Heterogeneous Classifiers with Virtual Flows in Intelligent Systems for Predicting Cardiovascular Complications during t
- PDF / 419,433 Bytes
- 4 Pages / 594 x 792 pts Page_size
- 23 Downloads / 128 Views
Heterogeneous Classifiers with Virtual Flows in Intelligent Systems for Predicting Cardiovascular Complications during the Rehabilitation Period E. V. Petrunina1, O. V. Shatalova2*, D. S. Zabanov2, and V. V. Serebrovskii2
Virtual models of “weak” classifiers for intelligent systems classifying the risk of recurrent myocardial infarction are considered. Bioimpedance investigation results were included in the construction of models of the risk of car diovascular complications as an additional risk factor. Five classifiers were studied, of which four were hetero geneous. The models of heterogeneous classifiers were obtained by sequentially increasing the number of deci sion modules used in the classification model. Use of all decision modules in a heterogeneous classifier gave diag nostic sensitivity of 0.90 with diagnostic specificity of 0.86. When a configuration of the attribute space includ ing only the conventional risk factors was used, the classification quality parameters were no worse than those for known risk scales for cardiovascular complications.
When addressing the task of synthesizing mathemat ical models prognosticating cardiovascular complications (CVC) during the rehabilitation period, it is important to bear in mind that doctors in polyclinics have access to standard, quite well tested, methods and techniques based on data from historytaking, examinations, laboratory investigations, and instrumented studies. Specific investi gation methods “bound” to specific organs and diseases also exist. However, there are virtually no standard sched ules of investigation methods for assessment of the course of the rehabilitation period and predicting complications in relation to various organs and diseases. There are only a few reports mainly focused on studies of mathematical methods for evaluating informative features in relation to the class of tasks under study and providing the corre sponding practical recommendations without sufficient generalizations which could be extended to various dis eases of the cardiovascular system (CVS) [1]. The present studies address the management of patients with acute myocardial infarction (MI), where assessment of the risk of recurrent MI (rMI) is an impor 1
Moscow State University of Humanities and Economics, Moscow, Russia. 2 Southwest State University, Kursk, Russia; Email: [email protected] * To whom correspondence should be addressed.
tant component in evaluation of the risk of CVC. Existing methods for prognosticating CVC have a number of sig nificant drawbacks: determination of individual prog noses in a large proportion of patients remains unsatisfac tory; there are essentially no unified systems prognosti cating the occurrence and outcome of disease; prognosti cation is a laborious operation for the doctor. Thus, development of methods for prognosticating rMI based on automated analysis of risk factors (RFs) using hetero geneous classifiers and mathematical models taking account of the latent variables in the risk of rMI is an important scienti
Data Loading...
