Reconstruction for 3D PET Based on Total Variation Constrained Direct Fourier Method.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• 3D Positron Emission Tomography (PET) generates complex data requiring advanced reconstruction techniques.
• Existing methods may struggle with noise and accuracy in PET image reconstruction.
• The piecewise constant nature of PET images suggests suitability for regularization methods like total variation.

Purpose of the Study:

• To develop and validate a novel 3D PET image reconstruction algorithm.
• To improve the accuracy and reduce artifacts in reconstructed PET images.
• To compare the performance of the proposed method against conventional direct Fourier (DF) reconstruction.

Main Methods:

• Employed Fourier rebinning (FORE) to convert 3D PET data into 2D sinogram datasets.
• Formulated the 2D PET reconstruction as an optimization problem incorporating Total Variation (TV) norm and data fidelity terms.
• Utilized the Bregman operator splitting algorithm with variable step size (BOSVS) for solving the minimization problem.

Main Results:

• The proposed TV-regularized reconstruction algorithm (BOSVS) demonstrated higher accuracy compared to conventional Direct Fourier (DF) methods.
• BOSVS achieved 70% of the bias and 80% of the variance observed in DF reconstructions.
• Validation experiments using both simulated and real PET data confirmed the method's effectiveness.

Conclusions:

• The developed TV-regularized reconstruction algorithm offers a significant improvement in accuracy for 3D PET imaging.
• The BOSVS method effectively leverages the piecewise constant characteristics of PET images for better reconstruction.
• This approach provides a more accurate alternative for quantitative analysis in PET studies.