Insertion of synthetic lesions on patient data: a method for evaluating clinical performance differences between PET...

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Radiology

Background:

• Positron emission tomography (PET) scanner performance assessment is vital for efficient clinical practice.
• A simulation method for creating controlled ground truth in PET system performance assessment was previously introduced.
• This study aimed to validate this simulation method using patient data for clinical accuracy assessment.

Purpose of the Study:

• To validate a simulation method for assessing PET scanner performance using patient data.
• To demonstrate the method's relevance in evaluating PET scanner accuracy under clinical conditions.
• To compare the performance of two PET systems (Discovery-IQ and Discovery-MI) using synthetic lesions.

Main Methods:

• Twenty-four patients were divided into two Body Mass Index (BMI) groups.
• ¹⁸F-FDG PET scans were performed on two systems (DIQ and DMI) using clinical protocols.
• Sixty synthetic lesions were inserted into scans, and their detection rates were compared with natural lesions, alongside inter-reader agreement (IOA) and Standardized Uptake Value (SUV) metrics.

Main Results:

• The relative true positive rate (RTPR) for synthetic lesions closely matched that of natural lesions in both BMI groups (≤25 and >25).
• Detection rates using insertion of synthetic lesions (ISL) were consistent with estimates and natural lesion detection rates.
• Inter-observer agreement (IOA) exceeded 80% across all scenarios, and SUV metrics showed good agreement between synthetic and natural lesions.

Conclusions:

• Insertion of synthetic lesions (ISL) is a relevant method for evaluating performance differences between PET scanners.
• This technique allows for the creation of a controlled ground truth within realistic anatomical models.
• The validated method can be utilized to exploit PET scanner potential for clinical applications.