Page 103 - 89Zr-Immuno-PET:Towards a Clinical Tool to Guide Antibody-based Therapy in Cancer

Noise-induced variability of 89Zr-immuno-PET
reconstruction filter was applied, in line with the recommendation for multicenter Zr-89-immuno-PET studies (10).
Image generation and analysis
Using the original scans, VOIs were defined manually for liver, spleen, kidney, brain and lung (brain and lung on the low dose CT, liver, spleen, kidney on the PET image). In addition, fixed size VOIs of 8.6 and 2.9 mL were placed in the lumbar vertebrae to estimate bone marrow activity concentration (AC), and in the aortic arch to estimate blood pool AC, respectively. Tumor uptake was defined as focal uptake exceeding local background reported by the nuclear medicine physician. Tumors were manually delineated on the immuno-PET scan, using the low dose CT for anatomical reference, using in-house developed software (ACCURATE tool, developed by RB).
Original list mode data were split (all counts, including delayed): even 5 counts were placed in one data set (e.g. H1), while the odd counts were placed in
the second data set (e.g. H2), creating interleaved datasets that are count statistically independent of each other, while preserving identical scan conditions (e.g. patient movement) (Figure 1). Next, these two split data sets were reconstructed (including
scatter and attenuation correction).
To assess the noise induced variability of Zr-89-immuno-PET at an injected dose of 37 MBq74inj, the original 89Zr-antiCD20 dataset (74 MBq) was split in two equal parts (H1 and H2), which were reconstructed in count-reduced images of 37 MBq74inj. To assess whether noise induced variability is independent of the investigated mAb, the original 89Zr-antiEGFR mAb dataset (37 MBq) and 89Zr- antiCD44 mAb dataset (37 MBq) were used to produce count-reduced images of 18 MBq37inj (H1 and H2). In addition, the 89Zr-antiCD20 mAb dataset (74 MBq) was split again (H1) resulting in count-reduced images at 25% of the injected dose (18 MBq74inj) (Q1 and Q2). To keep the statistical analysis similar for all three datasets, we did not split and analyze the second dataset (H2) (effectively creating Q3, Q4) as the 89Zr-antiEGFR mAb and 89Zr-antiCD44 mAb datasets only had the availability of two splits (H1, H2).
All tissue and tumor VOIs were applied to the count-reduced images, resulting in AC1 from the first count-reduced image (H1) and AC2 from the second count-reduced image (H2). For the 89Zr-antiCD20 mAb dataset all VOI were also applied to the count-reduced images at 25% of the injected dose (Q1 and Q2).
101