Identification of Two Lesions with Local Repolarization Changes Using Two Dipoles in the Inverse Solution

In the simulation study the method for inverse identification of two distinct simultaneous lesions with changed repolarization in the ventricular myocardium is described. The inverse localization of two lesions was obtained from the difference STT integral body surface potential maps (DIMs). The DIMs corresponding to 48 single lesions and 96 double lesions were simulated on the surface of an inhomogeneous torso using a numerical model of the ventricles. It was supposed that the double lesions can be represented by two dipoles. Twenty three characteristics of the obtained inverse solution were specified and used as the features in a discriminant analysis that should distinguish the inverse solutions that correctly identify two lesions from those yielding incorrect results or corresponding to single lesion. The mean lesion localization error in cases with two lesions was 1.2 cm. For optimally selected subset of 7 features the quadratic discriminant analysis with cross-validation and feature selection yielded an overall double lesion classification error smaller than 12% for input ECG data with added noise if the signal to noise ratio was above 20 dB.