[show abstract][hide abstract] ABSTRACT: Abstract
Objectives: One of the challenges in image-guided radiation therapy has been the search for a robust and reproducible method of delineation of Gross tumor volume from PET imaging in order to reduce the margins of the field treatments. Several segmentation methods have been proposed. In this study, we compared some of those segmentation algorithms on phantom images obtained in different PET centres using local clinical acquisition/reconstruction protocols.
Methods: The phantom defined in the IEC 61675-1 and NEMA NU2-2007 protocols was used. The phantom background was filled with 2.6 kBq/ml of 18F-FDG, which is close to background activity observed in clinical conditions. The spheres (10, 13, 17, 22, 28 and 37 mm of diameter) were filled with 3, 6 and 9 times the background activity. Acquisition and reconstruction settings were the ones used in clinical routine for whole body FDG PET-CT by each centre (10 centers, 6 different brand/model PET/CT cameras). The segmentation algorithms tested are of two types: thresholding methods (42 % of MAX, 50 % (MAX + BKG), Nestle, Biehl, Black, Jentzen) and variational methods (levelset, Geets). The segmentation algorithms have been implemented in a single dedicated software. To compare the methods, we calculated the root mean square (RMS) of the differences between theoretical and segmented diameters. This comparison was carried on in two steps: first we evaluated the effect of contrast at fixed spatial resolution (one center), then we assessed the effect of different resolutions with all centers and one contrast (contrast 9).
Results: For a fixed resolution, the best algorithm is Jentzen with a RMS of 0.67 mm. For a fixed contrast, the best algorithm is Black with a RMS of 1.01 mm.
Conclusions: The algorithms described by Black et al and Jentzen et al give excellent volume estimations for homogeneous spheres using local standard reconstructions from 10 Belgian PET/CT centers. These algorithms have yet to be tested on heterogeneous objects in simulated images and in vivo images
Journal of Nuclear Medicine 01/2012; 53. · 5.77 Impact Factor
[show abstract][hide abstract] ABSTRACT: Adequate patient selection and treatment planning is crucial for a safe and cost-effective administration of selective internal radiotherapy (SIRT) of malignant liver disease using 90Y-labelled microspheres. It requires the implementation of multimodality imaging, integrating metabolic, functional and structural characteristics. A multidisciplinary approach is a prerequisite for SIRT, bringing together the knowhow and expertise of radiologists, nuclear medicine physicians, medical physicists, imaging engineers, and radiotherapists. This review discusses the available radiologic (CT/MRI) and nuclear (SPECT/PET) imaging modalities and their specific utility in the different diagnostic phases related to SIRT: whole body and intrahepatic pre-treatment disease staging, CT and MRI-based angiography, liver-lung shunt assessment, treatment simulation, predictive dosimetry, post-treatment imaging, and SIRT response assessment.
Journal of Nuclear Medicine & Radiation Therapy. 01/2011; 2(113).