Page 138 - 89Zr-Immuno-PET:Towards a Clinical Tool to Guide Antibody-based Therapy in Cancer

Chapter 6
delineation was explored. For 18F-FDG-PET with perfect interobserver agreement for SUVmax (13) and higher tumor contrast, semi-automatic procedures are used to obtain SUVmean based on a semi-automatic method (e.g. with a threshold of 0.6 of SUVmax), total lesion glycolysis (TLG) and total metabolic tumor volume (TMVT) (14, 15). For our datasets, the area included by the semi-automatic VOI was often too large, indicating low tumor to local background ratios, resulting in inclusion of background voxels in the semi-automatic VOI.
For mAbs showing higher tumor contrast, as well as imaging with higher count statistics (due to e.g. higher injected doses or the availability of scanners with improved detection sensitivity or time of flight resolution), semi-automatic delineation may be feasible. Reduction of noise (e.g. by introduction of total body PET scanners) is the first step towards further improvement of tumor delineation procedures. Future studies into accuracy of tumor delineation should include ‘supervised’ delineation methods (semi-automatic procedures with a manual check) in which the optimal threshold is experimentally determined. If the success rate can thus be increased, this may lead to further development towards a robust automatic method, which is desired for clinical application. As semi-automatic delineation was not feasible in our datasets, we explored eligibility criteria to improve standardization for manual tumor delineation, especially in case of limited tumor contrast.
In our study, 81% of the VOI (84 out of 103) were considered suitable for quantification. Based on these results, we recommend a two-step procedure to exclude lesions with insufficient tumor contrast for manual delineation: 1) Verification of VOI delineation by a nuclear medicine physician to identify delineation of an incorrect structure due to limited tumor contrast, 2) Exclusion of VOI with the voxel with the highest uptake located at the border of the VOI, indicating low tumor uptake and/or high background uptake. These measures support optimal scan interpretation and standardization, which is an essential step towards potential clinical implementation of 89Zr-immuno-PET.
For this study, we performed a multicenter interobserver analysis for data that was originally obtained in single center studies. With this experience, the next step towards standardization of quantification for 89Zr-immuno-PET studies can be done in the context of a multicenter study (e.g. the IMPACT trials, (NCT02228954, NCT02117466 and NCT01957332)).
Reliable delineation of tumor uptake on 89Zr-immuno-PET allows future use as a non-invasive clinical tool to determine mAb concentrations in the tumor.
136