Author: Nadège Anizan, David Broggio, Eric Chojnacki, Estelle Davesne, Didier Franck, Stéphanie Lamart, Alexandre Pignard 👨🔬
Affiliation: Institut Bergonié, Service de Physique Médicale, Autorité de Sûreté Nucléaire et de Radioprotection (ASNR), PSE-SANTE/SDOS/LEDI, Autorité de Sûreté Nucléaire et de Radioprotection (ASNR), PSE-SANTE/SDOS, Autorité de Sûreté Nucléaire et de Radioprotection (ASNR), PSN-RES/SEMIA/LSMA 🌍
Purpose: To develop an innovative method enabling a comprehensive uncertainty budget associated to personalized absorbed doses for 177Lu-PSMA therapy and providing probability distributions of all dosimetric quantities.
Methods: In Targeted Radionuclide Therapy (TRT), three measurable and input variables contribute to the uncertainties on the final dose estimate: the calibration factor of the gamma camera, the count rate inside the lesional or healthy tissue measured on patient’s images, and the anatomical volume of this tissue. Uncertainties associated with the input variables were evaluated notably by sensitivity analysis. A Bayesian Network (BN) was developed to propagate these uncertainties to the computed absorbed dose, using deterministic or probabilistic relationships between variables involved in the dosimetric workflow. The uncertainty on each dosimetric estimate was defined as the standard deviation of the corresponding probability distribution generated by the BN.
Results: The BN was validated using a realistic calibration factor (8.35 counts/s/MBq, uncertainty: 2%), and a fictive lesion (volume of 2.5 mL, uncertainty: 30%) characterized by a theoretical biokinetic model, with an uptake (half-life T1) and elimination phase (T2) for 177Lu-PSMA. This model enabled the determination of true count rate values at selected timepoints (4h, 24h and 6 days after injection). Obtained probability distributions provided estimates comparable with theoretical values, e.g., the lesion absorbed dose was 19.0 Gy (uncertainty: 26%) versus 20.6 Gy in theory, T1 was 2.2h (58%) versus 2.0h, T2 was 58h (14%) versus 58h.
Conclusion: An innovative BN has been developed to evaluate the uncertainties on personalized absorbed doses in TRT and was validated on a realistic example. The obtained result for this theoretical lesion showed that dose uncertainties can reach more than 25%. This BN illustrates the effectiveness of the method in TRT and is ready for application to the actual patient data of an ongoing multicentric study focused on retrospective dosimetry.