The high resolution research tomograph (HRRT) is a 3-D PET scanner designed for human brain and small animal imaging. The HRRT consists of eight panel detector heads that are separated by gaps of 17 mm resulting in gaps in the sinogram. Furthermore, gaps can result from detector-block failure. To prevent artifacts in the reconstruction when using Fourier rebinning (FORE), filling the data gaps is required. The purpose of this study was to evaluate the accuracy of three gap filling methods: a) bilinear interpolation of sinogram data; b) a model-based method in which an intermediate volume is reconstructed [2-D ordered subsets expectation maximization (2-D OSEM)] based on direct planes only, after which this image is forward projected to fill the gaps; c) an improved model-based method in which gaps are first filled using interpolation, then reconstructed using FORE + 2-D OSEM and forward projected. The improved model-based method out-performs interpolation, but requires more computation time.