Feasibility of a predictive algorithm for identifying characteristic events of cardiac activation: results of a model study

By the use of body surface ECG signals, generated by a thoracic volume conductor and heart model, the validity of an algorithm suggested for the detection of the characteristic events of ventricular activation was tested. In the computations the volume conductor was considered by a real-shaped inhomogeneous torso, the generator of the electric sources by a propagating heart model with simplified geometry. The correct characteristic event classification was known from the simulated epicardial and cross-sectional activation maps. Results show, that by an optimal adjustment of the decision threshold of detection, a sensitivity of 60% and a specificity of 80% can be achieved. The detailed analysis of the false negative detections revealed, that the missed event have probably less importance in real shaped hearts, consequently, an increase in sensitivity may be anticipated without a compromise in specificity