Feasibility of Automated Image-Based Red Bone Marrow Dosimetry for [Lu]Lu-PSMA Radiopharmaceutical Therapy of Metastatic Castration-Resistant Prostate Cancer.

: Red bone marrow irradiation is a major concern for patients with advanced prostate cancer undergoing [Lu]Lu-PSMA therapy. However, low uptake in the red bone marrow and the presence of bone lesions complicate image-based red bone marrow dosimetry. This study aimed to investigate the general feasibility of image-based red bone marrow activity estimation for [Lu]Lu-PSMA treatment and to develop a fully automated workflow for clinical implementation. : In the first part of the study, 175 virtual patient phantoms with realistic Lu activity distributions were generated based on 639 pre-therapeutic [F]F-PSMA-1007 PET/CT scans. The SIMIND Monte Carlo tool was used to simulate the Lu SPECT acquisitions (24 h post-injection (p.i.)), which were used to assess the uncertainty of red bone marrow activity estimation. In the second part, red bone marrow self- and cross-absorbed doses were estimated for four therapy cycles of 20 patients. : The simulation study shows a significant overestimation of activity in skeletal sites with bone lesions, with median recovery coefficients (RCs) across all phantoms yielding a median of 225% (range: 106-1015%). In contrast, the median RCs were markedly lower in skeletal sites neighboring or distant to lesion-carrying sites (105% [72-163%] and 107% [77-130%], respectively). The median total absorbed dose to the red bone marrow was 20.8 mGy/GBq (range: 5.6-297.9 mGy/GBq). Median blood levels decreased with an increasing median cumulative total absorbed dose. : Reliable estimation of activity concentration in skeletal sites without bone lesion infiltration has been shown to be feasible. Based on this finding, an automated workflow for routine image-based red bone marrow dosimetry was developed.