GATE based simulation studies were conducted to guide breast PET system designs with different system geometries. A breast phantom (modeled as a half prolate spheroid with 7 cm equatorial radii and 11 cm polar radius) was generated for the study with warm background and hot lesions of different diameters (2, 3, 4 and 6 mm) and lesion/background activity ratios (31, 51 and 81, F-18 isotope). Each detector consists of an 8 × 8 array of 2 mm × 2 mm × 30 mm LSO scintillators with depth-of-interaction (DOI) measurement capability. Four systems were simulated: two stationary systems both with four detector panels arranged in a box-shaped geometry but different panel-to-panel separations of 17.2 cm and 20.4 cm, a system with two rotating detector panels separated at 17.2 cm panel-to-panel distance, and a whole-body (WB) PET with ∼90 cm detector ring diameter. Axial FOVs are 12.8 cm for all three breast systems and 16 cm for WB PET. A list-mode OSEM algorithm with 40 subsets and one-pass iteration was used for image reconstruction. The reconstructed images and measured contrast-noise-ratio (CNR) with different system configurations, DOI resolutions, lesion size, and activity ratios were used to compare the imaging performance of different systems. The results validate that DOI is critical to providing uniform spatial resolution across the FOV for a breast PET system with compact geometry, and show that the full-tomographic configuration with detectors encompassing the FOV provides superior imaging performance than that from the rotating detector PET or WB PET, with very high sensitivity (> 60% at the center of FOV), much better lesion visual identification, and significantly improved CNR. Based on these quantitative evaluations, this simulation study has demonstrated that even coarse DOI resolution (∼5 mm) can provide substantially improved imaging performance for a small bore system and design flexibility with different system geometries.