In spontaneously breathing or ventilated subjects, it is difficult to image cardiac-related conductivity changes using electrical impedance tomography (EIT) due to the high amplitude of the ventilation component. Previous attempts to separate these components included either electrocardiogram-gated averaging, frequency domain filtering or holding the breath while performing the measurements. However, such methods are either not able to produce continuous real-time images or to fully separate cardiac and pulmonary changes. The aim of this work was to develop a new dynamic filtering method for the online separation of pulmonary and cardiac changes avoiding the drawbacks of the previous attempts. The approach is based on estimating template functions for the pulmonary and cardiac components by means of principal component analysis and frequency domain filtering. Then, these templates are fitted into the input signals. The new method enables an observer to examine the variation of the cardiac signal beat-by-beat after a one-time setup period of 20 s. Preliminary in vivo results of two healthy subjects are presented. The results are superior to frequency domain filtering and in good agreement with signals averaged over several cardiac cycles. The method does not depend on ECG or other a priori knowledge. The apparent validity of the method's ability to separate cardiac and pulmonary changes in EIT images was shown and has to be confirmed in future studies. The algorithm opens up new possibilities for future clinical trials on continuous monitoring by means of EIT and for the examination of the relation between the cardiac component and lung perfusion.