Strengths and weaknesses of energy-based scatter estimation using three basis functions

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Computational Physics

Background:

• Accurate scatter correction is crucial for quantitative positron emission tomography (PET) imaging.
• Conventional tail-fitted single-scatter simulation (SSS) has limitations in accuracy due to attenuation image mismatches and out-of-field-of-view (FOV) activity.
• Energy-based scatter estimation (EBSE) offers a promising alternative to address these limitations.

Purpose of the Study:

• To enhance the accuracy of EBSE by incorporating line-of-response (LOR) dependent energy spectra of unscattered photons.
• To improve the computational speed of EBSE through optimized initialization and a more efficient algorithm.
• To evaluate the performance and stability of the improved EBSE method.

Main Methods:

• Developed an improved EBSE method modeling energy probability density functions (PDFs) for scattered and unscattered photons, including a position-dependent local energy PDF.
• Utilized two forward models based on 2D energy histograms for scatter estimation.
• Evaluated performance using GATE Monte Carlo simulations and NEMA phantom data on a GE SIGNA PET/MR scanner, assessing stability via data mashing.

Main Results:

• The improved EBSE method demonstrated superior performance over tail-fitted SSS, especially near out-of-FOV activity.
• Incorporating a local energy PDF for unscattered photons improved off-center quantification by approximately 2%.
• Optimized initialization and the NEGML optimizer reduced EBSE iterations from 200 to 50, achieving computation in 12 minutes on a six-core CPU.

Conclusions:

• The proposed EBSE method significantly enhances both accuracy and computational efficiency for scatter correction in PET.
• This advancement provides a more robust and faster approach for quantitative PET imaging.
• The study clarifies limitations of previous scattered basis function approaches.