Literature Review 7 pages

Introduction

Studies by researchers of different scientific schools

have shown that reflex diagnosis may provide an approach

to the prognostication and early diagnosis of diseases,

including diseases of the heart and cardiovascular system;

these are based on measurement of the electrical proper�

ties of biologically active points (BAP) [1�5]. One advan�

tage of reflex diagnosis methods is that the responses of

biologically active points to changes in the internal struc�

tures of the body occur before clinical symptoms of dis�

ease are manifest. This provides for the detection and

treatment of diseases at their earliest stages of develop�

ment and, sometimes, their prevention by application of

prophylactic measures. The medical�technical and time

costs of organizing investigations are significantly lower

than those of conventional approaches [2, 6].

In tasks in which the information value of BAP is

insufficient for the accuracy required, this can be

obtained using various signs used in conventional medi�

cine aggregated into a final decision rule using hybrid

fuzzy models, the general theory of whose construction

has been laid out in [3, 7, 8].

The total cost of achieving the necessary prognostic

and diagnostic accuracy is generally lower than when

standard study methods are used.

Methods

One of the most important tasks in constructing

decision�making models is that of selecting informative

properties, particularly informative BAP.

The procedure of selecting informative BAP will be

more effective if the characteristics of the production of

information on the state of the body’s internal structures

at these points are considered. These characteristics

include the production of large amounts of information at

one point (multiple diagnoses, symptoms, syndromes);

cyclic changes in the energy status of BAP during the day

even in conditions of normal energy balance along merid�

ians; the large volume of data required for analysis if the

Biomedical Engineering, Vol. 52, No. 1, May, 2018, pp. 68�71. Translated from Meditsinskaya Tekhnika, Vol. 52, No. 1, Jan.�Feb., 2018, pp. 50�53.

Original article submitted June 26, 2017.

68 0006�3398/18/5201�0068 © 2018 Springer Science+Business Media, LLC

1 Southwest State University, Kursk, Russia; E�mail: kstu�[email protected]

2 Kursk State University, Kursk, Russia.

3 Kursk State Medical University, Kursk, Russia.

4 Belgorod State University, Belgorod, Russia.

5 National Research Nuclear University MEPhI, Moscow, Russia.

6 Peoples’ Friendship University of Russia, Moscow, Russia.

* To whom correspondence should be addressed.

Prognosis of Recurrent Myocardial Infarction Based on Shortliffe Fuzzy Models Using the Electrical Characteristics of Biologically Active Points

N. A. Korenevskiy1*, S. P. Seregin1, V. A. Ivanov2, A. I. Kolesnik3, G. V. Siplivy3, K. F. Makkonen4, V. V. Dmitrieva5, D. I. Kicha6, and D. A. Zubarev1

This report discusses the employment of the electrical characteristics of biologically active points (BAP) in the

meridian of the heart to solve the task of prognosticating recurrence of myocardial infarctions during the reha�

bilitation period using fuzzy decision rules. Employment of only the electrical characteristics of informative BAP

was found to provide a reliability level of greater than 0.85 in decision�making, with a significant uncertainty

zone. If additional features characterizing long�term psychoemotional tension − lipid peroxidation and antioxi� dant activity − are included, the reliability of prognoses of recurrence of myocardial infarction during remission reaches 0.95.

DOI 10.1007/s10527-018-9784-1

Prognosis of Recurrent Myocardial Infarction Based on Shortliffe Fuzzy Models 69

pathology is not already known [1�3]. These properties of

the information available at a BAP hinder the procedure

of selecting the informative points in studies using con�

ventional methods of information theory and image

recognition theory.

Considering the characteristics of the presentation of

information on the state of the human body at biological�

ly active points, reports [2, 3] propose methods and algo�

rithms for finding combinations of BAP whose analysis

confirms the situation being studied (the diagnosis) and

rejects “interfering” situations “displayed” at BAP on the

basis of reference data but not present in the subject.

These combinations are termed diagnostically significant

points (DSP) [2], 3].

Specially conducted studies on the prognostication

and early and differential diagnosis of diseases of the car�

diovascular system, gastrointestinal tract, nervous sys�

tem, musculoskeletal system, respiratory system, and

others have shown that the use of DSP in combination

with other informative properties allows decision rules

providing high classification quality to be developed [2,

3].

Considering the recommendations in [2, 5], the

energy characteristics of BAP were taken as their electri�

cal resistance measured using a 1�kHz alternating current

at a current intensity of no greater than 10 μA [2, 5]. These studies showed that interval assessment of resist�

ance ΔRj,k giving the upper and lower boundaries of resist� ance for interval k for point j and the relative deviations of

ongoing values of the resistance of BAP Rj from nominal

values are highly informative.

Experience in solving a variety of prognostication

and diagnostic tasks using information on the energy sta�

tus of BAP has shown that the properties measured are

incomplete and unclear in relation to the tasks being

addressed. In these conditions, following recommenda�

tions in [3, 7, 8] led to determination of a fuzzy logic for

decision�making and, in particular, the authors’ develop�

ment of a method of constructing sets of hybrid fuzzy

decision rules.

In this study, working with the recommendations in

[3, 8], the prognostication task was evaluated as a classi�

fication task with two classes: 1) the subject is not going to

fall ill in the next T years (class ω0); 2) the subject will fall ill in the next T years (class ωl).

For selection of informative biologically active points

and other informative parameters, highly qualified

experts worked on those which, when recorded, increased

confidence in the validity of the test hypothesis (disease,

ωl). In these conditions, according to recommendations in [3, 4, 9], the starting formula for calculating the confi�

dence coefficient (CC) of decision�making in relation to

hypothesis ωl should be the Shortliffe cumulative iterative formula:

CCωl(i + 1) = CCωl(i) + CC*ωl(xi+l)[1 − CCωl(i)], (1)

where CCωl(i) is the coefficient of confidence in hypothe�

sis ωl with the condition that verification of the hypothesis has already involved i informative properties; CC*ωl(xi+l) is

the coefficient of confidence in diagnosis (prognosis) ωl using only one property xi+l; CCωl(1) = CC*ωl(x1).

In a number of applications, including the task

addressed in the present study, the coefficient of confi�

dence can conveniently be calculated using the member�

ship function μωl(xi+l) for the class ωl calculated for the starting variable (xi+l) [2�4].

The prognostic confidence in the values of the inter�

val assessments (ranges of resistances) ΔRj,k (where j is the number of the BAP and k is the number of the range ΔRj) is the basis forming the diagnostic assessment and is cal�

culated by transformation of Eq. (1) into

CCωl( j + 1) = CCωl( j) + CC*ωl(ΔRj+l,k)[1 − CCωl( j)], (2)

The technical means for recording BAP resistance

consisted of a computer device exchanging data with the

PC via Bluetooth. The device operates in two modes:

search mode using sound, light, and graphical recording

of BAP and measuring mode at a frequency of 1 kHz at a

current of 5 μA. A version of the technical implementation of the

device is described in [6].

Results

At the first stage of the study, an atlas of meridians

described in [1] was used to select a list of BAP related to

diseases of the cardiovascular system, from which, on the

basis of recommendations in [1, 5], BAP associated with

the cardiac meridian (C1�C9) were used to determine the

risk of developing myocardial infarction (MI). The sym�

pathetic heart meridian V15 and ear points AP19, AP21,

AP60, AP100, AP105, and AO115 were excluded because

of low informativeness and the inconvenience of measur�

ing their characteristics.

A total of 40 patients with MI were observed.

Observations of changes in patients’ status were made

over the period of a year. The study groups included peo�

ple with increases in the energy characteristics of the

major points of the heart meridian. The energy character�

istics of BAP were monitored monthly. People showing at

least a small increase in the resistance of the major BAP

70 Korenevskiy et al.

(more than 10% of nominal) without therapeutic or

healthcare measures were assigned to the class of patients

with high risk of onset and development of recurrent MI.

At the end of the one�year observation period, 94% of

subjects showed the initial clinical signs of deterioration

in the activity of the cardiovascular system [4].

At the second stage of the study, the Kullback meas�

ure of informativeness was determined with a system of

graduated features, as described in [2, 5].

Kullback analysis of informativeness allowed selection

of six BAP for decision rule (2) from the overall list: C4,

C6, C7, C8, and C9 (point numbers as described in [1]).

Our statistical studies showed that the system of grad�

uations described in [5] worked adequately for identifying

cardiac and cardiovascular system pathology. In the tasks

of prognosticating MI, as demonstrated by the correspon�

ding statistical studies, this graduation was very rough, not

providing the required quality of decision�making.

Considering this, the gradation system was modified

as follows: >500 kΩ – 0; 400�500 kΩ – 1; 300�399 kΩ – 2; 200�299 kΩ – 3; 100�199 kΩ – 4; 90�99 kΩ – 5; 80� 89 kΩ – 6; 70�79 kΩ – 7; 60�69 kΩ – 8; 50�59 kΩ – 9; 40�49 kΩ – 10; <40 kΩ – 11.

The unevenness in the graduated scale is due to dif�

ferent information values in relation to the classification

of the states under study.

Data from one year of studying patients with acute

myocardial infarction undergoing rehabilitation in med�

ical institutions in Kursk were used to construct a table of

coefficients of confidence (CCjr) for each of the BAP

recorded ( j is the number of the BAP and r is the number

of the gradation) (Table 1). The table is based on distribu�

tion histograms of gradations of features in terms of the

class ωMI. In constructing the table, the recommenda� tions in [2, 3] were taken into account.

These coefficients determine the prognostic ability

of each BAP electrical resistance gradation in relation to

the possible occurrence of the target class of diseases.

The overall confidence that a recurrence of MI will

occur is determined by Eq. (2). Analysis of the data pre�

sented in Table 1 shows that a drop in BAP resistance to

below 60 kΩ provides a reliable (CCMI > 0.92) prognosis of the occurrence of recurrent myocardial infarction.

For the range 60�90 kΩ, there is a zone of indeter� minacy in decision�making which could be significantly

decreased by using additional informative features.

In the present study, the following were selected at

the expert level: the level of long�term psychoemotional

tension – x1, antioxidant activity – x2, and lipid peroxi�

dation – x3.

The level of long�term psychoemotional tension was

determined as described in [2, 3]. Properties x2 and x3

TABLE 1. Prognostic Table for the ωMI Class

400�500

0

0

0

0

0

300�399

0

0

0

0

0

200�299

0

0

0

0.1

0.1

100�299

0

0.1

0

0.4

0.1

90�99

0.2

0.3

0

0.2

0.1

80�89

0.4

0.5

0

0.2

0.2

70�79

0.5

0.6

0.2

0.2

0.2

60�69

0.6

0.7

0.3

0.2

0.2

50�59

0.7

0.8

0.5

0.2

0.3

40�49

0.9

0.9

0.8

0.2

0.3

<40

0.95

0.95

0.95

0.2

0.3

Resistance ranges, kΩ BAP

С*9

С*7

С8

С4

С6

>500

0

0

0

0

0

* BAP forming groups of features excluding all “interfering” situations (DSP BAP).

μωMI(x1)

Fig. 1. Plot of ωMI membership function for the basic variable x1.

μωMI(x2)

Fig. 2. Plot of ωMI membership function for the basic variable x2.

Prognosis of Recurrent Myocardial Infarction Based on Shortliffe Fuzzy Models 71

were measured by conventional laboratory analytical

methods.

Using the recommendations described in [2�4],

membership functions μωMI(xj) were constructed to relate x1, x2, and x3 to the class of “high risk of developing

myocardial infarction (ωMI) during remission.” A plot of the membership function μωMI(x2) is shown in Fig. 1.

Features x2 and x3 were determined, as recommend�

ed in [4], by

x2 = _______ ⋅100%; x3 =

_______ ⋅100%,

where xP H

and xA H

are lipid peroxidation and antioxidant

activity measured in a representative group of healthy

subjects, and xP T

= x19 and xA T

= x20 are lipid peroxidation

and antioxidant activity in the test patient.

Plots of the corresponding membership functions are

shown in Figs. 2 and 3.

The final confidence that the subject will develop

myocardial infarction is determined by Eq. (1), in which

CCωMI(1) = CCωl, determined by Eq. (2);

CC*ωMI(2) = μωMI(x1); CC*ωMI(3) = μωMI(x2); CCωMI(4) = μωMI(x3).

Results from mathematical modeling and statistical

calculations for representative control cohorts showed

that use of all four components gave a level of confidence

in the correct prognosis of the occurrence of myocardial

infarction during remission of 0.95.

REFERENCES

1. Luvsan, G., Tradition and Current Aspects of Eastern Reflex

Therapy [in Russian], Nauka, Moscow (1986).

2. Korenevskiy, N. A., Krupchatnikov, R. A., and Al’�Kasasbekh, R. T.,

Theoretical Grounds for the Biophysics of Acupuncture with Appli�

cations to Medicine, Psychology, and Ecology Based on Fuzzy Net�

work Models. A Monograph [in Russian], TNT, Staryi Oskol (2013).

3. Korenevskiy, N. A., Shutkin, A. N., Gorbatenko, S. A., and

Serebrovskii, V. I., Assessment and Control of the State of Health in

Students Based on Hybrid intelligent Technologies. A Monograph

[in Russian], TNT, Staryi Oskol (2016).

4. Seregin, S. P., Vorob’eva, O. M., and Korenevskaya, S. N.,

Mathematical Models for the Prognostication and Prophylaxis of

Recurrent Myocardial Infarction during the Rehabilitation Period.

A Monograph [in Russian], N. A. Korenevskiy (scientific editor),

Southwest State University, Kursk (2015).

5. Portnov, F. P., Electropuncture Reflex Therapy [in Russian],

Zinante, Riga (1980).

6. Korenevskiy, N. A., Mukhataev, Yu. B., Startsev, E. A., and

Lazurina, L. P., “A multichannel analyzer of the meridional struc�

tures of the body based on AFE analog interfaces,” Med. Tekh.,

No. 6, 28�31 (2016).

7. Korenevskiy, N. A., “The use of fuzzy decision�making logic for

medical expert systems,” Med. Tekh., No. 1, 33�35 (2015).

8. Korenevskiy, N. A., “A method for constructing heterogeneous

fuzzy rules for the analysis and control of the state of biotechnical

systems,” Med. Priborostr., No. 2, 99�103 (2013).

xA H − xA

T

xA H

xP H − xP

T

xP H

μωMI(x3)

Fig. 3. Plot of ωMI membership function for the basic variable x3.