Medium-sized animals, such as dogs and rabbits, play an important role in the evaluation of new drugs, treatments and procedures but are often too large to be imaged by small-animal pinhole SPECT systems. An alternative is to use a standard clinical system, but the relatively small size of these subjects then requires a maximization of the resolution. One approach to achieving this is to use a clinical system with converging collimators, which provide an increase in resolution by magnification. Additional improvements in resolution can be obtained by accurately modeling the data acquisition process by incorporating into an OSEM algorithm the collimator blurring functions. We compared a high-resolution fanbeam collimator to a standard LEHR collimator. Reconstructed collimator performance was measured using a phantom consisting of plastic cylinders and spheres to mimic a complicated activity distribution. The edge spread function of a middle coronal slice of the phantom was used to measure the transaxial and axial FWHM. An increase in transaxial resolution of 32% was observed in the fanbeam compared to the parallel hole collimator. An in-vivo transaxial FWHM taken from the vertebral body of a rabbit measured 3.8 mm vs. 4.1 mm for fanbeam and parallel hole collimators respectively.