Assessing margin expansions of internal target volumes in 3D and 4D PET: a phantom study

Shyam S. Jani, James M. Lamb, Benjamin M. White, Magnus Dahlbom, Clifford G. Robinson, Daniel A. Low

Research output: Contribution to journalArticlepeer-review

3 Scopus citations

Abstract

Background and Purpose: To quantify tumor volume coverage and excess normal tissue coverage using margin expansions of mobile target internal target volumes (ITVs) in the lung.

Materials and methods: FDG-PET list-mode data were acquired for four spheres ranging from 1 to 4 cm as they underwent 1D motion based on four patient breathing trajectories. Both ungated PET images and PET maximum intensity projections (PET-MIPs) were examined. Amplitude-based gating was performed on sequential list-mode files of varying signal-to-background ratios to generate PET-MIPs. ITVs were first post-processed using either a Gaussian filter or a custom two-step module, and then segmented by applying a gradient-based watershed algorithm. Uniform and non-uniform 1 mm margins were added to segmented ITVs until complete target coverage was achieved.

Results: PET-MIPs required smaller uniform margins (4.7 vs. 11.3 mm) than ungated PET, with correspondingly smaller over-coverage volumes (OCVs). Non-uniform margins consistently resulted in smaller OCVs when compared to uniform margins. PET-MIPs and ungated PET had comparable OCVs with non-uniform margins, but PET-MIPs required smaller longitudinal margins (4.7 vs. 8.5 mm). Non-uniform margins were independent of sphere size.

Conclusions: Gated PET-MIP images and non-uniform margins result in more accurate ITV delineation while reducing normal tissue coverage.

Original languageEnglish
Pages (from-to)100-109
Number of pages10
JournalAnnals of Nuclear Medicine
Volume29
Issue number1
DOIs
StatePublished - Jan 1 2015

Keywords

  • 4D-PET
  • Lung cancer

Fingerprint

Dive into the research topics of 'Assessing margin expansions of internal target volumes in 3D and 4D PET: a phantom study'. Together they form a unique fingerprint.

Cite this