Abstract
This paper presents a fully 3D Bayesian image reconstruction and its localization properties of a pinhole SPECT for small animal imaging. Due to the unique geometric characteristics, pinhole SPECT for small animals can use existing clinical equipment to produce high resolution image results. However, low sensitivity is the major disadvantage of pinhole SPECT and leads to instability of iterative image reconstruction based on popular maximum likelihood expectation and maximization (MLEM). In this work, we describe a fully 3D system model of a pinhole SPECT combined with an image model based on medium root prior (MRP). Under the paradigm of Bayesian statistics, a modified EM algorithm with one-step late (OSL) is applied to reconstruct 3D SPECT images through maximum a posteriori (MAP) estimation. The experimental results of phantom studies indicate that the proposed 3D Bayesian image reconstruction can produce better edge-preserving image results and effectively suppress noises.