in operation and not completely checked up. The laboratory prototype includes two more important units — user interface and the mechanism of explanation results of a solution task.

The user interface consists of eight main windows: confirmation of lifting the arterial pressure (windows for information input)information input during the history takinginformation input during the physical examinationinformation input during the laboratory examinationinput of results of the analysis of bloodinput of results of the analysis of urine)one output window of results and their correction (results of examination of the patient)one window for operation with library of reports and patient cards.

Resume by the testing of the system:By the results of testing the island of knowledge ‘the information analysis on physical examinations’ and guidelines

to laboratory inspection have been corrected. Also the information in remarks of rules has been enlarged for rising a transparency of the diagnosis and improvement of the mechanism of an explanation.

On termination of testing the expert has given the conclusion about the quality of the automated system. The opinion of the doctor-expert is the following: ‘the system is ready to experimental application for students trainings, the interface is clear; the high performance of reception of the results (information processing does not exceed 0,5 seconds) and their validity was pleasant’.

On the basis of the received results has been concluded that the system gives sufficient and detailed diagnosis at the patient’s examination with suspicion on an arterial hypertension.

13