Region-of-interest tomography using exponential radial sampling

Area of Science:

• Image reconstruction
• Computational imaging
• Tomographic imaging

Background:

• Traditional tomography requires dense sampling across the entire image.
• Local tomography reconstructs a filtered version of the region of interest (ROI).
• Existing methods may lack speed and parallelizability for ROI-specific reconstruction.

Purpose of the Study:

• To develop a new image reconstruction procedure for a small region of interest (ROI).
• To achieve efficient and accurate reconstruction of the ROI itself, not a filtered version.
• To leverage nonuniform sampling and circular harmonic expansions for enhanced performance.

Main Methods:

• Combines nonuniform sampling with circular harmonic expansions.
• Employs exponentially decreasing radial sampling density away from the ROI.
• Utilizes the Fast Fourier Transform (FFT) for computational efficiency.
• Algorithm is designed for parallel computation of image harmonics.

Main Results:

• Successfully reconstructs the ROI image from projections with varying sampling densities.
• Demonstrates speed advantages due to FFT implementation.
• Shows parallelizability as each harmonic is computed independently.
• Numerical comparisons indicate advantages over filtered backprojection.

Conclusions:

• The new algorithm offers an efficient and parallelizable method for ROI image reconstruction.
• It accurately reconstructs the ROI image, distinguishing it from local tomography.
• The approach shows promise for applications requiring targeted, high-speed image reconstruction.