Classification of inverse solutions to two dipoles

The aim of the simulation study was to identify cases in which an inverse solution to two dipoles characterizes the existence of two simultaneous lesions with changed repolarization. Difference STT integral body surface potential maps computed for 48 single lesions and 96 pairs of lesions were used as the input data for an inverse solution to two dipoles. Additional noise with signal to noise ratio (SNR) 20, 30 and 40dB was applied to the input data. The inverse solution was obtained as several pairs of dipoles. 23 characteristics of the solution were used as features for the quadratic variant of the Fisher discriminant analysis that should distinguish the inverse solutions that correctly identify 2 lesions from those yielding incorrect results or corresponding to a single lesion. The mean localization error in cases of correct results was 1.2±0.8cm regardless of the SNR. If eight most informative features were used, the sensitivity of the classification method was from 97 to 89% and the specificity from 94 to 84% for input maps with SNR from 40 to 20dB. The inverse solution to two dipoles together with the proposed classification of obtained results yields the identification of 2 simultaneous lesions without the need of some a priori information about the number of lesions.