Time-of-flight PET image reconstruction using origin ensembles

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Computational Science

Background:

• Emission tomography data reconstruction is crucial for medical imaging.
• Current methods like Maximum Likelihood Expectation Maximization (ML-EM) have limitations.
• List-mode (LM) time-of-flight (TOF) Positron Emission Tomography (PET) generates large datasets requiring efficient reconstruction.

Purpose of the Study:

• To introduce and rigorously derive the Origin Ensemble (OE) algorithm for Minimum-Mean-Square-Error (MMSE) reconstruction.
• To investigate the application of OE for List-Mode Time-of-Flight (LM-TOF) PET reconstruction.
• To compare the performance of OE with standard ML-EM in terms of image quality and reconstruction time.

Main Methods:

• Developed a statistically rigorous derivation of the OE algorithm for MMSE reconstruction.
• Incorporated Time-of-Flight (TOF) information and corrections for random and scattered events into the OE algorithm.
• Applied MMSE-OE and ML-EM to LM-TOF phantom data and analyzed convergence, reconstruction time, and image quality.

Main Results:

• MMSE-OE demonstrated a gradual decrease in background variability (BV) with increasing iterations, while contrast recovery (CRV) remained stable.
• Final MMSE-OE images showed lower BV and slightly lower CRV compared to ML-EM.
• OE algorithm reconstruction time was approximately 1.3 times longer than ML-EM.

Conclusions:

• The OE algorithm is a feasible and promising approach for LM-TOF PET reconstruction, offering improved image quality (lower BV).
• OE inherently provides comprehensive statistical characterization of data, beneficial for kinetic analysis and image registration.
• Despite longer reconstruction times, the OE algorithm's advantages in data characterization support its potential in various medical imaging applications.