Abstract: Dual-layer-offset crystal array is a low cost and simple scheme for positron emission tomography (PET) system to achieve depth of interaction (DOI) information. Different thickness combinations of two layer crystals influence system performances. Monte Carlo simulations using GATE toolkit were implemented to optimize the design by evaluating the performances of the system with different thickness combinations in this study. Eleven magnetic resonance imaging (MRI) compatible brain PET inserted systems with an inner diameter of 345 mm, dual layer LYSO of 20 mm total thickness and inner layer thickness varied from 0 mm to 10 mm with a step size of 1 mm were built. Six point sources in the x-axis of the center slice from x=0 cm to x=10 cm with an interval of 2 cm were simulated. The sensitivity at the center was calculated and the radial resolutions were evaluated using the reconstructed points by the filtered back projection algorithm. The results show that dual-layer crystal design has a slight sensitivity decrease, but can significantly improve the radial spatial resolution and the resolution uniformity compared with the single-layer design. With the thickness increase of the inner crystal, the average radial spatial resolution decreases and then increases, and achieves minimal at inner layer with a thickness of 8 mm. In conclusion, 8 mm in the inner layer and 12 mm in the outer layer is the optimal design for the brain PET system with 20 mm dual-layer-offset crystal proposed in this study.