Abstract
2351
Introduction: Determination of organ and lesion absorbed doses (ADs) provide the potential to obtain dose-effect relationships that can lead to personalized radiopharmaceutical therapies (RPT). However, dosimetry procedures are not standard and variability of results between different institutions limits collection of data needed to standardize dosimetry. The 2021 SNMMI dosimetry challenge aimed at: 1) collecting data from the global nuclear medicine community to identify, quantify, and understand amount and sources of variability in dose calculations and 2) based on these data, make recommendations to reduce variability and harmonize and standardize dosimetry results and methods.
Methods: Data from 2 patients (PET, CT or MRI, 4 post-injection SPECT/CT and planar whole-body scans) who underwent RPT with 177Lu-DOTATATE were used. Participants performed AD calculations with the following input data sets provided for each of 5 tasks: 1) multiple SPECT/CT images, 2) 4 planar images, 3) 4 planar and 1 SPECT/CT image, 4) 4 SPECT/CT images plus 3D volumes of interests (VOIs), and 5) 3D VOIs plus a time-integrated activity (TIA) image. For both patients (Patient A and Patient B) ADs were requested for kidneys, liver, spleen, and 2 lesions for Patient A and 4 lesions for Patient B.
ADs were compared between tasks and variability of results was quantified using the coefficient of variation (COV). A Kruskal-Wallis test was used to determine whether the median ADs between tasks (1,2,3) and tasks (1,4,5) were equal. A Dunn’s test with adjusted p-values for multiple comparisons followed as a post-hoc test.
Results: Submissions from 18 countries were received corresponding to 120, 12, 64, 42, and 46 for tasks 1, 2, 3, 4, and 5, respectively. Largest COV of 196% was found for the left kidney AD of Patient A in task 2 (planar), with median dose of 0.6 Gy [0.2-16.7 Gy]. Smallest COV of 13% was found for the total kidney AD of Patient B in task 5, with median AD of 4.2 Gy [3.0 to 4.6 Gy]. In general, planar dosimetry had the highest COVs over all tasks and VOIs. The average COVs over all VOIs changed from 133±38% to 48±25% and from 72±21% to 30±20% between tasks 2 and 5, for Patient A and Patient B, respectively.
When comparing tasks (1,2,3) significant statistical differences (p<0.05) were found in ADs of all VOIs of Patient A, and for all VOIs except for the left kidney for Patient B. Dunn’s test revealed differences (p<0.05) between tasks 1-2 (liver, spleen, right kidney), tasks 1-3 (spleen, right kidney and both lesions), tasks 2-3 (spleen) of Patient A, while Patient B only showed differences for tasks 1-2 (liver, right kidney) and tasks 1-3 (right kidney).
For tasks (1,4,5) using exclusively SPECT/CT data, statistical difference was only found for Lesion1 of Patient A, indicating comparable median ADs between tasks. For Patient B differences were found for all lesions, right kidney and total kidneys. Dunn’s test only showed differences for Lesion1 for tasks 1-4 and 1-5 for Patient A. For Patient B, differences were found only for the lesions of tasks 1-4 and 1-5. Task 1 vs. 4 average COVs over all VOIs reduced from 51±14% to 37±16% and from 62±24% to 29±10% for patients A and B, respectively.
Conclusions: The results of this preliminary analysis suggest that dosimetry based on SPECT produces less variability in ADs than planar dosimetry. The results of tasks 1 vs. 4 suggest that segmentation played a large role in the reduction of variability in ADs. However, when the VOIs were provided the variability was not completely eliminated; other factors in the dose calculation method still contribute to variability in ADs, although to a lower extent. Consequently, providing recommendations to standardize segmentation could be a first step towards harmonizing dosimetry calculations. Analyses are ongoing to further reveal magnitude and variations for individual steps in the dosimetry workflow.