A performance study of 3D reconstruction algorithms for positron emission tomography

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Positron Emission Tomography (PET) scanners without septa are crucial for whole-body imaging.
• Accurate 3D image reconstruction is essential for quantitative analysis in PET.
• Evaluating reconstruction algorithms ensures optimal image quality and diagnostic performance.

Purpose of the Study:

• To statistically and systematically assess the accuracy of five 3D PET reconstruction algorithms for septa-less multi-ring scanners.
• To compare the performance of reprojection, direct Fourier, FAVOR, single-slice rebinning, and multi-slice rebinning algorithms.
• To identify the strengths and weaknesses of each algorithm regarding image quality and quantitative accuracy.

Main Methods:

• Simulated PET data from uniform cylinders, Gaussian sources, and spherical sources were used.
• Image quality metrics evaluated include noise properties, modulation transfer function (MTF), and recovery coefficients (RC).
• Real brain scan data were analyzed using linear regression to compare mean values within regions of interest (ROIs).

Main Results:

• The reprojection algorithm showed a slight loss of transaxial resolution in the scanner's external slices.
• Multi-slice rebinning algorithms resulted in increased image noise compared to other methods.
• Differences in noise properties, MTF, and RC varied among the evaluated algorithms.

Conclusions:

• Algorithm choice significantly impacts 3D PET image quality and quantitative accuracy in septa-less scanners.
• Multi-slice rebinning may introduce unacceptable noise levels for certain applications.
• Further optimization or alternative algorithms are needed to balance resolution and noise performance.