Real-time detection of ischemic ECG changes using quasi-orthogonal leads and artificial intelligence

Abstract
The authors describe a novel real-time ECG monitor for detecting ischemic ECG changes using three quasi-orthogonal leads. ECG recordings were made during angioplasty in 27 patients, with 19 patients used as a learning set and eight patients as a test set. Ischemia-detection algorithms were generated from the learning set using both logistic regression and inductive learning approaches, but only the latter approach gave acceptable accuracy on the test set (sensitivity 92% specificity 91%). The use of QRS-plane-referenced and polarcardiographic ST measurements improved performance when combined with conventional ST criteria. It is concluded that real-time ischemia detection is both feasible and practicable.<>