Expectation maximization (EM) algorithms using polar symmetries for computed tomography (CT) image reconstruction.

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Iterative algorithms like expectation maximization (EM) are crucial for image reconstruction.
• Traditional methods often divide 3D problems into 2D slices, increasing complexity and potentially reducing accuracy.
• Computed tomography (CT) and other tomographic techniques require efficient image reconstruction from projection data.

Purpose of the Study:

• To introduce a symmetric-polar pixellation scheme to enhance the efficiency of EM iterative algorithms.
• To enable holistic 3D image reconstruction without resorting to 2D slice-based methods.
• To evaluate the performance, stability, and image quality of the proposed scheme.

Main Methods:

• Development of a symmetric-polar pixellation scheme for 3D image reconstruction.
• Comparative analysis of the proposed scheme against 2D slice-based approaches in a 2D model.
• Implementation and evaluation of the best-performing 2D approaches in a 3D reconstruction model.
• Validation using real data from a small animal CT scanner.

Main Results:

• The symmetric-polar pixellation scheme significantly reduces computational cost for EM algorithms.
• The holistic 3D approach demonstrates superior stability and image quality compared to 2D slice-based methods.
• Performance evaluations confirmed the effectiveness of the proposed method in a 3D reconstruction model.
• Comparisons with Filtered Back Projection (FBP) in 3D reconstruction were also presented.

Conclusions:

• The symmetric-polar pixellation scheme offers a computationally efficient and effective method for 3D image reconstruction.
• This approach is applicable to various tomographic techniques beyond CT, provided suitable scanning geometry.
• The study validates the mathematical model with real-world small animal CT data.