Personalised Dosimetry in Nuclear Medicine: Bridging Physics, Biology and AI for Next Generation Radiopharmaceutical Therapy

Radiopharmaceutical dosimetry is rapidly evolving from a physics-dominated calculation tool to a central pillar of precision nuclear medicine. As targeted radionuclide therapies expand across indications, there is a growing clinical imperative to personalize dose estimation, predict therapeutic efficacy and mitigate organ toxicity. This review critically examines the current landscape of dosimetry methods including organ level Medical Internal Radiation Dose (MIRD) schema, voxel-based S-values, Monte Carlo (MC) simulations and emerging artificial intelligence (AI)-assisted segmentation tools and their translational relevance. Through a comprehensive literature search of 177Lu peptide receptor radio nuclide therapy (PRRT) studies published between 2020 and 2025, we evaluate methodological heterogeneity and quantify dose variations across organs. Findings reveal persistent inconsistencies in absorbed dose estimates with reported kidney doses varying, 0.3–0.9 Gy/GBq and tumor doses ranging 1–10 Gy/GBq largely driven by differences in imaging protocol timing, segmentation strategy, and time-point sampling across studies. We also discuss regulatory trends, biologically informed dosimetry models incorporating relative biological effectiveness (RBE), and future integration with dose-point kernel (DPK) based and dose-volume histogram (DVH) driven computational frameworks. The field must now shift toward harmonized, reproducible standards that bridge physics, biology, and computation, transforming dosimetry into a predictive engine for individualized radiopharmaceutical therapy (RPT).