Reconstruction of gated single-photon emission tomography (gSPET) is intrinsically a four-dimensional (4D) problem. In practice, the time frames are reconstructed independently as a sequence of frame-by-frame reconstructions. This approach is not optimal since the strong signal correlations among the individual time frames are not exploited. In this study we propose a simple but efficient algorithm to improve the image quality of myocardial perfusion gSPET by incorporating the cyclic temporal information within the reconstruction using Fourier filtering. The gSPET images were reconstructed using the Ordered Subsets Expectation Maximisation (OSEM) algorithm employing six iterations with eight subsets. Temporal filtering was applied either before (PreOSEM) or after image reconstruction (PostOSEM) or was incorporated within the OSEM algorithm (OSEM4D). The effect of temporal filtering was compared with conventional frame-by-frame OSEM using clinical data. Image quality was evaluated by estimating the systematic and statistical error. The results indicated that temporal filtering introduces a small (<1%) systematic error, while the statistical error was reduced from 15.0%+/-3.1% when conventional frame-by-frame OSEM was applied to 12.6%+/-2.7%, 12.0%+/-2.5% and 9.3%+/-2.4% when PreOSEM, PostOSEM and OSEM4D were used, respectively. It is concluded that temporal filtering incorporated within OSEM reconstruction dramatically reduces noise in gated SPET myocardial images.