Maximum likelihood-based motion estimation in cardiac SPECT imaging

Patient motion estimation strategies found in literature may be categorized primarily into two types: data-driven using the SPECT acquisition data itself to determine motion, and external surrogate-based, where motion is obtained from a dedicated motion-tracking system. In previous work, we investigated a data-driven strategy based on 2D to 3D registration using partial reconstruction of the heart. In this paper we implemented a combined motion and activity estimation method based on the maximum-likelihood approach commonly used in SPECT reconstruction. In this scheme, both the transformation and activity are updated within the same iteration for all motion states. This method has the potential of correcting gradual motion or frequent pose changes, where the quality of the partial reconstruction for registration-based motion estimation may be affected greatly. Here we present preliminary results for this technique using the cardiac NCAT phantom. The goal of this work is to validate and compare data-driven strategies for patient motion correction in cardiac SPECT.