Page 128 - 89Zr-Immuno-PET:Towards a Clinical Tool to Guide Antibody-based Therapy in Cancer
P. 128

                                Chapter 6
settings was allowed: zoom, contrast and orientation (coronal/axial/sagittal). Use of a threshold (upper or lower limit) or fixed size VOI was not allowed. For 89Zr- rituximab and 89Zr-cetuximab, tumors were manually delineated on both the D3 and D6 scans, starting with the latest time point. On D0, no tumor uptake was visible, therefore the VOI delineated on D6 were imported to the D0 scan. Observers could manually adjust localization of the VOI to optimize matching of the anatomical position of the tumor lesion on the D0 scan.
For all VOI, max, peak and mean activity concentrations (AC in Bq/mL) were derived and converted to standardized uptake values (SUV), by correcting for body weight and injected dose (ID). In addition, delineated volume (mL) and TLU (defined as ACmean * volume, in %ID) were obtained.
Manual tumor delineation on immuno-PET after viewing the 18F-FDG-PET. In order to support delineation of the tumor, the observers had access to the corresponding 18F-FDG-PET and could adapt the original manually delineated VOI if necessary (for example by creating a smaller or larger VOI, or changing the position of the VOI) (Figure 2b). This procedure was performed on scans with visible tumor uptake (D3, D4, D6). The number of VOI that were adapted after viewing the 18F-FDG-PET was obtained.
Semi-automatic VOI delineation. Finally, we investigated the feasibility of a mask- restricted semi-automatic VOI delineation method. Each observer, for every tumor lesion, manually delineated a mask, which is a VOI including the tumor, excluding non-tumor structures (e.g. nearby blood vessels) on the immuno-PET scan. Subsequently, the semi-automatic VOI was generated including all voxels with a value ≥50% of background-corrected ACpeak within the mask (Figure 2c). The semi-automatic isocontour was defined as 0.5 * (peak value + average background value). The background region was determined with a region growing algorithm of the tumour border, expanding three voxels away from the border of the tumour in all three dimensions (10). The observers rated the semi-automatic VOI and accepted the VOI if it contained the tumor and no other structures or background. The number of tumor lesions for which the semi-automatic VOI was accepted by all observers was obtained.
126





























































































   126   127   128   129   130