The "image surface" is a data collection of spa-tial vectors that reflect relationship between heart vec- tor and measured electrocardiograms, supposing the heart source is a fixed dipole source. For long, the Frank image surface has been used. However, since the Frank model represents an average model of west- ern population, it may not fit other populations like Japanese, and therefore may affect the accuracy of ap- plications In this study, to improve the accuracy of the derived 12-lead ECG, we developed the digital image surface based on the personalized torso models constructed from MRI images. The boundary element method was used to calculate the values of digital image surface by applying electrical current dipole source on the torso- heart model. Further, the valises of the digital image surface for eight subjects were averaged as a represen- tation of typical models of Japanese population. As an application, derived 12-lead ECG is obtained based on the digital image surface and the results is com- pared with those based on the Frank image surface. The results show that the digital image surface can get better accuracy than the Frank image surface.