Abstract : Respiratory motion may blur the tomographic reconstruction of PET or SPECT images, which subsequently impairs quantitative measurements, e.g. in the upper abdomen area. Respiratory phase-based gated reconstruction addresses this problem for CT images, but deteriorates the signal-to-noise ratio and other intensity-based quality measures for ET images. This article describes an image-based motion detection method for dynamic images, which works for both 4D-CT and 4D-SPECT images. This method allowed us to identify the motionless phases from the image-based motion signal. Interestingly, we found that the peak of motion according to the pressure signal, recorded during the acquisition and considered as a reference, is temporally shifted compared to the motion-phases identified by the proposed method. Moreover, these observations permit to reconstruct a breath hold like 3D SPECT image improving the quality of the 3D reconstruction compared to both 4D reconstruction and standard 3D reconstruction.