Using automated analysis of the resting twelve-lead ECG to identify patients at risk of developing transient myocardial ischaemia--an application of an adaptive logic network.

The aim of this study was to introduce an adaptive logic network computing method for detecting patients who were likely to show transient ischaemic episodes during ambulatory Holter monitoring, using parameters from a previously recorded standard twelve-lead resting electrocardiogram (ECG). In the present study, the adaptive logic network computing method is compared with other commonly used classification methods, such as backpropagation network and discriminant analysis techniques. Of 1367 study subjects aged 65 and above, 733 were women and 634 were men. Ambulatory Holter recordings were made to detect episodic ischaemia in study patients. Those subjects showing ischaemic episodes were classified as 'ischaemic' patients, and the remaining subjects were 'non-ischaemic'. Accuracy was 67% using the adaptive logic network computing method, 56% using the backpropagation network computing method, and 65% using statistical discriminant analysis. We concluded that the adaptive logic network technique offers a slightly higher accuracy and shows several potential advantages for automated detection of ischaemia in resting electrocardiograms.