Radiobiological dosimetry using hybrid imaging–derived spatiotemporal dose distributions for 131 I therapy of metastatic thyroid cancer

OBJECTIVE: To develop a patient-specific radiobiological dosimetry framework that incorporates tempora dose delivery and spatial heterogeneity using hybrid imaging for 131I therapy of metastatic thyroid cancer. APPROACH: Four-dimensional particle source distributions were reconstructed for eight adults from one quantitative pretherapy single-photon emission computed tomography/computed tomography acquisition and three serial planar images. An in-house Monte Carlo engine transported the sources through each patient's computed tomography geometry and scored absorbed dose in 1-h bins for 23 lesions. Voxelwise Lea-Catcheside repair factors and biologically effective dose (BED) were calculated from the temporal dose profiles, followed by lesion-level equivalent uniform BED (EUBED) and tumor control probability (TCP). The administered activity required to achieve 50% TCP (TCA50) was calculated under no, partial, and complete sublethal damage repair. Tumor repopulation was evaluated separately, and sensitivity analysis of the radiobiological parameters assessed their effects on TCA50. MAIN RESULTS: Predicted tumor control varied substantially across and within patients and depended on lesion-specific dose distributions and the assumed radiobiological parameters. Of the parameters varied from 0.5 to 1.5 times baseline, the linear radiosensitivity coefficient α had the greatest effect on TCA50. Median TCA50 values were 6.99, 14.38, and 15.02 GBq under no, partial, and complete repair, respectively. Small G-factors (median, 0.0133) resulted in partial-repair TCP curves and TCA50 estimates close to those under complete repair. Increasing activity amplified spatial BED heterogeneity and produced sublinear increases in EUBED, indicating that prescriptions based only on mean absorbed dose or mean BED may overestimate treatment effectiveness. Temporal TCP trajectories revealed changes in predicted control during tumor repopulation that were not captured by endpoint TCP alone. SIGNIFICANCE: The proposed framework enables patient-specific voxelwise radiobiological modeling through TCP using one quantitative SPECT/CT acquisition and serial planar imaging. It may support biologically informed activity prescription, although its model-dependent predictions require clinical validation.