Saratov JOURNAL of Medical and Scientific Research

North-Caucasus Federal University

Application of self-learning neural network in obstetrics and neonatal practice

Year: 2013, volume 9 Issue: №2 Pages: 282-286
Heading: Clinical Laboratory Diagnostics Article type: Original article
Authors: Bondar Т.Р., Tsaturyan Е.О., Deryabin М.А., Zaytsev А.А.
Organization: North-Caucasus Federal University

The purpose of the study is to examine the feasibility of a self-learning neural network to predict the risk of hemo-static disorders in newborns and their mothers. Material: To solve the problems 214 patients have been under the study, of which — 107 women on a day of labor and 107 of their infants. Morphofunctional state of platelets was assessed by automatic hematological analysis, computer cytomorphometry and platelet aggregation. These laboratory studies of patients were used as a training set for a neural network. This study used a neural network with classical structure «layer perceptron.» Results: Testing of the neural network was carried out on two female patients with physiological pregnancy and signs of imbalance of the hemostatic system. At the output of neural network based on laboratory confirmed the presence or absence of pathology of the hemostatic system in two specific patients and their newborns. Conclusion: The introduction of practical obstetric program being developed based on the neural network will reduce and prevent the development of thrombohemorrhagic complications, the protocol will optimize diagnostic and treatment activities, and an opportunity for the state of hemostasis monitor on a background of pathogenetic therapy for women and their newborns during childbirth and neonatal period.

2013-02_282-286.pdf282.74 KB