|
|
||||||||
|
|
|||||||||
|
|
| ||||||||||||||||||||||||||||||||||||||
Instrumentation & Data Analysis: Image GenerationImage Generation Posters |
1 Radiology, SUNY, Syracuse, New York; 2 Math; 3 Physics, SU, Syracuse, New York
1659
Objectives: To apply attenuation compensation in irregular mesh-based SPECT forward projection and tomographic reconstruction.
Methods: We performed analytical and Monte Carlo simulations of parallel-beam SPECT for a digital chest phantom and a simple geometrical phantom. The two-step reconstruction process consists of: First, reconstruction in pixel domain using OSEM algorithm with line-length system model. Second, a singularity-based mesh is generated from this initial reconstruction using wavelets transform and OSEM reconstruction is performed in mesh-domain with the mesh-domain system matrix. Three approaches for attenuation compensation were investigated: Approach 0, nodal ray approximation, assumes that all photons emitted from mesh elements originate from the associated nodes. Approach 1, average central ray approximation, and Approach 2, individual central ray approximation, assume that all photons emitted from mesh elements intersecting with a tube-of-response (TOR) of the detector originate from mid points of the sections of TOR within the mesh elements of interest. In Approach 1, we estimate the attenuation correction factors as an average attenuation along the ray traversing all mesh elements to which the node belongs. In Approach 2, we estimate the attenuation correction factors individually for each central ray segment within all mesh elements associated with the node.
Results: Approaches 0 and 1 resulted in similar sinograms and images with 15% relative error, as compared to pixel-domain technique. Approach 2 resulted in a significantly smaller relative error (below 5%).
Conclusions: We conclude that Approach 2 to attenuation compensation in mesh-domain SPECT reconstruction results in relatively small errors that are acceptable for practical applications.
| ||||||||||||||||||||||||||||||||||||||