Patient-Specific Liver Vasculature Phantoms for Y-90 Radioembolization Dosimetry
Abstract
Purpose
Current Y-90 radioembolization (RE) standard dosimetry assumes homogeneous microsphere distribution, ignoring patient-specific vasculature heterogeneity. We developed a framework to reproduce and augment patient-specific hepatic vascular models from cone-beam CT angiography (CBCT-A) and used them to simulate synthetic SPECT images that reflect vascular-driven microsphere deposition.
Methods
A hepatocellular carcinoma patient treated with RE in the right lobe was analyzed as a proof-of-concept. The hepatic arterial tree was segmented from the CBCT-A image using automated vesselness filtering in ITK. A sex-specific reference liver model partitioned into eight Couinaud segments was deformably registered to the patient's liver contour using ANTs symmetric normalization. Segmentation reference from the CBCT-A arterial tree seeded a 3D Voronoi partition of the perfused territory, within which virtual vasculature was generated. We introduced a heterogeneity parameter α to control terminal endpoint density, with extreme cases α = 1 and α = 0 yielding radius-weighted (heterogeneous) distributions, and volume-weighted (homogeneous) distributions, respectively. Six model variants were created for α∈{1.00,0.75,0.50,0.25,0.00} plus a synthetic baseline model. Perfusion of 2.5GBq was simulated to generate SPECT images for each model. Monte Carlo dose calculations were performed using terminal points as the source, sampling different in the normal tissue and tumor according to a tumor-to-normal ratio (TNR) of 2.17.
Results
Mean doses were stable across α: 67–69 Gy in the perfusion target; 187–198 Gy in the tumor. Maximum dose decreased with decreasing α, from 3.56×10³ Gy (α = 1.00) to 6.80×10² Gy (α = 0.00) in the perfusion target. Dose standard deviation decreased monotonically with α, reflecting reduced spatial heterogeneity.
Conclusion
Patient-specific vasculature geometry directly impacts Y-90 dose heterogeneity. This framework provides a patient-specific way to study this effect, completing intrahepatic vascular trees from CBCT-A images, and paving the way for future calibratable digital twins for microsphere transport simulations and personalized RE treatment planning.