A novel iterative method for lesion delineation and volumetric quantification with FDG PET

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Radiochemistry

Background:

• Positron Emission Tomography (PET) imaging faces challenges in accurately determining lesion boundaries due to blurring and unknown activity concentration.
• Standard threshold-based volumetric quantification methods are often size-dependent and influenced by signal-to-background ratio (SBR).

Purpose of the Study:

• To develop and validate a novel, model-based method for lesion delineation and volumetric quantification on FDG PET.
• To establish threshold levels that are independent of SBR and accurate for small lesions.
• To compare the performance of the novel method against standard techniques.

Main Methods:

• A background-subtracted relative threshold level (RTL) method was derived using a convolution model of the point-spread function and spherical lesions.
• Validation involved PET imaging of a Jaszczak phantom with varying sphere sizes and SBRs.
• An iterative procedure was used for quantification, and the method's feasibility was tested in patients with liver metastases.

Main Results:

• Phantom data confirmed that the optimal RTL is sphere-size dependent but independent of SBR.
• Specific RTL values were identified for different sphere sizes (e.g., 40% for 15-60 mm, >50% for <12 mm).
• The RTL method demonstrated superior applicability to patient data compared to standard fixed-percentage thresholding.

Conclusions:

• The RTL method, utilizing an iterative procedure, provides optimal, SBR-independent threshold levels for PET lesion quantification.
• This method is effective in both phantom and patient studies, offering a promising tool for accurate lesion delineation.
• The RTL method enables volumetric quantification of PET lesions down to the scanner's resolution limits.