A level set approach to segmenting a deforming myocardium from dynamically acquired spect projection data

Dynamic cardiac single photon emission computed tomography (SPECT) offers an effective way for observing fundamental physiological functions of organs and could aid in the early diagnosis of cardiovascular disease, in particular, for those patients with minimal disease. This would improve the chances of recovery by initiating appropriate therapy and an altered life style. To make dynamic cardiac SPECT viable with present clinical scanners methods need to be developed that reconstruct time activity curves from dynamically moving organs representing the change of tracer concentration as a function of time from projection data acquired from slowly rotating gamma cameras. This type of data analysis faces the challenge of modeling both rigid and non-rigid body deformation as well as modeling of a time varying tracer concentration. In the work presented here, we develop methods for segmenting the beating heart using an approach based upon level sets, which can deal naturally with topological changes. A variational formulation of the level set method was implemented. This allowed the inclusion of a priori information and was computationally efficient. The algorithm was first evaluated with simulated dynamic cardiac image data. The MCAT phantom was used to generate data containing 32 time frames over one cardiac cycle. Each frame had a matrix size of 64×64×32 voxels with a resolution of 6.25 mm. Starting with an initial estimate of the boundary, the algorithm then converged to an accurate segmentation of the deforming heart. The initial estimate was not important and we could segment simultaneously both interior and exterior boundaries. This algorithm forms the foundation for the segmentation of the boundary of the deforming myocardium directly from projection data.