Abstract
1441
Objectives In thoracic PET, due to respiratory motions, accurate lesion delineation is difficult. The aim of this study was to evaluate the accuracy of the AISM on GATE Monte-Carlo simulations of gated 18F-FDG PET scans, using the realistic 4D-NCAT numerical phantom.
Methods 48 respiratory gated scans (RGS) and 48 ungated scans (UngS), for a total of 80 thoracic and hepatic lesions, were simulated and reconstructed to closely reproduce clinical images. The lesions were segmented using a fixed 40% threshold of the lesion maximum uptake (Th40%) and the AISM. AISM is an extension of Daisne's method (Radiother Oncol, 2003) and is automatic, iterative and adapted to clinical gated data. We evaluated the lesion displacement volumes (DV) obtained by merging the volumes from each time frame of the RGS. DVAISM,RGS and DVAISM,UngS from RGS and UngS, and DVTh40% from UngS were compared to the DV computed from the reference lesion volumes and displacements.
Results Unlike AISM which is fully user independent, Th40% needed initial manual exclusion of noisy structures. DV accuracy was largely improved with the AISM on RGS (DVAISM,RGS median error = -9.7mL) compared to UngS with AISM (DVAISM,UngS median error = -21.5mL) or Th40% (DVTh40% median error = -19.6mL).
Conclusions The accuracy improvements in lesion delineation offered by AISM on RGS might have a significant impact when patient treatment is performed using ungated external radiotherapy.
- © 2009 by Society of Nuclear Medicine