In emission tomography, the process of image reconstruction has evolved since the utilization of the filtered back projection algorithm (FBP). Nowadays, iterative techniques incorporating statistical knowledge of the photon emission process has produced much interest among researchers, not only for the improvements achieved in image quality, but also by the possibility of incorporate corrections factors, like scattering, attenuation, etc. One of the problems that has not been discussed in depth is the incorporation of correction methodologies due to patient breathing, which produces blur regions in lungs and cardiac images. Solutions like respiratory gating, that synchronize the breathing cycle of the patient and the data acquisition process, or correlated dynamic PET techniques that use external radioactive markers and list mode data, have been tested with improvements over the spatial activity distribution in lungs lesions, but with the disadvantages of requiring extra hardware or more expensive scanner systems. The objective of this study was to incorporate breathing-movement corrections directly to the phase of image reconstruction, without any additional acquisition protocol consideration. For this, a procedure of correction inside the probability matrix in the classical MLEM algorithm has been implemented which takes into account the spatio-temporal relationship of each voxel in the structure under study. We present 2D and 3D results from synthetic and realistic simulations showing the potential benefits of using this approach.