Computerized recognition of acute infarction. Criteria advances and test results

The advent of thrombolytic therapy has increased the desire for an accurate computerized recognition of the ST elevation associated with acute myocardial infarction (AMI). Quantified electrocardiographic (ECG) criteria for AMI, described in the literature and in computerized ECG criteria packages, have concentrated on only those leads that exhibit ST elevation. The accuracy of this lead-specific approach has been maximized by an analysis of the associated ST-segment and T wave. Further advancement of the Marquette 12SL program's ability to classify ST elevation due to AMI has required the inclusion of those leads that are not elevated. This global approach was developed by modifying the 12SL program so that it would properly diagnose a small training set of clinically correlated AMI ECGs that did not meet the lead-specific criteria. Further training, via an evaluation of the false-positive rate, was done with a large clinical database (greater than 30,000 ECGs). Both the new and the old criteria were tested with a separate ECG database taken from a total of 296 patients. Of these, 77 had an AMI (determined via the hospital discharge diagnosis). The lead-specific criteria resulted in a sensitivity of 21% and a specificity of 99.5%. The global approach resulted in over twice the sensitivity (53%), while continuing to maintain a high rate of specificity (98%).