A proof-of-concept study of sub-resolution hepatic vasculature modeling to assess dose heterogeneity in transarterial radioembolization

OBJECTIVE: Patient-specific vasculature plays a critical role in the distribution of yttrium-90 (90Y) microspheres and the resulting absorbed dose (AD) in transarterial radioembolization (TARE). We present a novel framework for generating a patient-specific hepatic vasculature model to evaluate the impact of vascular heterogeneity on 90Y dose distributions. Subsequently, we investigated the feasibility of calibrating the model by using the pre-treatment [99mTc]Tc-MAA SPECT/CT study. APPROACH: A single patient receiving TARE for hepatocellular carcinoma (HCC) was used as proof-of-concept. A standard cone-beam CT angiography (CBCT-A) study was used to identify major hepatic arteries, which were segmented using automated image-filtering methods implemented in the Insight Toolkit. An expert radiologist manually contoured the liver, perfused target area (PTA), and tumor. Deformable image registration was applied to map patient-specific anatomy onto a reference liver mesh partitioned according to the Couinaud classification. Vasculature models were generated in two ways, each comprising hepatic arterial, portal venous, and hepatic venous trees: (i) a fully synthetic model constrained to the patient's liver contour, and (ii) a segmentation-seeded patient-specific model in which virtual vessels were algorithmically extended from the extracted arterial endpoints. A vascular heterogeneity model was introduced in (ii) to generate six arterial trees spanning conditions from highly heterogeneous to uniform vascularization across the PTA. A virtual 90Y administration of 2.5 GBq was simulated in each arterial tree to generate synthetic single-photon emission computed tomography (SPECT) images and Monte Carlo-based AD distributions. Furthermore, we used the generated vasculature models to generate a synthetic [99mTc]Tc-MAA SPECT, which was compared with the clinical [99mTc]Tc-MAA SPECT by using 3D gamma index analysis (10%/10mm). MAIN RESULTS: Analysis of dose-volume histograms and coefficients of variation demonstrated that distinct vascular architectures yield substantially different dose heterogeneity despite identical mean ADs for both normal tissue and tumor. Comparison between the synthetic and clinical [99mTc]Tc-MAA SPECT images showed a systematic dependence of gamma pass rates on α, with a maximum pass rate of 78.5%. SIGNIFICANCE: This proof-of-concept study highlights the importance of intra-organ distribution in TARE beyond the standard use of mean AD and demonstrated the feasibility of calibrating the vascular heterogeneity model from clinical imaging. Because [99mTc]Tc-MAA particles and iodinated contrast differ in hemodynamic behaviour, future studies should use multi-phase CBCT-A as the calibration reference.