616.8-005-072:517.938 . 2010. 5. . 81-90.
.. 1, .. 1, .. 2,
.. 3,.. 2
1
2
3- , ,
. . (ASA/AHA), 2009 , , , (). . , , , 1992 2006, 2 623 000 TIA 0.37 1.1 1000 . Oxford Vascular Study 2002 2004 0.66 1000 . . 1000 0.67 , – 0.61.
, Cardiovascular Health Study, 2.7% 65 69 4.1% 75-79 . 65 69 1.6% 4.1% 75-79 . 0.4% 45 64 [30].
.
. , . . , , , , - . , , , [3]. « » « », . - ( - , ..) , . , . .
- - in statu nascendi « » «» . «» , ( , ..) [7, 8, 13, 14]. , , , - .
, , . , . [1, 2].
. , é -. - «- », - , , , « » , . , , - . - . «ó », . « ó », , , .
– , , , , , , , , , , , , .
, . [7-10, 15, 16], , . , .
, « ó » - . , . , : - , , , . : « », ?
, 1987 , :
1. (1).
2. (2).
3. (3).
4. ().
« » ( ). , - «», () [13, 14, 26, 27]. , - [4-6, 8-11].
66 (30 , 36 , 54.9 ). 22 , 22 – (, , , ), 22 ( ) . 35 , , . 101 .
, (, , , ), -, . [13, 14, 26, 27], , , « » . . 1 - .
- , , , . , , , « » « » , « » . .
1. (-, )
|
,
|
|
1 |
6 |
|
2 |
2 |
|
3 |
|
5 |
4 |
4 |
|
5 |
4 |
|
6 |
|
3 |
7 |
10 |
5 |
8 |
10 |
7 |
9 |
( , ) |
4 |
10 |
3 |
|
11 |
|
3 |
12 |
|
3 |
13 |
|
4 |
14 |
|
3 |
15 |
|
5 |
16 |
|
3 |
17 |
|
3 |
18 |
|
4 |
19 |
3 |
|
20 |
|
4 |
21 |
|
3 |
22 |
|
3 |
23 |
, |
4 |
24 |
|
3 |
25 |
( ) |
4 |
26 |
5 |
|
27 |
( ) |
2 |
28 |
|
3 |
29 |
() |
3 |
30 |
() |
4 |
31 |
3 |
|
32 |
3 |
|
33 |
() |
3 |
34 |
( ) |
3 |
35 |
() |
3 |
36 |
( ) |
4 |
37 |
|
3 |
38 |
() |
4 |
39 |
( ) |
4 |
40 |
4 |
|
41 |
|
2 |
. - . - , - . , - , “ ” (“gray box”), . 1.
![]() |
. 1. -
-
. 2 , , [6]. . , . 41- , ( ) . 4- (1-3) ().
![]() |
. 2. - -
, .. -, - , . . (), , . (Recirculation Neural Network - RNN), . 41 - 12 . . (MultiLayer Perceptron – MLP), () , , () ..
. (RNN), (. 3). , , .
![]() |
. 3. , RNN.
- . j- ,
,
F – ; Sj – j- ; wij – i- j- ; xi – i- .
:
,
w’ji – j- i- ; – i- .
RNN. [33], – (backpropagation algorithm). RNN :
,
.
RNN . , . 12- , 4 5.
RNN :
,
,
.
L – . 41 - 101 .
, RNN . :
,
- j- ,
– .
.
, 101 41 - (. 1). - . , , : 1 – 22 , 2 – 22 , 3 – 22 – 35 . 51 83 (. 2).
, , . , 101 50 18 . 100%- , , 76 77 .
2.
- |
- |
|
|
51 |
50 |
100%, |
76% |
83 |
18 |
100%, |
77% |
. . 5 RNN « » ( 1-3) (4). , .
. 4. RNN
RNN, . 5. (1-3) (4). RNN, RNN – ( ) «» . 2 (2, ) .
. 5. RNN
.
- « ». - « » , . . - « » , . - . 3, 10 « » .
3.
, , () |
1 |
2 |
3 |
|
“0” |
25% |
25% |
25% |
25% |
, , 48 |
29.1% |
29.9% |
40.9% |
0% |
, , 54 |
21.5% |
25.1% |
53.4% |
0% |
, , 59 |
50% |
30.5% |
10.7% |
8.8% |
, , 51 |
27.5% |
30% |
34.2% |
8.2% |
, , 54 |
44.8% |
13.9% |
27.8% |
13.5% |
, , 60 |
36.8% |
2.1% |
59.5% |
1.6% |
M, , 73 |
17.9% |
21.9% |
51.8% |
8.3% |
, , 46 |
34.5% |
5.6% |
59.9% |
0% |
M, , 63 |
17.9% |
21.9% |
51.8% |
8.3% |
M, , 71 |
34.5% |
5.6% |
59.9% |
0% |
, - . . .
. , .
, , «» - . - .
, - , . , , , , [19-22]. . .. (2003) , « , » [3]. , . , () , in statu nascendi. , , , , , « » [11, 19-22].
, - (, , ) , , , - , , , . , , . , . , , , - , - .. . «» . , 100%- , « », , , , [3-6, 8-12, 18, 24, 28, 29, 31-34].
- , , () . ASA/AHA [30], , , , (0.3, 1.1, 0.6, 0.61 .. 1000 ) , . , , [17, 23]. , 12 2007 92 « », - .
, , , -, , , . - . - . , . , « , , , . «» , . » [25]. - - [12].
, (RNN) , . 5. , - ( , , , ). , , , . , , , « ó » - . (, , ) - .
ó «» - (, , , ).
, .
1. .. . , , , 25 2005.
2. .. : . . . . . 95.13.09. , 1991. - 27 .
3. .. . . . . 2003. . 521-533.
4. .. . .... . . 05.13.01. , 2002. 46 .
5. .. . VII - -2005. . .: , 2005. . 43-91.
6. .., .. ..// . , , . 2006. 5(41). . 14-19.
7. .., .., .., .. // . 2009. 1. . 143-148.
8. .., .., .., .. // ARS MEDICA. 2009. 3. . 60-72.
9. , .., .., .. // . 2009. 3. . 11-119.
10. .., .., .. . - « ». 20-21 2010. . . . 349-355.
11. .., .., .., .., .., ..// i i. . . , 2010, -. .
12. . . . …. . . -. 05.13.18 . – 2002. 48 .
13. .., .., .. // . 2004. 4. . 18-21.
14. .., .., .. // i i (. . ). 2006. 1. . 76-82.
15. . . . . . . . . 05.13.18. , 2008. 18 .
16. . ., .., ..// . 2008. 3. . 105-112.
17. . . // . 2006. 7. . 18 – 23.
18. .., .., .. // . .. . (), 2004, . 12, . 23-28.
19. . . // . 14 1968 . . . 35–38.
20. .., .., .. // . . . 1982. . 26. 7. . 665-668.
21. .. : . .- 2002. 164 .
22. . . // . 2004. 8. C. 5–9.
23. ..// . 2007. 9. . 41-43.
24. .., .. . .-. . . 2000. 39 .
25. .., .., ..// // . 1986. 9. . 65-85.
26. .., .. // 1997, 2 . 47-49.
27. .., .., .., .. // . . 1998. 18 .
28. . . .: . 1980. 317 .
29. Cpalka K, Rebrova O, Rutkowski L. // Lecture Notes in Computer Science 5769, Springer-Verlag Berlin Heidelberg 2009, 435-444.
30. Easton J.D., Saver J.L., Albers G.W. et al. //. Stroke. 2009. Vol. 40. 6. P. 2276-2293.
31. Kosko B. Fuzzy Thinking: The New Science of Fuzzy Logic. Hyperion. 1993.
32. Laurentsyeva S., Golovko V., Evstigneev V. // Proceedings of the Tenth International Conference of Pattern Recognition And Information Processing (PRIP’2009). Minsk. Belarus, 19 – 21 May. Minsk, 2009. P. 327–331.
33. Oja E.// Neural Networks archive. 1992. Vol. 5. 6. P. 927-935.
34. Zadeh L. A. // Communications of the ACM. 1994. Vol. 37. 3. P. 77-84.
616.8-005-072:517.938
.., .., .., .., .. . .. 2010. .
(, ) . , . , , , . , – . . , , . .
100%- , (). , 75%- , , -, .
. 3. . 6. – 34 .
A.S. MASTYKIN1, V.V. EVSTIGNEEV1, V.A. GOLOVKO2,
E.N. APANEL3, G. Yu. VOYTSEKHOVICH2
THE NEURONET APPROACH FOR RESOLVE THE PROBLEM OF DIAGNOSTICS END PREVENTION OF TRANSIENT ISCHEMIC ATTACKS
1Belarusian Medical Academy of Post-Graduate Education
2Brest State Technical University, Belarus
3Scientific and Clinical Center o f Neurology and Neurosurgery, Minsk, Belarus
Summary
Nocuous influences (attacks, intrusions) on normally proceeding processes in various areas of practical activities induce to search for new more refined methods of protection against them. Normal work on the Internet and normal ability to live of a living organism in surrounding conditions is exposed to such attacks. Informative methodological approaches to a problem of protection against nocuous calls, both in a society, and in a separate living organism, can be similar and even identical. Here and there the main is to unmask and to determinate the pathogenic agent. In this context in the present work the decision of a problem of protection against nocuous attacks to normal blood supply of a brain by convergence with the neuronet approach to protection against virus network attacks on the Internet is concretized. Steady classification decisions are necessary, first of all, for this purpose on recognition of images. In our medical research it is correct diagnostics.
On an available initial material on training sample 100 %-s' recognition on four classes of recognition of images as which are subtypes of transient ischemic attacks (TIA) act has been received. However on test samples the result has appeared more modest, slightly exceeding unstable 75 %-s' level, that, apparently, is caused by insufficient amount of the verified clinical cases on subtypes TIA in training sample.
25.09.2023