Effect of inter-crystal scatter on estimation methods for random coincidences and subsequent correction

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Positron Emission Tomography (PET)

Background:

• Random coincidences significantly contribute to background noise in PET imaging.
• Accurate correction of random coincidences is crucial for quantitative PET analysis.
• Existing estimation methods for random coincidences require validation.

Purpose of the Study:

• To investigate the validity of random coincidence estimation techniques.
• To evaluate the impact of various low-energy thresholds (LETs) on estimation accuracy.
• To assess the influence of inter-crystal scatter on random coincidence overestimation.

Main Methods:

• Simulated singles list-mode data from the MADPET-II PET scanner using the GATE toolkit.
• Evaluation of random event estimation using delayed window (DW), singles rate (SR), and time histogram fitting (TH) methods.
• Comparison of estimated random coincidences with GATE simulation ground truth.

Main Results:

• DW and SR methods consistently overestimated random events, especially at lower LETs.
• Overestimation decreased with LETs above 255 keV and when Compton-scattered events were removed.
• Inter-crystal scatter was identified as the cause of overestimation, leading to activity underestimation in reconstructed images.
• Activity underestimation ranged from 1.7% (DW at 200 keV LET) to 7% (SR at 200 keV LET).

Conclusions:

• Standard methods like DW and SR overestimate random coincidences in PET, primarily due to inter-crystal scatter.
• Higher LETs and removal of Compton-scattered events mitigate this overestimation.
• Compensating for random events during reconstruction yields more accurate quantitative PET images than precorrected data.