A novel surface reconstruction algorithm based on a variational framework using the level set method is proposed and its application to atrial fibrillation (AF) treatment is presented in this work. The need of reconstructing the cardiac surface from multi-modality data with magnetic resonance imaging (MRI) and computer tomography (CT) for surgical AF treatment is first explained. An implicit form of the cardiac surface is analytically derived by optimizing an energy functional that takes into account the distance between the surface and electroanatomic mapping (EAM) data and its similarity to a prior shape model. This problem is then solved numerically. The resultant algorithm allows non-rigid deformation to occur in the reconstruction of the cardiac surface. Promising numerical and experimental results show the potential of the proposed approach.