Data-driven, energy-based method for estimation of scattered events in positron emission tomography

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Physics

Background:

• Scattered events introduce significant bias in Positron Emission Tomography (PET) image reconstruction.
• Accurate scatter distribution estimation is crucial for effective scatter correction in PET imaging.

Purpose of the Study:

• To propose and validate a practical energy-based (EB) scatter estimation method for accurate scatter correction in PET.
• To compare the performance of the EB method against single scatter simulation (SSS) and assess its applicability on real-world scanners.

Main Methods:

• Developed an energy-based (EB) scatter estimation technique utilizing the distinct energy distributions of scattered, unscattered, and random events.
• Modeled unscattered events using measured point source data, differentiating from prior EB methods.
• Validated the EB method using Monte Carlo simulations and experimental data from the PennPET EXPLORER scanner.

Main Results:

• EB scatter estimated sinograms showed excellent agreement with ground truth in simulations across various phantoms.
• Both EB and SSS methods yielded good contrast recovery on the NEMA IQ phantom, but EB provided superior lung residual correction.
• On larger phantoms (35 cm × 30 cm), EB correction reduced background variability by 12%-36% due to better handling of multiple scattering and out-of-FOV activity.

Conclusions:

• The proposed energy-based (EB) scatter estimation method is feasible and effectively translates to real-world PET scanners.
• EB offers significant efficiency advantages over SSS, with estimation times in seconds versus minutes.
• The EB method provides more uniform PET images, particularly beneficial for large-volume scanners and complex scattering scenarios.