Improvement of level-crossing reconstruction of potential maps by use of a priori information

The authors refine the approach for simplification of the body surface potential mapping (BSPM) acquisition system, which is based on partial measurements of level-crossing (LC) data. The algorithm is based on a combination of the optimal signal representation by the K-L transform and the theory of projections onto a convex set. A training set consisting of 200 subjects has been used for calculating a basis of decomposition functions. The reconstructions have been performed on LC measurements of BSPMs from a test set of 80 subjects, with similarly distributed pathologies. By incorporating a priori information about the underlying electrophysiological nature of the electrocardiogram (ECG) signals, the reconstruction error was decreased from an average of 120 μV root mean square (RMS) for the whole test set, to about 100 μV