Impact of Bone Marrow Segmentation on Dose Estimates In High‑ Vs Low‑ Disease Burden Patients Treated with 177lu‑PSMA‑617
Abstract
Purpose
Bone marrow (BM) is an organ-at-risk for 177Lu-PSMA radiopharmaceutical therapy, but estimating the absorbed dose (AD) is challenged by the segmentation of disseminated BM. In this work, we evaluated the impact of two segmentation approaches on AD estimates in patients treated with 177Lu-PSMA-617.
Methods
We included 85 patients from the RESIST-PC and LUNAR phase II trials, representing a broad range of tumor burden and bone disease involvement (NCT03042312, NCT05496959). All patients underwent post-therapy quantitative SPECT/CT imaging 24h after the first treatment cycle, with additional scans acquired at 4h and >48h for those in the LUNAR cohort. Two BM segmentation algorithms were implemented: the first targeting lumbar vertebrae (LV) as a surrogate, the second targeting soft tissue within the whole skeleton (WS). In the first method, L1-L5 were manually contoured, excluding those with significant disease involvement. In the WS method, an in-house automated algorithm is used to produce a BM mask for the field of view. AD calculations were performed using the Hänscheid single time point approach or, when data permitted, a multiple time point SPECT/CT workflow (MIM SurePlan MRT). We then applied a one sample t-test on the absolute value of the normalized dose difference (ADD).
Results
Although the LV and WS approaches produced similar mean ADs (0.12 Gy/cycle, 0.11 Gy/cycle respectively), they yielded significant patient-level differences, with a mean ADD of 40%, or 0.06 ± 0.11 Gy (p<0.001).The direction and magnitude of the ADD varied across patients, without a bias toward either approach.
Conclusion
Estimated BM AD comparison shows significant patient-level differences attributable to the segmentation method (n=85 patients). These differences highlight the importance of selecting appropriate surrogates for BM estimation, which may depend on disease distribution. Further work will correlate BM AD with acute hematologic toxicities to investigate which method more closely relates to clinical outcomes.