Segmentation of PET volumes by iterative image thresholding

Area of Science:

• Nuclear Medicine
• Medical Imaging Analysis

Background:

• Positron Emission Tomography (PET) volume segmentation typically relies on thresholding methods.
• Clinical applications often require anatomical images like CT for accurate thresholding, limiting PET-only analysis.

Purpose of the Study:

• To develop and validate an iterative thresholding method (ITM) for estimating PET lesion volumes.
• To assess the ITM's accuracy without requiring a priori anatomical information from CT scans.

Main Methods:

• Developed an ITM based on source-to-background (S/B) ratio threshold-volume curves derived from phantom studies.
• Validated the ITM using PET phantoms and 39 clinical PET/CT tumor datasets ((18)F-FDG and (124)I).
• Iteratively calculated lesion volumes from PET images and compared them to gold standards (phantom sphere volumes, CT-derived tumor volumes).

Main Results:

• ITM showed similar S/B-threshold-volume curves for (18)F-FDG and (124)I.
• Average absolute deviation was ~10% for volumes >1.0 mL in phantoms and ~9% for patient lesions between 0.8-7.5 mL.
• Larger lesions (>7.5 mL) exhibited greater deviations (~15%), which were still considered acceptable.

Conclusions:

• The ITM provides sufficient accuracy for estimating clinical PET volumes, particularly in the 0.8-7.5 mL range.
• The method is especially valuable for lesions detectable only by PET, facilitating accurate lesion dosimetry.
• ITM enables lesion dosimetry using PET images alone, reducing reliance on multi-modal imaging.