Conventional Voxel in Tomographic Reconstruction Based upon Plane-Integral Projections – Use It or Lose It?

Document Type : Conference Proceedings


1 Radiation Application Research School, Nuclear Science and Technology Research Institute, Tehran, Iran.

2 Radiation Medicine Department, Shahid Beheshti University, Tehran, Iran.

3 Department of Medical Physics and Biomedical Engineering, Tehran University of Medical Sciences, Tehran, Iran. Research Center for Molecular and Cellular Imaging, Tehran University of Medical Sciences, Tehran, Iran.


Introduction: While the necessity of replacing voxels with blobs in conventional tomographic reconstruction based upon line-integrals is clear, it is not however well-investigated in plane- integral-based reconstruction. The problem is more challenging in convergent-plane projection reconstruction. In this work, we are aiming at utilizing blobs as alternative to voxels.
Materials and Methods: To doing so, a spinning slithole collimated small-animal SPECT scanner was modeled within GATE simulator and SPECT data acquisition was performed for NEMA IQ phantom. The slithole provides convergent-plane-integral projections of the object being imaged. The collimation data were then reconstructed using an ad hoc 3D OSEM algorithm based upon plane-integral projections using cubic voxels. The image reconstruction was repeated for blobs. Various figures of merit were assessed in order to compare voxel and blob reconstructions.
Results: Both voxel and blob reconstruction algorithms were verified using a noisy 3D Shepp- Logan phantom prior to reconstructing the simulation data, with an error of 6% and 9%, respectively. Blob modeling leads to a lower computation burden by a factor of ~2.1 in on- the-fly calculation of the system matrix. A 13% increase in SNR of the hottest rod of the NEMA IQ phantom is obtained when voxels are replaced by blobs. Blob reconstruction shows an advantage over voxel reconstruction by offering a monotonically decreasing NSE across all iterations indicating a more robust statistical noise handling.
Conclusion: Voxels are simple and easy-to-implement, but blobs outperforms them by exhibiting a much more efficient projector/backprojector pair due to their inherent symmetric and smooth shape. Therefore, blobs should be the basis functions of choice in the case of plane-integral-based reconstruction.