Change of SUV and metabolic volume products after use of rigid coregistration algorithms in F18-FDG-PET


Change of SUV and metabolic volume products after use of rigid coregistration algorithms in F18-FDG-PET

Steffen, I. G.; Hofheinz, F.; Grosser, O. S.; Furth, C.; Denecke, T.; Plotkin, M.; Amthauer, H.; Ruf, J.

Aim: Image fusion of anatomical and functional data is an established procedure in nuclear medicine. Whereas elastic image fusion is expected to alter the information of the transformed data, we assessed whether rigid transformations also change the information contained in the PET-data.
Methods: F-18-FDG-PET/CT (Biograph 16, Siemens Medical, Erlangen, Germany) data of 14 tumor patients with a total of 18 solid pulmonary lesions were included in this retrospective analysis. For coregistration, rigid fast (RF) and rigid slow (RS) transformation algorithms of the Fusion7d-software were employed (Fusion7d, Mirada Solutions Ltd./Siemens Medical, Erlangen, Germany). Original PET-images (voxel size: 0.417cm x 0.417cm x 0.5cm) were fused with the attenuation-corrected CT (LR, voxel size: 0.417cm x 0.417cm x 0.5cm) and diagnostic CT (HR, voxel size: 0.098cm x 0.098cm x 0.1cm).PET data were saved in the respective CT-geometry and CT-resolution. Segmentation of lesions was performed using a 3D ROI volume determination software with automatic background detection (Rover, ABX GmbH, Radeberg, Germany). SUVmax, SUVmean, volume and metabolic volumes (volume*SUVmax and volume*SUVmean) of the lesions were determined in the original and in the coregistered PET data and compared. SUVmax in original PET ranged from to 1.6-30.9 (mean±SD, 8.8±7.0). CT-segmented volumes ranged from 1-58 ml (mean±SD, 9.9±15 ml).
Results: The relative differences of volume between original data and rigid fast fused-data ranged from 2-45% (mean±SD, 19%±13%) for LR-coregistration and from 0-12% (mean±SD, 3%±3%) for HR-coregistration. Fusing data with the rigid slow algorithm resulted in relative volume differences ranging from 1-48% (mean±SD, 20%±18%) for LR-coregistration and from 0-14% (mean±SD, 4%±4%) for HR-coregistration. Analyzing relative differences of metabolic volume products (volume*SUVmax) between original data and rigid fast fused-data ranged from 0-33% (mean±SD, 12%±9%) for LR-coregistration and from 0-11% (mean±SD, 3%±3%) for HR-coregistration. After fusion with the rigid slow method differences of volume*SUVmax ranged from 0-41% (mean±SD, 13%±13%) for LR-coregistration and from 0-16% (mean±SD, 4%±5%).Conclusions:Transformation of PET data by rigid coregistration algorithms results in a redistribution of the respective voxel information in the new coordinate system. Volumetric analysis based on a source-to-background algorithm showed substantial differences between original and coregistered data, which may have a potential impact on e.g. PET-based planning of radiotherapy or the assessment of treatment response by measurement of metabolic burden in follow-up studies. This effect is dependent on the source data (CT) resolution. The extent of the observed effect on the data of modern high resolution PET-systems remains to be evaluated.

  • Poster
    Annual Congress of the European Association of Nuclear Mediciene (EANM) 2011, 15.-19.10.2011, Birmingham, UK
  • Abstract in refereed journal
    European Journal of Nuclear Medicine and Molecular Imaging 38(2011), S269

Permalink: https://www.hzdr.de/publications/Publ-16702
Publ.-Id: 16702