ECG monitoring, classification and early warning by sensitive neural nets

Abstract

Medical engineering support systems that are controlled by neural networks are being applied with increasing frequency in medical practice. However, a solution still needs to be found for the problem of constructing medical support systems that can be set up by the physicians themselves without the need of a knowledge of the mathematical theories of neural networks and signal processing. We describe a medical support system which can be set up by novices in the field of neural networks and which can be controlled and the results correctly interpreted by nurses and other medical staff. On the basis of typical pathological types of ECG signals, simulated by a common simulator which is used in Europe for heart beat monitoring, we show the basic structure of normal and pathological heart beat signatures and how they can be presented in a new and readily interpretable display. Furthermore, we explain how this support system can be used to create networks which are typical for most patients, are small, and can be quickly set up to monitor the patient's condition during therapy, or to provide a measure of the depth of anaesthesia of the patient.