Ischemic episode detection using an artificial neural network trained with isolated ST-T segments

The presence of changes in the ST segment of the ECG followed or not by chest pain is related to ischemic heart disease and indicates increased risk of malignant arrhythmia and sudden death. This paper proposes a method for automatic ST changes detection to help the analysis of ambulatory monitoring. ECG data segments with ST-T intervals are extracted and, through principal component analysis, reduced to six components that represent 98.1% of total data variance. Data from 45 patients of the European ST-T Database were organized into three groups to develop, validate and test feedforward ANN. With six inputs, 10 neurons in the hidden layer and three in the output the ANN showed 69.6% total accuracy when classifying isolated segments. Fixed thresholds were applied to the output neutrons for detecting sequences of abnormal ST segments. The performance, 85.83% sensitivity for negative and 78.38% for positive, are similar to others reported in literature