METHOD OF DIAGNOSING PATHOLOGIES OF HEMOSTASIS SYSTEM BY MEANS OF NEURAL NETWORK Russian patent published in 2012 - IPC G01N33/48 G06F19/00 

FIELD: medicine.

SUBSTANCE: in addition to AT-III activity level, platelet aggregation and content of products of fibrin and fibrinogen degradation determined are time of coagulation, echitoxic time, activated partial thromboplastin time, thrombin time, orthophenanthroline, Hageman-dependent fibrinolysis, streptokinase-induced euglubuline fibrinolysis, prothrombin index for any supposed pathology from limited set. For definite pathology analysis and statistical processing of numerical values of hemostasis parameters are carried out, reducing the number of input indices to necessary and sufficient quantity, statistically selected data are supplied on inlet of neuron nets, which are used to carry out formulation of preliminary diagnosis, differentiation of several pathologies and/or prediction of disease outcome. By results of statistical processing non-informative parameters are excluded from symptom complexes of investigated pathology, which results in increase of prognostic reliability of analysis results.

EFFECT: method makes it possible to apply neural network technologies for detection of pathology, diagnosed by means of hemostasis parameters.

1 ex, 3 tbl