A Generalized and Efficient Microdosimetry Framework to Support Accurate Alpha-Emitter Radiobiology
Abstract
Purpose
Targeted alpha therapy (TAT) produces fundamentally different dose-response behavior than beta-emitting therapy because energy deposition occurs from rare, short-ranged, high-LET particle traversals. Under typical TAT administered activities, many cells receive no direct hits while a small fraction receive very large depositions, making mean absorbed dose formalisms (e.g. MIRD S-values) poorly representative. Averaging this heterogeneity can obscure dose-effect relationships and complicate interpretations of non-targeted phenomena such as bystander signaling. Direct Monte Carlo microdosimetry can capture this structure but is computationally prohibitive for large cell clusters. To address these shortcomings, we developed an efficient framework that preserves Monte Carlo-derived microdosimetric fidelity while enabling stochastic modeling of large monolayers and spheroids.
Methods
Precomputed Geant4 specific-energy spectra were used in an analytical simulation to generate per-cell dose probability distributions in large clusters. Validation against MIRDCell used matched geometries and membrane/cytoplasm activity distributions by comparing dose-PDF expectation values with S-values and dose–volume histograms (DVHs). The platform was applied to PC3-PIP and 22Rv1 clonogenic assays (n=6/condition) exposed to free 225Ac (0-0.0814-0.251-0.814 kBq/mL) for 96 h. A dose-probability-weighted linear model was fit to survival data to quantify predictive performance.
Results
Agreement with MIRDCell S-values was 2.2 ± 2.7% (cytoplasm) and 2.0 ± 0.2% (membrane) for 5-9 µm cell radii at 0-50 µm separations; for 515-cell spheroids, DVH dose-at-volume agreed within 2.0 ± 1.4% (cytoplasm) and 1.5 ± 0.3% (membrane). Predicted zero–nuclear dose fractions tracked survival across activity conditions: PC3-PIP zero-dose fractions 79.60%/49.50%/10.18% versus survivals 85.2%/61.04%/20.2% (r=0.998), and 22Rv1 zero-dose fractions 69.3%/32.3%/2.6% versus survivals 70.1%/30.1%/7.8% (r=0.999). Survival prediction was accurate within error bounds with RMSE 0.14 (95% CI: 0.12–0.16) for PC3-PIP and 0.10 (95% CI: 0.11–0.13) for 22Rv1.
Conclusion
The platform reproduces conventional dosimetry metrics while quantifying stochastic dose heterogeneity that is not accessible with S-value methods, providing accessible microdosimetry for use in understanding TAT radiobiology.