Neural networks for ECG diagnosis of inferior myocardial infarction

Author

Heden, B. ; Edenbrandt, L. ; Haisty, W.K., Jr. ; Pahlm, O.

Author_Institution

Dept. of Clinical Physiol., Lund Univ., Sweden

fYear

1993

fDate

5-8 Sep 1993

Firstpage

345

Lastpage

347

Abstract

Artificial neural networks are computer-based expert systems which have been used for several pattern recognition tasks, e.g. classification of electrocardiograms (ECG). Another method which has proved to be a valuable tool for improving the quality of ECG interpretation is the synthesized vectorcardiogram (VCG). These two methods were combined and applied on a material of 1458 ECGs recorded from subjects with or without inferior myocardial infarction (MI) evidences by ECG independent data. Neural networks fed with measurements from the ECG as well as the synthesized VCG showed after a learning process a sensitivity of 83% at a level of specificity of 97%. The result shows that planar ST-T measurements and QRS measurements from the synthesized VCG improved the performance compared to networks fed with scalar QRS measurements only

Keywords

electrocardiography; medical signal processing; ECG interpretation quality improvement; QRS measurements; artificial neural networks; computer-based expert systems; electrocardiograms classification; inferior myocardial infarction; learning process; neural networks ECG diagnosis; planar ST-T measurements; sensitivity; specificity; synthesized vectorcardiogram; Arteries; Artificial neural networks; Cardiac disease; Computer networks; Electrocardiography; Heart; Medical diagnostic imaging; Myocardium; Network synthesis; Neural networks;

fLanguage

English

Publisher

ieee

Conference_Titel

Computers in Cardiology 1993, Proceedings.

Conference_Location

London

Print_ISBN

0-8186-5470-8

Type

conf

DOI

10.1109/CIC.1993.378433

Filename

378433