A Study on Tumor Boron Dose Evaluation Using a Pharmacokinetic Model Based on 18F-BPA Dynamic Positron Emission Tomography (PET) Imaging In Boron Neutron Capture Therapy (BNCT)
Abstract
Purpose
In BNCT, 18F-BPA-PET evaluates boron drug distribution for treatment eligibility, but is rarely incorporated into treatment planning. A key limitation is differences in drug administration protocols between PET and BNCT, which may weaken clinical outcome-to-dose associations. This study aimed to improve boron dose calculation in treatment planning by incorporating heterogeneous boron distribution using a voxel-wise pharmacokinetic model.
Methods
Dynamic 18F-BPA-PET from six head and neck cancer patients was voxelized (4 mm). A three-compartment model with an image-derived input function obtained from the carotid VOI was fitted in each tumoral voxel to derive rate constants (K1–k4). Model-estimated voxel-wise SUVs were compared with 1-h static PET SUVs after bolus injection using MAE and correlation coefficients. For clinical BNCT, tumoral voxel-wise boron concentrations under 2-h continuous infusion were estimated and used for boron dose calculations. Estimated boron concentrations were compared with values derived from 2-h static PET SUVs. Resulting doses were also compared with a conventional treatment planning method assuming a constant tumor-to-blood boron concentration ratio under a 1-h neutron irradiation using CICS-1. Blood boron concentration was set to 25 ppm after 2 h continuous infusion based on published pharmacokinetic data, and CBE was set to 4.0. Statistical analysis used the Wilcoxon signed-rank test.
Results
Estimated SUVs strongly correlated with 1-h static PET SUVs (r = 0.938), with a median MAE of 0.527. Estimated tumoral boron concentrations under 2-h continuous infusion differed from those derived from 2-h static PET SUVs (p < 0.05) but remained strongly correlated (r = 0.857). Median tumoral voxel-wise boron dose (5583 voxels) across six cases was 242.6 Gy-Eq for the conventional method and 102.4 Gy-Eq for the model (p < 0.05).
Conclusion
This voxel-wise pharmacokinetic approach enables protocol-consistent estimation of heterogeneous boron distribution for treatment planning. It may improve dose evaluation accuracy and support personalized BNCT.