Using statistical measures for automated comparison of in-beam PET data

Area of Science:

• Medical Physics
• Nuclear Medicine
• Radiotherapy

Background:

• Positron emission tomography (PET) is the gold standard for in vivo particle therapy monitoring.
• Current range assessment relies on time-consuming visual inspection by experts.
• There is a need for automated, objective methods to assess particle beam delivery.

Purpose of the Study:

• To develop and evaluate an automated approach for objective comparison of PET images in particle therapy.
• To support clinical decisions in particle therapy monitoring by analyzing beam delivery accuracy.
• To assess the efficacy of Pearson's correlation coefficient (PCC) for detecting range deviations.

Main Methods:

• Utilized 12 in-beam PET patient datasets with 70 artificial range modifications (0-10 mm WEPL).
• Calculated Pearson's correlation coefficient (PCC) between measured and simulated/reference PET images.
• Determined sensitivity and specificity of PCC for detecting ion beam range deviations.

Main Results:

• Test 1 (vs. simulation): >90% sensitivity/specificity for ±10 and ±6 mm WEPL modifications.
• Test 2 (vs. measured PET): >80% sensitivity/specificity for ±10 and -6 mm WEPL modifications.
• Method's detection limit was approximately 4 mm WEPL.

Conclusions:

• Automated PCC comparison shows comparable sensitivity and specificity to expert visual inspection.
• The proposed automated method is a promising tool to enhance efficiency in particle therapy monitoring.
• This approach can improve the clinical workflow for in-beam PET data analysis.