One of the major issues in the inverse problem of electrocardiography is the sensitivity to noise (or in mathematically more correct terms, the ill-posedness of the problem). In general, research teams add constraints to the possible solutions to reduce the sensitivity to noise. These constrains are mainly based on physical or mathematical characteristics of the inverse problem. But what if we could use constraints that are actually based on more “biological” knowledge of the heart? In this conference paper we have seen some encouraging results when we try to achieve this by using simulated electrical training data on a heart surface.Reference: Matthijs Cluitmans, Ralf Peeters, Paul Volders, and Ronald Westra. Realistic training data improve noninvasive reconstruction of heart-surface potentials. In Engineering in Medicine and Biology Society (EMBC), 2012 Annual International Conference of the IEEE, pages 6373-6376. IEEE, 2012.